fix: pass terminal_lifetime through to Modal sandbox timeout

The lifetime_seconds config was computed and stored as an env var but never reached Modal's Sandbox.create(timeout=...), causing sandboxes to always die after the hardcoded 1h default regardless of configuration.
fix: use hermes agent system prompt and nudges
2026-04-13 11:57:39 -07:00 · 2026-04-08 18:34:01 -07:00 · 2026-04-08 18:34:01 -07:00 · 2026-04-08 18:33:02 -07:00 · 2026-04-08 18:33:02 -07:00 · 2026-04-08 18:33:02 -07:00
1220 changed files with 26562 additions and 223406 deletions
@@ -5,7 +5,6 @@

 # Dependencies
 node_modules
-.venv

 # CI/CD
 .github
@@ -24,15 +24,6 @@
 # Optional base URL override (default: Google's OpenAI-compatible endpoint)
 # GEMINI_BASE_URL=https://generativelanguage.googleapis.com/v1beta/openai

-# =============================================================================
-# LLM PROVIDER (Ollama Cloud)
-# =============================================================================
-# Cloud-hosted open models via Ollama's OpenAI-compatible endpoint.
-# Get your key at: https://ollama.com/settings
-# OLLAMA_API_KEY=your_ollama_key_here
-# Optional base URL override (default: https://ollama.com/v1)
-# OLLAMA_BASE_URL=https://ollama.com/v1
-
 # =============================================================================
 # LLM PROVIDER (z.ai / GLM)
 # =============================================================================
@@ -52,15 +43,6 @@
 # KIMI_BASE_URL=https://api.kimi.com/coding/v1  # Default for sk-kimi- keys
 # KIMI_BASE_URL=https://api.moonshot.ai/v1      # For legacy Moonshot keys
 # KIMI_BASE_URL=https://api.moonshot.cn/v1       # For Moonshot China keys
-# KIMI_CN_API_KEY=                               # Dedicated Moonshot China key
-
-# =============================================================================
-# LLM PROVIDER (Arcee AI)
-# =============================================================================
-# Arcee AI provides access to Trinity models (trinity-mini, trinity-large-*)
-# Get an Arcee key at: https://chat.arcee.ai/
-# ARCEEAI_API_KEY=
-# ARCEE_BASE_URL=                                 # Override default base URL

 # =============================================================================
 # LLM PROVIDER (MiniMax)
@@ -107,15 +89,6 @@
 # Optional base URL override:
 # HERMES_QWEN_BASE_URL=https://portal.qwen.ai/v1

-# =============================================================================
-# LLM PROVIDER (Xiaomi MiMo)
-# =============================================================================
-# Xiaomi MiMo models (mimo-v2-pro, mimo-v2-omni, mimo-v2-flash).
-# Get your key at: https://platform.xiaomimimo.com
-# XIAOMI_API_KEY=your_key_here
-# Optional base URL override:
-# XIAOMI_BASE_URL=https://api.xiaomimimo.com/v1
-
 # =============================================================================
 # TOOL API KEYS
 # =============================================================================
@@ -154,10 +127,6 @@
 # Only override here if you need to force a backend without touching config.yaml:
 # TERMINAL_ENV=local

-# Override the container runtime binary (e.g. to use Podman instead of Docker).
-# Useful on systems where Docker's storage driver is broken or unavailable.
-# HERMES_DOCKER_BINARY=/usr/local/bin/podman
-
 # Container images (for singularity/docker/modal backends)
 # TERMINAL_DOCKER_IMAGE=nikolaik/python-nodejs:python3.11-nodejs20
 # TERMINAL_SINGULARITY_IMAGE=docker://nikolaik/python-nodejs:python3.11-nodejs20
@@ -1,5 +1 @@
-watch_file pyproject.toml uv.lock
-watch_file ui-tui/package-lock.json ui-tui/package.json
-watch_file flake.nix flake.lock nix/devShell.nix nix/tui.nix nix/package.nix nix/python.nix
-
 use flake
@@ -1,2 +0,0 @@
-# Auto-generated files — collapse diffs and exclude from language stats
-web/package-lock.json linguist-generated=true
@@ -11,7 +11,6 @@ body:
        **Before submitting**, please:
        - [ ] Search [existing issues](https://github.com/NousResearch/hermes-agent/issues) to avoid duplicates
        - [ ] Update to the latest version (`hermes update`) and confirm the bug still exists
-        - [ ] Run `hermes debug share` and paste the links below (see Debug Report section)

  - type: textarea
    id: description
@@ -83,25 +82,6 @@ body:
        - Slack
        - WhatsApp

-  - type: textarea
-    id: debug-report
-    attributes:
-      label: Debug Report
-      description: |
-        Run `hermes debug share` from your terminal and paste the links it prints here.
-        This uploads your system info, config, and recent logs to a paste service automatically.
-
-        If you're in an interactive chat session, you can also use the `/debug` slash command — it does the same thing.
-
-        If the upload fails, run `hermes debug share --local` and paste the output directly.
-      placeholder: |
-        Report   https://paste.rs/abc123
-        agent.log   https://paste.rs/def456
-        gateway.log   https://paste.rs/ghi789
-      render: shell
-    validations:
-      required: true
-
  - type: input
    id: os
    attributes:
@@ -117,6 +97,8 @@ body:
      label: Python Version
      description: Output of `python --version`
      placeholder: "3.11.9"
+    validations:
+      required: true

  - type: input
    id: hermes-version
@@ -124,14 +106,14 @@ body:
      label: Hermes Version
      description: Output of `hermes version`
      placeholder: "2.1.0"
+    validations:
+      required: true

  - type: textarea
    id: logs
    attributes:
-      label: Additional Logs / Traceback (optional)
-      description: |
-        The debug report above covers most logs. Use this field for any extra error output, 
-        tracebacks, or screenshots not captured by `hermes debug share`.
+      label: Relevant Logs / Traceback
+      description: Paste any error output, traceback, or log messages. This will be auto-formatted as code.
      render: shell

  - type: textarea
@@ -71,15 +71,3 @@ body:
      label: Contribution
      options:
        - label: I'd like to implement this myself and submit a PR
-
-  - type: textarea
-    id: debug-report
-    attributes:
-      label: Debug Report (optional)
-      description: |
-        If this feature request is related to a problem you're experiencing, run `hermes debug share` and paste the links here.
-        In an interactive chat session, you can use `/debug` instead.
-        This helps us understand your environment and any related logs.
-      placeholder: |
-        Report   https://paste.rs/abc123
-      render: shell
@@ -9,8 +9,7 @@ body:
        Sorry you're having trouble! Please fill out the details below so we can help.

        **Quick checks first:**
-        - Run `hermes debug share` and paste the links in the Debug Report section below
-        - If you're in a chat session, you can use `/debug` instead — it does the same thing
+        - Run `hermes doctor` and include the output below
        - Try `hermes update` to get the latest version
        - Check the [README troubleshooting section](https://github.com/NousResearch/hermes-agent#troubleshooting)
        - For general questions, consider the [Nous Research Discord](https://discord.gg/NousResearch) for faster help
@@ -75,21 +74,10 @@ body:
      placeholder: "2.1.0"

  - type: textarea
-    id: debug-report
+    id: doctor-output
    attributes:
-      label: Debug Report
-      description: |
-        Run `hermes debug share` from your terminal and paste the links it prints here.
-        This uploads your system info, config, and recent logs to a paste service automatically.
-
-        If you're in an interactive chat session, you can also use the `/debug` slash command — it does the same thing.
-
-        If the upload fails or install didn't get that far, run `hermes debug share --local` and paste the output directly.
-        If even that doesn't work, run `hermes doctor` and paste that output instead.
-      placeholder: |
-        Report   https://paste.rs/abc123
-        agent.log   https://paste.rs/def456
-        gateway.log   https://paste.rs/ghi789
+      label: Output of `hermes doctor`
+      description: Run `hermes doctor` and paste the full output. This will be auto-formatted.
      render: shell

  - type: textarea
@@ -1,73 +0,0 @@
-name: Contributor Attribution Check
-
-on:
-  pull_request:
-    branches: [main]
-    paths:
-      # Only run when code files change (not docs-only PRs)
-      - '*.py'
-      - '**/*.py'
-      - '.github/workflows/contributor-check.yml'
-
-permissions:
-  contents: read
-
-jobs:
-  check-attribution:
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5  # v4
-        with:
-          fetch-depth: 0  # Full history needed for git log
-
-      - name: Check for unmapped contributor emails
-        run: |
-          # Get the merge base between this PR and main
-          MERGE_BASE=$(git merge-base origin/main HEAD)
-
-          # Find any new author emails in this PR's commits
-          NEW_EMAILS=$(git log ${MERGE_BASE}..HEAD --format='%ae' --no-merges | sort -u)
-
-          if [ -z "$NEW_EMAILS" ]; then
-            echo "No new commits to check."
-            exit 0
-          fi
-
-          # Check each email against AUTHOR_MAP in release.py
-          MISSING=""
-          while IFS= read -r email; do
-            # Skip teknium and bot emails
-            case "$email" in
-              *teknium*|*noreply@github.com*|*dependabot*|*github-actions*|*anthropic.com*|*cursor.com*)
-                continue ;;
-            esac
-
-            # Check if email is in AUTHOR_MAP (either as a key or matches noreply pattern)
-            if echo "$email" | grep -qP '\+.*@users\.noreply\.github\.com'; then
-              continue  # GitHub noreply emails auto-resolve
-            fi
-
-            if ! grep -qF "\"${email}\"" scripts/release.py 2>/dev/null; then
-              AUTHOR=$(git log --author="$email" --format='%an' -1)
-              MISSING="${MISSING}\n  ${email} (${AUTHOR})"
-            fi
-          done <<< "$NEW_EMAILS"
-
-          if [ -n "$MISSING" ]; then
-            echo ""
-            echo "⚠️  New contributor email(s) not in AUTHOR_MAP:"
-            echo -e "$MISSING"
-            echo ""
-            echo "Please add mappings to scripts/release.py AUTHOR_MAP:"
-            echo -e "$MISSING" | while read -r line; do
-              email=$(echo "$line" | sed 's/^ *//' | cut -d' ' -f1)
-              [ -z "$email" ] && continue
-              echo "    \"${email}\": \"<github-username>\","
-            done
-            echo ""
-            echo "To find the GitHub username for an email:"
-            echo "  gh api 'search/users?q=EMAIL+in:email' --jq '.items[0].login'"
-            exit 1
-          else
-            echo "✅ All contributor emails are mapped in AUTHOR_MAP."
-          fi
@@ -1,12 +1,11 @@
 name: Deploy Site

 on:
-  release:
-    types: [published]
  push:
    branches: [main]
    paths:
      - 'website/**'
+      - 'landingpage/**'
      - 'skills/**'
      - 'optional-skills/**'
      - '.github/workflows/deploy-site.yml'
@@ -21,46 +20,32 @@ concurrency:
  cancel-in-progress: false

 jobs:
-  deploy-vercel:
-    if: github.event_name == 'release'
-    runs-on: ubuntu-latest
-    steps:
-      - name: Trigger Vercel Deploy
-        run: curl -X POST "${{ secrets.VERCEL_DEPLOY_HOOK }}"
-
-  deploy-docs:
+  build-and-deploy:
+    # Only run on the upstream repository, not on forks
    if: github.repository == 'NousResearch/hermes-agent'
    runs-on: ubuntu-latest
    environment:
      name: github-pages
      url: ${{ steps.deploy.outputs.page_url }}
    steps:
-      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5  # v4
+      - uses: actions/checkout@v4

-      - uses: actions/setup-node@49933ea5288caeca8642d1e84afbd3f7d6820020  # v4
+      - uses: actions/setup-node@v4
        with:
          node-version: 20
          cache: npm
          cache-dependency-path: website/package-lock.json

-      - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065  # v5
+      - uses: actions/setup-python@v5
        with:
          python-version: '3.11'

      - name: Install PyYAML for skill extraction
-        run: pip install pyyaml==6.0.2 httpx==0.28.1
+        run: pip install pyyaml

      - name: Extract skill metadata for dashboard
        run: python3 website/scripts/extract-skills.py

-      - name: Build skills index (if not already present)
-        env:
-          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        run: |
-          if [ ! -f website/static/api/skills-index.json ]; then
-            python3 scripts/build_skills_index.py || echo "Skills index build failed (non-fatal)"
-          fi
-
      - name: Install dependencies
        run: npm ci
        working-directory: website
@@ -72,13 +57,18 @@ jobs:
      - name: Stage deployment
        run: |
          mkdir -p _site/docs
+          # Landing page at root
+          cp -r landingpage/* _site/
+          # Docusaurus at /docs/
          cp -r website/build/* _site/docs/
+          # CNAME so GitHub Pages keeps the custom domain between deploys
+          echo "hermes-agent.nousresearch.com" > _site/CNAME

      - name: Upload artifact
-        uses: actions/upload-pages-artifact@56afc609e74202658d3ffba0e8f6dda462b719fa  # v3
+        uses: actions/upload-pages-artifact@v3
        with:
          path: _site

      - name: Deploy to GitHub Pages
        id: deploy
-        uses: actions/deploy-pages@d6db90164ac5ed86f2b6aed7e0febac5b3c0c03e  # v4
+        uses: actions/deploy-pages@v4
@@ -8,9 +8,6 @@ on:
  release:
    types: [published]

-permissions:
-  contents: read
-
 concurrency:
  group: docker-${{ github.ref }}
  cancel-in-progress: true
@@ -20,29 +17,22 @@ jobs:
    # Only run on the upstream repository, not on forks
    if: github.repository == 'NousResearch/hermes-agent'
    runs-on: ubuntu-latest
-    timeout-minutes: 60
+    timeout-minutes: 30
    steps:
      - name: Checkout code
-        uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5  # v4
+        uses: actions/checkout@v4
        with:
          submodules: recursive

-      - name: Set up QEMU
-        uses: docker/setup-qemu-action@c7c53464625b32c7a7e944ae62b3e17d2b600130  # v3
-
      - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@8d2750c68a42422c14e847fe6c8ac0403b4cbd6f  # v3
+        uses: docker/setup-buildx-action@v3

-      # Build amd64 only so we can `load` the image for smoke testing.
-      # `load: true` cannot export a multi-arch manifest to the local daemon.
-      # The multi-arch build follows on push to main / release.
-      - name: Build image (amd64, smoke test)
-        uses: docker/build-push-action@10e90e3645eae34f1e60eeb005ba3a3d33f178e8  # v6
+      - name: Build image
+        uses: docker/build-push-action@v6
        with:
          context: .
          file: Dockerfile
          load: true
-          platforms: linux/amd64
          tags: nousresearch/hermes-agent:test
          cache-from: type=gha
          cache-to: type=gha,mode=max
@@ -56,31 +46,34 @@ jobs:

      - name: Log in to Docker Hub
        if: github.event_name == 'push' && github.ref == 'refs/heads/main' || github.event_name == 'release'
-        uses: docker/login-action@c94ce9fb468520275223c153574b00df6fe4bcc9  # v3
+        uses: docker/login-action@v3
        with:
          username: ${{ secrets.DOCKERHUB_USERNAME }}
          password: ${{ secrets.DOCKERHUB_TOKEN }}

-      - name: Push multi-arch image (main branch)
+      - name: Push image (main branch)
        if: github.event_name == 'push' && github.ref == 'refs/heads/main'
-        uses: docker/build-push-action@10e90e3645eae34f1e60eeb005ba3a3d33f178e8  # v6
+        uses: docker/build-push-action@v6
        with:
          context: .
          file: Dockerfile
          push: true
-          platforms: linux/amd64,linux/arm64
-          tags: nousresearch/hermes-agent:latest
+          tags: |
+            nousresearch/hermes-agent:latest
+            nousresearch/hermes-agent:${{ github.sha }}
          cache-from: type=gha
          cache-to: type=gha,mode=max

-      - name: Push multi-arch image (release)
+      - name: Push image (release)
        if: github.event_name == 'release'
-        uses: docker/build-push-action@10e90e3645eae34f1e60eeb005ba3a3d33f178e8  # v6
+        uses: docker/build-push-action@v6
        with:
          context: .
          file: Dockerfile
          push: true
-          platforms: linux/amd64,linux/arm64
-          tags: nousresearch/hermes-agent:${{ github.event.release.tag_name }}
+          tags: |
+            nousresearch/hermes-agent:latest
+            nousresearch/hermes-agent:${{ github.event.release.tag_name }}
+            nousresearch/hermes-agent:${{ github.sha }}
          cache-from: type=gha
          cache-to: type=gha,mode=max
@@ -7,16 +7,13 @@ on:
      - '.github/workflows/docs-site-checks.yml'
  workflow_dispatch:

-permissions:
-  contents: read
-
 jobs:
  docs-site-checks:
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5  # v4
+      - uses: actions/checkout@v4

-      - uses: actions/setup-node@49933ea5288caeca8642d1e84afbd3f7d6820020  # v4
+      - uses: actions/setup-node@v4
        with:
          node-version: 20
          cache: npm
@@ -26,12 +23,12 @@ jobs:
        run: npm ci
        working-directory: website

-      - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065  # v5
+      - uses: actions/setup-python@v5
        with:
          python-version: '3.11'

-      - name: Install ascii-guard
-        run: python -m pip install ascii-guard==2.3.0 pyyaml==6.0.3
+      - name: Install Python dependencies
+        run: python -m pip install ascii-guard pyyaml

      - name: Extract skill metadata for dashboard
        run: python3 website/scripts/extract-skills.py
@@ -14,9 +14,6 @@ on:
      - 'run_agent.py'
      - 'acp_adapter/**'

-permissions:
-  contents: read
-
 concurrency:
  group: nix-${{ github.ref }}
  cancel-in-progress: true
@@ -29,9 +26,9 @@ jobs:
    runs-on: ${{ matrix.os }}
    timeout-minutes: 30
    steps:
-      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5  # v4
-      - uses: DeterminateSystems/nix-installer-action@ef8a148080ab6020fd15196c2084a2eea5ff2d25  # v22
-      - uses: DeterminateSystems/magic-nix-cache-action@565684385bcd71bad329742eefe8d12f2e765b39  # v13
+      - uses: actions/checkout@v4
+      - uses: DeterminateSystems/nix-installer-action@main
+      - uses: DeterminateSystems/magic-nix-cache-action@main
      - name: Check flake
        if: runner.os == 'Linux'
        run: nix flake check --print-build-logs
@@ -1,101 +0,0 @@
-name: Build Skills Index
-
-on:
-  schedule:
-    # Run twice daily: 6 AM and 6 PM UTC
-    - cron: '0 6,18 * * *'
-  workflow_dispatch:  # Manual trigger
-  push:
-    branches: [main]
-    paths:
-      - 'scripts/build_skills_index.py'
-      - '.github/workflows/skills-index.yml'
-
-permissions:
-  contents: read
-
-jobs:
-  build-index:
-    # Only run on the upstream repository, not on forks
-    if: github.repository == 'NousResearch/hermes-agent'
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5  # v4
-
-      - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065  # v5
-        with:
-          python-version: '3.11'
-
-      - name: Install dependencies
-        run: pip install httpx==0.28.1 pyyaml==6.0.2
-
-      - name: Build skills index
-        env:
-          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        run: python scripts/build_skills_index.py
-
-      - name: Upload index artifact
-        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02  # v4
-        with:
-          name: skills-index
-          path: website/static/api/skills-index.json
-          retention-days: 7
-
-  deploy-with-index:
-    needs: build-index
-    runs-on: ubuntu-latest
-    permissions:
-      pages: write
-      id-token: write
-    environment:
-      name: github-pages
-      url: ${{ steps.deploy.outputs.page_url }}
-    # Only deploy on schedule or manual trigger (not on every push to the script)
-    if: github.event_name == 'schedule' || github.event_name == 'workflow_dispatch'
-    steps:
-      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5  # v4
-
-      - uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093  # v4
-        with:
-          name: skills-index
-          path: website/static/api/
-
-      - uses: actions/setup-node@49933ea5288caeca8642d1e84afbd3f7d6820020  # v4
-        with:
-          node-version: 20
-          cache: npm
-          cache-dependency-path: website/package-lock.json
-
-      - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065  # v5
-        with:
-          python-version: '3.11'
-
-      - name: Install PyYAML for skill extraction
-        run: pip install pyyaml==6.0.2
-
-      - name: Extract skill metadata for dashboard
-        run: python3 website/scripts/extract-skills.py
-
-      - name: Install dependencies
-        run: npm ci
-        working-directory: website
-
-      - name: Build Docusaurus
-        run: npm run build
-        working-directory: website
-
-      - name: Stage deployment
-        run: |
-          mkdir -p _site/docs
-          cp -r landingpage/* _site/
-          cp -r website/build/* _site/docs/
-          echo "hermes-agent.nousresearch.com" > _site/CNAME
-
-      - name: Upload artifact
-        uses: actions/upload-pages-artifact@56afc609e74202658d3ffba0e8f6dda462b719fa  # v3
-        with:
-          path: _site
-
-      - name: Deploy to GitHub Pages
-        id: deploy
-        uses: actions/deploy-pages@d6db90164ac5ed86f2b6aed7e0febac5b3c0c03e  # v4
@@ -14,7 +14,7 @@ jobs:
    runs-on: ubuntu-latest
    steps:
      - name: Checkout
-        uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5  # v4
+        uses: actions/checkout@v4
        with:
          fetch-depth: 0

@@ -149,62 +149,6 @@ jobs:
          "
          fi

-          # --- CI/CD workflow files modified ---
-          WORKFLOW_HITS=$(git diff --name-only "$BASE".."$HEAD" | grep -E '\.github/workflows/.*\.ya?ml$' || true)
-          if [ -n "$WORKFLOW_HITS" ]; then
-            FINDINGS="${FINDINGS}
-          ### ⚠️ WARNING: CI/CD workflow files modified
-          Changes to workflow files can alter build pipelines, inject steps, or modify permissions. Verify no unauthorized actions or secrets access were added.
-
-          **Files:**
-          \`\`\`
-          ${WORKFLOW_HITS}
-          \`\`\`
-          "
-          fi
-
-          # --- Dockerfile / container build files modified ---
-          DOCKER_HITS=$(git diff --name-only "$BASE".."$HEAD" | grep -iE '(Dockerfile|\.dockerignore|docker-compose)' || true)
-          if [ -n "$DOCKER_HITS" ]; then
-            FINDINGS="${FINDINGS}
-          ### ⚠️ WARNING: Container build files modified
-          Changes to Dockerfiles or compose files can alter base images, add build steps, or expose ports. Verify base image pins and build commands.
-
-          **Files:**
-          \`\`\`
-          ${DOCKER_HITS}
-          \`\`\`
-          "
-          fi
-
-          # --- Dependency manifest files modified ---
-          DEP_HITS=$(git diff --name-only "$BASE".."$HEAD" | grep -E '(pyproject\.toml|requirements.*\.txt|package\.json|Gemfile|go\.mod|Cargo\.toml)$' || true)
-          if [ -n "$DEP_HITS" ]; then
-            FINDINGS="${FINDINGS}
-          ### ⚠️ WARNING: Dependency manifest files modified
-          Changes to dependency files can introduce new packages or change version pins. Verify all dependency changes are intentional and from trusted sources.
-
-          **Files:**
-          \`\`\`
-          ${DEP_HITS}
-          \`\`\`
-          "
-          fi
-
-          # --- GitHub Actions version unpinning (mutable tags instead of SHAs) ---
-          ACTIONS_UNPIN=$(echo "$DIFF" | grep -n '^\+' | grep 'uses:' | grep -v '#' | grep -E '@v[0-9]' | head -10 || true)
-          if [ -n "$ACTIONS_UNPIN" ]; then
-            FINDINGS="${FINDINGS}
-          ### ⚠️ WARNING: GitHub Actions with mutable version tags
-          Actions should be pinned to full commit SHAs (not \`@v4\`, \`@v5\`). Mutable tags can be retargeted silently if a maintainer account is compromised.
-
-          **Matches:**
-          \`\`\`
-          ${ACTIONS_UNPIN}
-          \`\`\`
-          "
-          fi
-
          # --- Output results ---
          if [ -n "$FINDINGS" ]; then
            echo "found=true" >> "$GITHUB_OUTPUT"
@@ -239,7 +183,7 @@ jobs:
          ---
          *Automated scan triggered by [supply-chain-audit](/.github/workflows/supply-chain-audit.yml). If this is a false positive, a maintainer can approve after manual review.*"

-          gh pr comment "${{ github.event.pull_request.number }}" --body "$BODY" || echo "::warning::Could not post PR comment (expected for fork PRs — GITHUB_TOKEN is read-only)"
+          gh pr comment "${{ github.event.pull_request.number }}" --body "$BODY"

      - name: Fail on critical findings
        if: steps.scan.outputs.critical == 'true'
@@ -6,9 +6,6 @@ on:
  pull_request:
    branches: [main]

-permissions:
-  contents: read
-
 # Cancel in-progress runs for the same PR/branch
 concurrency:
  group: tests-${{ github.ref }}
@@ -20,13 +17,13 @@ jobs:
    timeout-minutes: 10
    steps:
      - name: Checkout code
-        uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5  # v4
+        uses: actions/checkout@v4

      - name: Install system dependencies
        run: sudo apt-get update && sudo apt-get install -y ripgrep

      - name: Install uv
-        uses: astral-sh/setup-uv@d4b2f3b6ecc6e67c4457f6d3e41ec42d3d0fcb86  # v5
+        uses: astral-sh/setup-uv@v5

      - name: Set up Python 3.11
        run: uv python install 3.11
@@ -52,10 +49,10 @@ jobs:
    timeout-minutes: 10
    steps:
      - name: Checkout code
-        uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5  # v4
+        uses: actions/checkout@v4

      - name: Install uv
-        uses: astral-sh/setup-uv@d4b2f3b6ecc6e67c4457f6d3e41ec42d3d0fcb86  # v5
+        uses: astral-sh/setup-uv@v5

      - name: Set up Python 3.11
        run: uv python install 3.11
@@ -51,15 +51,10 @@ ignored/
 .worktrees/
 environments/benchmarks/evals/

-# Web UI build output
-hermes_cli/web_dist/
-
 # Release script temp files
 .release_notes.md
 mini-swe-agent/

 # Nix
 .direnv/
-.nix-stamps/
 result
-website/static/api/skills-index.json
@@ -1,108 +0,0 @@
-# .mailmap — canonical author mapping for git shortlog / git log / GitHub
-# Format: Canonical Name <canonical@email> <commit@email>
-# See: https://git-scm.com/docs/gitmailmap
-#
-# This maps commit emails to GitHub noreply addresses so that:
-# 1. `git shortlog -sn` shows deduplicated contributor counts
-# 2. GitHub's contributor graph can attribute commits correctly
-# 3. Contributors with personal/work emails get proper credit
-#
-# When adding entries: use the contributor's GitHub noreply email as canonical
-# so GitHub can link commits to their profile.
-
-# === Teknium (multiple emails) ===
-Teknium <127238744+teknium1@users.noreply.github.com> <teknium1@gmail.com>
-Teknium <127238744+teknium1@users.noreply.github.com> <teknium@nousresearch.com>
-
-# === Contributors — personal/work emails mapped to GitHub noreply ===
-# Format: Canonical Name <GH-noreply> <commit-email>
-
-# Verified via GH API email search
-luyao618 <364939526@qq.com> <364939526@qq.com>
-ethernet8023 <arilotter@gmail.com> <arilotter@gmail.com>
-nicoloboschi <boschi1997@gmail.com> <boschi1997@gmail.com>
-cherifya <chef.ya@gmail.com> <chef.ya@gmail.com>
-BongSuCHOI <chlqhdtn98@gmail.com> <chlqhdtn98@gmail.com>
-dsocolobsky <dsocolobsky@gmail.com> <dsocolobsky@gmail.com>
-pefontana <fontana.pedro93@gmail.com> <fontana.pedro93@gmail.com>
-Helmi <frank@helmschrott.de> <frank@helmschrott.de>
-hata1234 <hata1234@gmail.com> <hata1234@gmail.com>
-
-# Verified via PR investigation / salvage PR bodies
-DeployFaith <agents@kylefrench.dev> <agents@kylefrench.dev>
-flobo3 <floptopbot33@gmail.com> <floptopbot33@gmail.com>
-gaixianggeng <gaixg94@gmail.com> <gaixg94@gmail.com>
-KUSH42 <xush@xush.org> <xush@xush.org>
-konsisumer <der@konsi.org> <der@konsi.org>
-WorldInnovationsDepartment <vorvul.danylo@gmail.com> <vorvul.danylo@gmail.com>
-m0n5t3r <iacobs@m0n5t3r.info> <iacobs@m0n5t3r.info>
-sprmn24 <oncuevtv@gmail.com> <oncuevtv@gmail.com>
-fancydirty <fancydirty@gmail.com> <fancydirty@gmail.com>
-fxfitz <francis.x.fitzpatrick@gmail.com> <francis.x.fitzpatrick@gmail.com>
-limars874 <limars874@gmail.com> <limars874@gmail.com>
-AaronWong1999 <aaronwong1999@icloud.com> <aaronwong1999@icloud.com>
-dippwho <dipp.who@gmail.com> <dipp.who@gmail.com>
-duerzy <duerzy@gmail.com> <duerzy@gmail.com>
-geoffwellman <geoff.wellman@gmail.com> <geoff.wellman@gmail.com>
-hcshen0111 <shenhaocheng19990111@gmail.com> <shenhaocheng19990111@gmail.com>
-jamesarch <han.shan@live.cn> <han.shan@live.cn>
-stephenschoettler <stephenschoettler@gmail.com> <stephenschoettler@gmail.com>
-Tranquil-Flow <tranquil_flow@protonmail.com> <tranquil_flow@protonmail.com>
-Dusk1e <yusufalweshdemir@gmail.com> <yusufalweshdemir@gmail.com>
-Awsh1 <ysfalweshcan@gmail.com> <ysfalweshcan@gmail.com>
-WAXLYY <ysfwaxlycan@gmail.com> <ysfwaxlycan@gmail.com>
-donrhmexe <don.rhm@gmail.com> <don.rhm@gmail.com>
-hqhq1025 <1506751656@qq.com> <1506751656@qq.com>
-BlackishGreen33 <s5460703@gmail.com> <s5460703@gmail.com>
-tomqiaozc <zqiao@microsoft.com> <zqiao@microsoft.com>
-MagicRay1217 <mingjwan@microsoft.com> <mingjwan@microsoft.com>
-aaronagent <1115117931@qq.com> <1115117931@qq.com>
-YoungYang963 <young@YoungdeMacBook-Pro.local> <young@YoungdeMacBook-Pro.local>
-LongOddCode <haolong@microsoft.com> <haolong@microsoft.com>
-Cafexss <coffeemjj@gmail.com> <coffeemjj@gmail.com>
-Cygra <sjtuwbh@gmail.com> <sjtuwbh@gmail.com>
-DomGrieco <dgrieco@redhat.com> <dgrieco@redhat.com>
-
-# Duplicate email mapping (same person, multiple emails)
-Sertug17 <104278804+Sertug17@users.noreply.github.com> <srhtsrht17@gmail.com>
-yyovil <birdiegyal@gmail.com> <tanishq231003@gmail.com>
-DomGrieco <dgrieco@redhat.com> <dgrieco@redhat.com>
-dsocolobsky <dsocolobsky@gmail.com> <dylan.socolobsky@lambdaclass.com>
-olafthiele <programming@olafthiele.com> <olafthiele@gmail.com>
-
-# Verified via git display name matching GH contributor username
-cokemine <aptx4561@gmail.com> <aptx4561@gmail.com>
-dalianmao000 <dalianmao0107@gmail.com> <dalianmao0107@gmail.com>
-emozilla <emozilla@nousresearch.com> <emozilla@nousresearch.com>
-jjovalle99 <juan.ovalle@mistral.ai> <juan.ovalle@mistral.ai>
-kagura-agent <kagura.chen28@gmail.com> <kagura.chen28@gmail.com>
-spniyant <niyant@spicefi.xyz> <niyant@spicefi.xyz>
-olafthiele <programming@olafthiele.com> <programming@olafthiele.com>
-r266-tech <r2668940489@gmail.com> <r2668940489@gmail.com>
-xingkongliang <tianliangjay@gmail.com> <tianliangjay@gmail.com>
-win4r <win4r@outlook.com> <win4r@outlook.com>
-zhouboli <zhouboli@gmail.com> <zhouboli@gmail.com>
-yongtenglei <yongtenglei@gmail.com> <yongtenglei@gmail.com>
-
-# Nous Research team
-benbarclay <ben@nousresearch.com> <ben@nousresearch.com>
-jquesnelle <jonny@nousresearch.com> <jonny@nousresearch.com>
-
-# GH contributor list verified
-spideystreet <dhicham.pro@gmail.com> <dhicham.pro@gmail.com>
-dorukardahan <dorukardahan@hotmail.com> <dorukardahan@hotmail.com>
-MustafaKara7 <karamusti912@gmail.com> <karamusti912@gmail.com>
-Hmbown <hmbown@gmail.com> <hmbown@gmail.com>
-kamil-gwozdz <kamil@gwozdz.me> <kamil@gwozdz.me>
-kira-ariaki <kira@ariaki.me> <kira@ariaki.me>
-knopki <knopki@duck.com> <knopki@duck.com>
-Unayung <unayung@gmail.com> <unayung@gmail.com>
-SeeYangZhi <yangzhi.see@gmail.com> <yangzhi.see@gmail.com>
-Julientalbot <julien.talbot@ergonomia.re> <julien.talbot@ergonomia.re>
-lesterli <lisicheng168@gmail.com> <lisicheng168@gmail.com>
-JiayuuWang <jiayuw794@gmail.com> <jiayuw794@gmail.com>
-tesseracttars-creator <tesseracttars@gmail.com> <tesseracttars@gmail.com>
-xinbenlv <zzn+pa@zzn.im> <zzn+pa@zzn.im>
-SaulJWu <saul.jj.wu@gmail.com> <saul.jj.wu@gmail.com>
-angelos <angelos@oikos.lan.home.malaiwah.com> <angelos@oikos.lan.home.malaiwah.com>
-MestreY0d4-Uninter <241404605+MestreY0d4-Uninter@users.noreply.github.com> <MestreY0d4-Uninter@users.noreply.github.com>
@@ -13,7 +13,7 @@ source venv/bin/activate  # ALWAYS activate before running Python
 ```
 hermes-agent/
 ├── run_agent.py          # AIAgent class — core conversation loop
-├── model_tools.py        # Tool orchestration, discover_builtin_tools(), handle_function_call()
+├── model_tools.py        # Tool orchestration, _discover_tools(), handle_function_call()
 ├── toolsets.py           # Toolset definitions, _HERMES_CORE_TOOLS list
 ├── cli.py                # HermesCLI class — interactive CLI orchestrator
 ├── hermes_state.py       # SessionDB — SQLite session store (FTS5 search)
@@ -55,20 +55,7 @@ hermes-agent/
 ├── gateway/              # Messaging platform gateway
 │   ├── run.py            # Main loop, slash commands, message dispatch
 │   ├── session.py        # SessionStore — conversation persistence
-│   └── platforms/        # Adapters: telegram, discord, slack, whatsapp, homeassistant, signal, qqbot
-├── ui-tui/               # Ink (React) terminal UI — `hermes --tui`
-│   ├── src/entry.tsx        # TTY gate + render()
-│   ├── src/app.tsx          # Main state machine and UI
-│   ├── src/gatewayClient.ts # Child process + JSON-RPC bridge
-│   ├── src/app/             # Decomposed app logic (event handler, slash handler, stores, hooks)
-│   ├── src/components/      # Ink components (branding, markdown, prompts, pickers, etc.)
-│   ├── src/hooks/           # useCompletion, useInputHistory, useQueue, useVirtualHistory
-│   └── src/lib/             # Pure helpers (history, osc52, text, rpc, messages)
-├── tui_gateway/          # Python JSON-RPC backend for the TUI
-│   ├── entry.py             # stdio entrypoint
-│   ├── server.py            # RPC handlers and session logic
-│   ├── render.py            # Optional rich/ANSI bridge
-│   └── slash_worker.py      # Persistent HermesCLI subprocess for slash commands
+│   └── platforms/        # Adapters: telegram, discord, slack, whatsapp, homeassistant, signal
 ├── acp_adapter/          # ACP server (VS Code / Zed / JetBrains integration)
 ├── cron/                 # Scheduler (jobs.py, scheduler.py)
 ├── environments/         # RL training environments (Atropos)
@@ -192,62 +179,9 @@ if canonical == "mycommand":

 ---

-## TUI Architecture (ui-tui + tui_gateway)
-
-The TUI is a full replacement for the classic (prompt_toolkit) CLI, activated via `hermes --tui` or `HERMES_TUI=1`.
-
-### Process Model
-
-```
-hermes --tui
-  └─ Node (Ink)  ──stdio JSON-RPC──  Python (tui_gateway)
-       │                                  └─ AIAgent + tools + sessions
-       └─ renders transcript, composer, prompts, activity
-```
-
-TypeScript owns the screen. Python owns sessions, tools, model calls, and slash command logic.
-
-### Transport
-
-Newline-delimited JSON-RPC over stdio. Requests from Ink, events from Python. See `tui_gateway/server.py` for the full method/event catalog.
-
-### Key Surfaces
-
-| Surface | Ink component | Gateway method |
-|---------|---------------|----------------|
-| Chat streaming | `app.tsx` + `messageLine.tsx` | `prompt.submit` → `message.delta/complete` |
-| Tool activity | `thinking.tsx` | `tool.start/progress/complete` |
-| Approvals | `prompts.tsx` | `approval.respond` ← `approval.request` |
-| Clarify/sudo/secret | `prompts.tsx`, `maskedPrompt.tsx` | `clarify/sudo/secret.respond` |
-| Session picker | `sessionPicker.tsx` | `session.list/resume` |
-| Slash commands | Local handler + fallthrough | `slash.exec` → `_SlashWorker`, `command.dispatch` |
-| Completions | `useCompletion` hook | `complete.slash`, `complete.path` |
-| Theming | `theme.ts` + `branding.tsx` | `gateway.ready` with skin data |
-
-### Slash Command Flow
-
-1. Built-in client commands (`/help`, `/quit`, `/clear`, `/resume`, `/copy`, `/paste`, etc.) handled locally in `app.tsx`
-2. Everything else → `slash.exec` (runs in persistent `_SlashWorker` subprocess) → `command.dispatch` fallback
-
-### Dev Commands
-
-```bash
-cd ui-tui
-npm install       # first time
-npm run dev       # watch mode (rebuilds hermes-ink + tsx --watch)
-npm start         # production
-npm run build     # full build (hermes-ink + tsc)
-npm run type-check # typecheck only (tsc --noEmit)
-npm run lint      # eslint
-npm run fmt       # prettier
-npm test          # vitest
-```
-
---
-
 ## Adding New Tools

-Requires changes in **2 files**:
+Requires changes in **3 files**:

 **1. Create `tools/your_tool.py`:**
 ```python
@@ -270,9 +204,9 @@ registry.register(
 )
 ```

-**2. Add to `toolsets.py`** — either `_HERMES_CORE_TOOLS` (all platforms) or a new toolset.
+**2. Add import** in `model_tools.py` `_discover_tools()` list.

-Auto-discovery: any `tools/*.py` file with a top-level `registry.register()` call is imported automatically — no manual import list to maintain.
+**3. Add to `toolsets.py`** — either `_HERMES_CORE_TOOLS` (all platforms) or a new toolset.

 The registry handles schema collection, dispatch, availability checking, and error wrapping. All handlers MUST return a JSON string.

@@ -417,9 +351,8 @@ Cache-breaking forces dramatically higher costs. The ONLY time we alter context

 ### Background Process Notifications (Gateway)

-When `terminal(background=true, notify_on_complete=true)` is used, the gateway runs a watcher that
-detects process completion and triggers a new agent turn. Control verbosity of background process
-messages with `display.background_process_notifications`
+When `terminal(background=true, check_interval=...)` is used, the gateway runs a watcher that
+pushes status updates to the user's chat. Control verbosity with `display.background_process_notifications`
 in config.yaml (or `HERMES_BACKGROUND_NOTIFICATIONS` env var):

 - `all` — running-output updates + final message (default)
@@ -524,45 +457,13 @@ def profile_env(tmp_path, monkeypatch):

 ## Testing

-**ALWAYS use `scripts/run_tests.sh`** — do not call `pytest` directly. The script enforces
-hermetic environment parity with CI (unset credential vars, TZ=UTC, LANG=C.UTF-8,
-4 xdist workers matching GHA ubuntu-latest). Direct `pytest` on a 16+ core
-developer machine with API keys set diverges from CI in ways that have caused
-multiple "works locally, fails in CI" incidents (and the reverse).
-
-```bash
-scripts/run_tests.sh                                  # full suite, CI-parity
-scripts/run_tests.sh tests/gateway/                   # one directory
-scripts/run_tests.sh tests/agent/test_foo.py::test_x  # one test
-scripts/run_tests.sh -v --tb=long                     # pass-through pytest flags
-```
-
-### Why the wrapper (and why the old "just call pytest" doesn't work)
-
-Five real sources of local-vs-CI drift the script closes:
-
-| | Without wrapper | With wrapper |
-|---|---|---|
-| Provider API keys | Whatever is in your env (auto-detects pool) | All `*_API_KEY`/`*_TOKEN`/etc. unset |
-| HOME / `~/.hermes/` | Your real config+auth.json | Temp dir per test |
-| Timezone | Local TZ (PDT etc.) | UTC |
-| Locale | Whatever is set | C.UTF-8 |
-| xdist workers | `-n auto` = all cores (20+ on a workstation) | `-n 4` matching CI |
-
-`tests/conftest.py` also enforces points 1-4 as an autouse fixture so ANY pytest
-invocation (including IDE integrations) gets hermetic behavior — but the wrapper
-is belt-and-suspenders.
-
-### Running without the wrapper (only if you must)
-
-If you can't use the wrapper (e.g. on Windows or inside an IDE that shells
-pytest directly), at minimum activate the venv and pass `-n 4`:
-
 ```bash
 source venv/bin/activate
-python -m pytest tests/ -q -n 4
+python -m pytest tests/ -q          # Full suite (~3000 tests, ~3 min)
+python -m pytest tests/test_model_tools.py -q   # Toolset resolution
+python -m pytest tests/test_cli_init.py -q       # CLI config loading
+python -m pytest tests/gateway/ -q               # Gateway tests
+python -m pytest tests/tools/ -q                 # Tool-level tests
 ```

-Worker count above 4 will surface test-ordering flakes that CI never sees.
-
 Always run the full suite before pushing changes.
@@ -1,56 +1,25 @@
-FROM ghcr.io/astral-sh/uv:0.11.6-python3.13-trixie@sha256:b3c543b6c4f23a5f2df22866bd7857e5d304b67a564f4feab6ac22044dde719b AS uv_source
-FROM tianon/gosu:1.19-trixie@sha256:3b176695959c71e123eb390d427efc665eeb561b1540e82679c15e992006b8b9 AS gosu_source
 FROM debian:13.4

-# Disable Python stdout buffering to ensure logs are printed immediately
-ENV PYTHONUNBUFFERED=1
-
-# Store Playwright browsers outside the volume mount so the build-time
-# install survives the /opt/data volume overlay at runtime.
-ENV PLAYWRIGHT_BROWSERS_PATH=/opt/hermes/.playwright
-
 # Install system dependencies in one layer, clear APT cache
 RUN apt-get update && \
    apt-get install -y --no-install-recommends \
-        build-essential nodejs npm python3 ripgrep ffmpeg gcc python3-dev libffi-dev procps git && \
+        build-essential nodejs npm python3 python3-pip ripgrep ffmpeg gcc python3-dev libffi-dev && \
    rm -rf /var/lib/apt/lists/*

-# Non-root user for runtime; UID can be overridden via HERMES_UID at runtime
-RUN useradd -u 10000 -m -d /opt/data hermes
-
-COPY --chmod=0755 --from=gosu_source /gosu /usr/local/bin/
-COPY --chmod=0755 --from=uv_source /usr/local/bin/uv /usr/local/bin/uvx /usr/local/bin/
-
+COPY . /opt/hermes
 WORKDIR /opt/hermes

-# ---------- Layer-cached dependency install ----------
-# Copy only package manifests first so npm install + Playwright are cached
-# unless the lockfiles themselves change.
-COPY package.json package-lock.json ./
-COPY scripts/whatsapp-bridge/package.json scripts/whatsapp-bridge/package-lock.json scripts/whatsapp-bridge/
-COPY web/package.json web/package-lock.json web/
-
-RUN npm install --prefer-offline --no-audit && \
+# Install Python and Node dependencies in one layer, no cache
+RUN pip install --no-cache-dir -e ".[all]" --break-system-packages && \
+    npm install --prefer-offline --no-audit && \
    npx playwright install --with-deps chromium --only-shell && \
-    (cd scripts/whatsapp-bridge && npm install --prefer-offline --no-audit) && \
-    (cd web && npm install --prefer-offline --no-audit) && \
+    cd /opt/hermes/scripts/whatsapp-bridge && \
+    npm install --prefer-offline --no-audit && \
    npm cache clean --force

-# ---------- Source code ----------
-# .dockerignore excludes node_modules, so the installs above survive.
-COPY --chown=hermes:hermes . .
+WORKDIR /opt/hermes
+RUN chmod +x /opt/hermes/docker/entrypoint.sh

-# Build web dashboard (Vite outputs to hermes_cli/web_dist/)
-RUN cd web && npm run build
-
-# ---------- Python virtualenv ----------
-RUN chown hermes:hermes /opt/hermes
-USER hermes
-RUN uv venv && \
-    uv pip install --no-cache-dir -e ".[all]"
-
-# ---------- Runtime ----------
-ENV HERMES_WEB_DIST=/opt/hermes/hermes_cli/web_dist
 ENV HERMES_HOME=/opt/data
 VOLUME [ "/opt/data" ]
 ENTRYPOINT [ "/opt/hermes/docker/entrypoint.sh" ]
@@ -13,7 +13,7 @@

 **The self-improving AI agent built by [Nous Research](https://nousresearch.com).** It's the only agent with a built-in learning loop — it creates skills from experience, improves them during use, nudges itself to persist knowledge, searches its own past conversations, and builds a deepening model of who you are across sessions. Run it on a $5 VPS, a GPU cluster, or serverless infrastructure that costs nearly nothing when idle. It's not tied to your laptop — talk to it from Telegram while it works on a cloud VM.

-Use any model you want — [Nous Portal](https://portal.nousresearch.com), [OpenRouter](https://openrouter.ai) (200+ models), [NVIDIA NIM](https://build.nvidia.com) (Nemotron), [Xiaomi MiMo](https://platform.xiaomimimo.com), [z.ai/GLM](https://z.ai), [Kimi/Moonshot](https://platform.moonshot.ai), [MiniMax](https://www.minimax.io), [Hugging Face](https://huggingface.co), OpenAI, or your own endpoint. Switch with `hermes model` — no code changes, no lock-in.
+Use any model you want — [Nous Portal](https://portal.nousresearch.com), [OpenRouter](https://openrouter.ai) (200+ models), [z.ai/GLM](https://z.ai), [Kimi/Moonshot](https://platform.moonshot.ai), [MiniMax](https://www.minimax.io), OpenAI, or your own endpoint. Switch with `hermes model` — no code changes, no lock-in.

 <table>
 <tr><td><b>A real terminal interface</b></td><td>Full TUI with multiline editing, slash-command autocomplete, conversation history, interrupt-and-redirect, and streaming tool output.</td></tr>
@@ -33,10 +33,8 @@ Use any model you want — [Nous Portal](https://portal.nousresearch.com), [Open
 curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash
 ```

-Works on Linux, macOS, WSL2, and Android via Termux. The installer handles the platform-specific setup for you.
+Works on Linux, macOS, and WSL2. The installer handles everything — Python, Node.js, dependencies, and the `hermes` command. No prerequisites except git.

-> **Android / Termux:** The tested manual path is documented in the [Termux guide](https://hermes-agent.nousresearch.com/docs/getting-started/termux). On Termux, Hermes installs a curated `.[termux]` extra because the full `.[all]` extra currently pulls Android-incompatible voice dependencies.
->
 > **Windows:** Native Windows is not supported. Please install [WSL2](https://learn.microsoft.com/en-us/windows/wsl/install) and run the command above.

 After installation:
@@ -141,18 +139,11 @@ See `hermes claw migrate --help` for all options, or use the `openclaw-migration

 We welcome contributions! See the [Contributing Guide](https://hermes-agent.nousresearch.com/docs/developer-guide/contributing) for development setup, code style, and PR process.

-Quick start for contributors — clone and go with `setup-hermes.sh`:
+Quick start for contributors:

 ```bash
 git clone https://github.com/NousResearch/hermes-agent.git
 cd hermes-agent
-./setup-hermes.sh     # installs uv, creates venv, installs .[all], symlinks ~/.local/bin/hermes
-./hermes              # auto-detects the venv, no need to `source` first
-```
-
-Manual path (equivalent to the above):
-
-```bash
 curl -LsSf https://astral.sh/uv/install.sh | sh
 uv venv venv --python 3.11
 source venv/bin/activate
@@ -174,7 +165,6 @@ python -m pytest tests/ -q
 - 📚 [Skills Hub](https://agentskills.io)
 - 🐛 [Issues](https://github.com/NousResearch/hermes-agent/issues)
 - 💡 [Discussions](https://github.com/NousResearch/hermes-agent/discussions)
- 🔌 [HermesClaw](https://github.com/AaronWong1999/hermesclaw) — Community WeChat bridge: Run Hermes Agent and OpenClaw on the same WeChat account.

 ---

@@ -1,27 +0,0 @@
-# Hermes Agent v0.10.0 (v2026.4.16)
-
-**Release Date:** April 16, 2026
-
-> The Tool Gateway release — paid Nous Portal subscribers can now use web search, image generation, text-to-speech, and browser automation through their existing subscription with zero additional API keys.
-
---
-
-## ✨ Highlights
-
- **Nous Tool Gateway** — Paid [Nous Portal](https://portal.nousresearch.com) subscribers now get automatic access to **web search** (Firecrawl), **image generation** (FAL / FLUX 2 Pro), **text-to-speech** (OpenAI TTS), and **browser automation** (Browser Use) through their existing subscription. No separate API keys needed — just run `hermes model`, select Nous Portal, and pick which tools to enable. Per-tool opt-in via `use_gateway` config, full integration with `hermes tools` and `hermes status`, and the runtime correctly prefers the gateway even when direct API keys exist. Replaces the old hidden `HERMES_ENABLE_NOUS_MANAGED_TOOLS` env var with clean subscription-based detection. ([#11206](https://github.com/NousResearch/hermes-agent/pull/11206), based on work by @jquesnelle; docs: [#11208](https://github.com/NousResearch/hermes-agent/pull/11208))
-
---
-
-## 🐛 Bug Fixes & Improvements
-
-This release includes 180+ commits with numerous bug fixes, platform improvements, and reliability enhancements across the agent core, gateway, CLI, and tool system. Full details will be published in the v0.11.0 changelog.
-
---
-
-## 👥 Contributors
-
- **@jquesnelle** (emozilla) — Original Tool Gateway implementation ([#10799](https://github.com/NousResearch/hermes-agent/pull/10799)), salvaged and shipped in this release
-
---
-
-**Full Changelog**: [v2026.4.13...v2026.4.16](https://github.com/NousResearch/hermes-agent/compare/v2026.4.13...v2026.4.16)
@@ -1,329 +0,0 @@
-# Hermes Agent v0.9.0 (v2026.4.13)
-
-**Release Date:** April 13, 2026
-**Since v0.8.0:** 487 commits · 269 merged PRs · 167 resolved issues · 493 files changed · 63,281 insertions · 24 contributors
-
-> The everywhere release — Hermes goes mobile with Termux/Android, adds iMessage and WeChat, ships Fast Mode for OpenAI and Anthropic, introduces background process monitoring, launches a local web dashboard for managing your agent, and delivers the deepest security hardening pass yet across 16 supported platforms.
-
---
-
-## ✨ Highlights
-
- **Local Web Dashboard** — A new browser-based dashboard for managing your Hermes Agent locally. Configure settings, monitor sessions, browse skills, and manage your gateway — all from a clean web interface without touching config files or the terminal. The easiest way to get started with Hermes.
-
- **Fast Mode (`/fast`)** — Priority processing for OpenAI and Anthropic models. Toggle `/fast` to route through priority queues for significantly lower latency on supported models (GPT-5.4, Codex, Claude). Expands across all OpenAI Priority Processing models and Anthropic's fast tier. ([#6875](https://github.com/NousResearch/hermes-agent/pull/6875), [#6960](https://github.com/NousResearch/hermes-agent/pull/6960), [#7037](https://github.com/NousResearch/hermes-agent/pull/7037))
-
- **iMessage via BlueBubbles** — Full iMessage integration through BlueBubbles, bringing Hermes to Apple's messaging ecosystem. Auto-webhook registration, setup wizard integration, and crash resilience. ([#6437](https://github.com/NousResearch/hermes-agent/pull/6437), [#6460](https://github.com/NousResearch/hermes-agent/pull/6460), [#6494](https://github.com/NousResearch/hermes-agent/pull/6494))
-
- **WeChat (Weixin) & WeCom Callback Mode** — Native WeChat support via iLink Bot API and a new WeCom callback-mode adapter for self-built enterprise apps. Streaming cursor, media uploads, markdown link handling, and atomic state persistence. Hermes now covers the Chinese messaging ecosystem end-to-end. ([#7166](https://github.com/NousResearch/hermes-agent/pull/7166), [#7943](https://github.com/NousResearch/hermes-agent/pull/7943))
-
- **Termux / Android Support** — Run Hermes natively on Android via Termux. Adapted install paths, TUI optimizations for mobile screens, voice backend support, and the `/image` command work on-device. ([#6834](https://github.com/NousResearch/hermes-agent/pull/6834))
-
- **Background Process Monitoring (`watch_patterns`)** — Set patterns to watch for in background process output and get notified in real-time when they match. Monitor for errors, wait for specific events ("listening on port"), or watch build logs — all without polling. ([#7635](https://github.com/NousResearch/hermes-agent/pull/7635))
-
- **Native xAI & Xiaomi MiMo Providers** — First-class provider support for xAI (Grok) and Xiaomi MiMo, with direct API access, model catalogs, and setup wizard integration. Plus Qwen OAuth with portal request support. ([#7372](https://github.com/NousResearch/hermes-agent/pull/7372), [#7855](https://github.com/NousResearch/hermes-agent/pull/7855))
-
- **Pluggable Context Engine** — Context management is now a pluggable slot via `hermes plugins`. Swap in custom context engines that control what the agent sees each turn — filtering, summarization, or domain-specific context injection. ([#7464](https://github.com/NousResearch/hermes-agent/pull/7464))
-
- **Unified Proxy Support** — SOCKS proxy, `DISCORD_PROXY`, and system proxy auto-detection across all gateway platforms. Hermes behind corporate firewalls just works. ([#6814](https://github.com/NousResearch/hermes-agent/pull/6814))
-
- **Comprehensive Security Hardening** — Path traversal protection in checkpoint manager, shell injection neutralization in sandbox writes, SSRF redirect guards in Slack image uploads, Twilio webhook signature validation (SMS RCE fix), API server auth enforcement, git argument injection prevention, and approval button authorization. ([#7933](https://github.com/NousResearch/hermes-agent/pull/7933), [#7944](https://github.com/NousResearch/hermes-agent/pull/7944), [#7940](https://github.com/NousResearch/hermes-agent/pull/7940), [#7151](https://github.com/NousResearch/hermes-agent/pull/7151), [#7156](https://github.com/NousResearch/hermes-agent/pull/7156))
-
- **`hermes backup` & `hermes import`** — Full backup and restore of your Hermes configuration, sessions, skills, and memory. Migrate between machines or create snapshots before major changes. ([#7997](https://github.com/NousResearch/hermes-agent/pull/7997))
-
- **16 Supported Platforms** — With BlueBubbles (iMessage) and WeChat joining Telegram, Discord, Slack, WhatsApp, Signal, Matrix, Email, SMS, DingTalk, Feishu, WeCom, Mattermost, Home Assistant, and Webhooks, Hermes now runs on 16 messaging platforms out of the box.
-
- **`/debug` & `hermes debug share`** — New debugging toolkit: `/debug` slash command across all platforms for quick diagnostics, plus `hermes debug share` to upload a full debug report to a pastebin for easy sharing when troubleshooting. ([#8681](https://github.com/NousResearch/hermes-agent/pull/8681))
-
---
-
-## 🏗️ Core Agent & Architecture
-
-### Provider & Model Support
- **Native xAI (Grok) provider** with direct API access and model catalog ([#7372](https://github.com/NousResearch/hermes-agent/pull/7372))
- **Xiaomi MiMo as first-class provider** — setup wizard, model catalog, empty response recovery ([#7855](https://github.com/NousResearch/hermes-agent/pull/7855))
- **Qwen OAuth provider** with portal request support ([#6282](https://github.com/NousResearch/hermes-agent/pull/6282))
- **Fast Mode** — `/fast` toggle for OpenAI Priority Processing + Anthropic fast tier ([#6875](https://github.com/NousResearch/hermes-agent/pull/6875), [#6960](https://github.com/NousResearch/hermes-agent/pull/6960), [#7037](https://github.com/NousResearch/hermes-agent/pull/7037))
- **Structured API error classification** for smart failover decisions ([#6514](https://github.com/NousResearch/hermes-agent/pull/6514))
- **Rate limit header capture** shown in `/usage` ([#6541](https://github.com/NousResearch/hermes-agent/pull/6541))
- **API server model name** derived from profile name ([#6857](https://github.com/NousResearch/hermes-agent/pull/6857))
- **Custom providers** now included in `/model` listings and resolution ([#7088](https://github.com/NousResearch/hermes-agent/pull/7088))
- **Fallback provider activation** on repeated empty responses with user-visible status ([#7505](https://github.com/NousResearch/hermes-agent/pull/7505))
- **OpenRouter variant tags** (`:free`, `:extended`, `:fast`) preserved during model switch ([#6383](https://github.com/NousResearch/hermes-agent/pull/6383))
- **Credential exhaustion TTL** reduced from 24 hours to 1 hour ([#6504](https://github.com/NousResearch/hermes-agent/pull/6504))
- **OAuth credential lifecycle** hardening — stale pool keys, auth.json sync, Codex CLI race fixes ([#6874](https://github.com/NousResearch/hermes-agent/pull/6874))
- Empty response recovery for reasoning models (MiMo, Qwen, GLM) ([#8609](https://github.com/NousResearch/hermes-agent/pull/8609))
- MiniMax context lengths, thinking guard, endpoint corrections ([#6082](https://github.com/NousResearch/hermes-agent/pull/6082), [#7126](https://github.com/NousResearch/hermes-agent/pull/7126))
- Z.AI endpoint auto-detect via probe and cache ([#5763](https://github.com/NousResearch/hermes-agent/pull/5763))
-
-### Agent Loop & Conversation
- **Pluggable context engine slot** via `hermes plugins` ([#7464](https://github.com/NousResearch/hermes-agent/pull/7464))
- **Background process monitoring** — `watch_patterns` for real-time output alerts ([#7635](https://github.com/NousResearch/hermes-agent/pull/7635))
- **Improved context compression** — higher limits, tool tracking, degradation warnings, token-budget tail protection ([#6395](https://github.com/NousResearch/hermes-agent/pull/6395), [#6453](https://github.com/NousResearch/hermes-agent/pull/6453))
- **`/compress <focus>`** — guided compression with a focus topic ([#8017](https://github.com/NousResearch/hermes-agent/pull/8017))
- **Tiered context pressure warnings** with gateway dedup ([#6411](https://github.com/NousResearch/hermes-agent/pull/6411))
- **Staged inactivity warning** before timeout escalation ([#6387](https://github.com/NousResearch/hermes-agent/pull/6387))
- **Prevent agent from stopping mid-task** — compression floor, budget overhaul, activity tracking ([#7983](https://github.com/NousResearch/hermes-agent/pull/7983))
- **Propagate child activity to parent** during `delegate_task` ([#7295](https://github.com/NousResearch/hermes-agent/pull/7295))
- **Truncated streaming tool call detection** before execution ([#6847](https://github.com/NousResearch/hermes-agent/pull/6847))
- Empty response retry (3 attempts with nudge) ([#6488](https://github.com/NousResearch/hermes-agent/pull/6488))
- Adaptive streaming backoff + cursor strip to prevent message truncation ([#7683](https://github.com/NousResearch/hermes-agent/pull/7683))
- Compression uses live session model instead of stale persisted config ([#8258](https://github.com/NousResearch/hermes-agent/pull/8258))
- Strip `<thought>` tags from Gemma 4 responses ([#8562](https://github.com/NousResearch/hermes-agent/pull/8562))
- Prevent `<think>` in prose from suppressing response output ([#6968](https://github.com/NousResearch/hermes-agent/pull/6968))
- Turn-exit diagnostic logging to agent loop ([#6549](https://github.com/NousResearch/hermes-agent/pull/6549))
- Scope tool interrupt signal per-thread to prevent cross-session leaks ([#7930](https://github.com/NousResearch/hermes-agent/pull/7930))
-
-### Memory & Sessions
- **Hindsight memory plugin** — feature parity, setup wizard, config improvements — @nicoloboschi ([#6428](https://github.com/NousResearch/hermes-agent/pull/6428))
- **Honcho** — opt-in `initOnSessionStart` for tools mode — @Kathie-yu ([#6995](https://github.com/NousResearch/hermes-agent/pull/6995))
- Orphan children instead of cascade-deleting in prune/delete ([#6513](https://github.com/NousResearch/hermes-agent/pull/6513))
- Doctor command only checks the active memory provider ([#6285](https://github.com/NousResearch/hermes-agent/pull/6285))
-
---
-
-## 📱 Messaging Platforms (Gateway)
-
-### New Platforms
- **BlueBubbles (iMessage)** — full adapter with auto-webhook registration, setup wizard, and crash resilience ([#6437](https://github.com/NousResearch/hermes-agent/pull/6437), [#6460](https://github.com/NousResearch/hermes-agent/pull/6460), [#6494](https://github.com/NousResearch/hermes-agent/pull/6494), [#7107](https://github.com/NousResearch/hermes-agent/pull/7107))
- **Weixin (WeChat)** — native support via iLink Bot API with streaming, media uploads, markdown links ([#7166](https://github.com/NousResearch/hermes-agent/pull/7166), [#8665](https://github.com/NousResearch/hermes-agent/pull/8665))
- **WeCom Callback Mode** — self-built enterprise app adapter with atomic state persistence ([#7943](https://github.com/NousResearch/hermes-agent/pull/7943), [#7928](https://github.com/NousResearch/hermes-agent/pull/7928))
-
-### Discord
- **Allowed channels whitelist** config — @jarvis-phw ([#7044](https://github.com/NousResearch/hermes-agent/pull/7044))
- **Forum channel topic inheritance** in thread sessions — @hermes-agent-dhabibi ([#6377](https://github.com/NousResearch/hermes-agent/pull/6377))
- **DISCORD_REPLY_TO_MODE** setting ([#6333](https://github.com/NousResearch/hermes-agent/pull/6333))
- Accept `.log` attachments, raise document size limit — @kira-ariaki ([#6467](https://github.com/NousResearch/hermes-agent/pull/6467))
- Decouple readiness from slash sync ([#8016](https://github.com/NousResearch/hermes-agent/pull/8016))
-
-### Slack
- **Consolidated Slack improvements** — 7 community PRs salvaged into one ([#6809](https://github.com/NousResearch/hermes-agent/pull/6809))
- Handle assistant thread lifecycle events ([#6433](https://github.com/NousResearch/hermes-agent/pull/6433))
-
-### Matrix
- **Migrated from matrix-nio to mautrix-python** ([#7518](https://github.com/NousResearch/hermes-agent/pull/7518))
- SQLite crypto store replacing pickle (fixes E2EE decryption) — @alt-glitch ([#7981](https://github.com/NousResearch/hermes-agent/pull/7981))
- Cross-signing recovery key verification for E2EE migration ([#8282](https://github.com/NousResearch/hermes-agent/pull/8282))
- DM mention threads + group chat events for Feishu ([#7423](https://github.com/NousResearch/hermes-agent/pull/7423))
-
-### Gateway Core
- **Unified proxy support** — SOCKS, DISCORD_PROXY, multi-platform with macOS auto-detection ([#6814](https://github.com/NousResearch/hermes-agent/pull/6814))
- **Inbound text batching** for Discord, Matrix, WeCom + adaptive delay ([#6979](https://github.com/NousResearch/hermes-agent/pull/6979))
- **Surface natural mid-turn assistant messages** in chat platforms ([#7978](https://github.com/NousResearch/hermes-agent/pull/7978))
- **WSL-aware gateway** with smart systemd detection ([#7510](https://github.com/NousResearch/hermes-agent/pull/7510))
- **All missing platforms added to setup wizard** ([#7949](https://github.com/NousResearch/hermes-agent/pull/7949))
- **Per-platform `tool_progress` overrides** ([#6348](https://github.com/NousResearch/hermes-agent/pull/6348))
- **Configurable 'still working' notification interval** ([#8572](https://github.com/NousResearch/hermes-agent/pull/8572))
- `/model` switch persists across messages ([#7081](https://github.com/NousResearch/hermes-agent/pull/7081))
- `/usage` shows rate limits, cost, and token details between turns ([#7038](https://github.com/NousResearch/hermes-agent/pull/7038))
- Drain in-flight work before restart ([#7503](https://github.com/NousResearch/hermes-agent/pull/7503))
- Don't evict cached agent on failed runs — prevents MCP restart loop ([#7539](https://github.com/NousResearch/hermes-agent/pull/7539))
- Replace `os.environ` session state with `contextvars` ([#7454](https://github.com/NousResearch/hermes-agent/pull/7454))
- Derive channel directory platforms from enum instead of hardcoded list ([#7450](https://github.com/NousResearch/hermes-agent/pull/7450))
- Validate image downloads before caching (cross-platform) ([#7125](https://github.com/NousResearch/hermes-agent/pull/7125))
- Cross-platform webhook delivery for all platforms ([#7095](https://github.com/NousResearch/hermes-agent/pull/7095))
- Cron Discord thread_id delivery support ([#7106](https://github.com/NousResearch/hermes-agent/pull/7106))
- Feishu QR-based bot onboarding ([#8570](https://github.com/NousResearch/hermes-agent/pull/8570))
- Gateway status scoped to active profile ([#7951](https://github.com/NousResearch/hermes-agent/pull/7951))
- Prevent background process notifications from triggering false pairing requests ([#6434](https://github.com/NousResearch/hermes-agent/pull/6434))
-
---
-
-## 🖥️ CLI & User Experience
-
-### Interactive CLI
- **Termux / Android support** — adapted install paths, TUI, voice, `/image` ([#6834](https://github.com/NousResearch/hermes-agent/pull/6834))
- **Native `/model` picker modal** for provider → model selection ([#8003](https://github.com/NousResearch/hermes-agent/pull/8003))
- **Live per-tool elapsed timer** restored in TUI spinner ([#7359](https://github.com/NousResearch/hermes-agent/pull/7359))
- **Stacked tool progress scrollback** in TUI ([#8201](https://github.com/NousResearch/hermes-agent/pull/8201))
- **Random tips on new session start** (CLI + gateway, 279 tips) ([#8225](https://github.com/NousResearch/hermes-agent/pull/8225), [#8237](https://github.com/NousResearch/hermes-agent/pull/8237))
- **`hermes dump`** — copy-pasteable setup summary for debugging ([#6550](https://github.com/NousResearch/hermes-agent/pull/6550))
- **`hermes backup` / `hermes import`** — full config backup and restore ([#7997](https://github.com/NousResearch/hermes-agent/pull/7997))
- **WSL environment hint** in system prompt ([#8285](https://github.com/NousResearch/hermes-agent/pull/8285))
- **Profile creation UX** — seed SOUL.md + credential warning ([#8553](https://github.com/NousResearch/hermes-agent/pull/8553))
- Shell-aware sudo detection, empty password support ([#6517](https://github.com/NousResearch/hermes-agent/pull/6517))
- Flush stdin after curses/terminal menus to prevent escape sequence leakage ([#7167](https://github.com/NousResearch/hermes-agent/pull/7167))
- Handle broken stdin in prompt_toolkit startup ([#8560](https://github.com/NousResearch/hermes-agent/pull/8560))
-
-### Setup & Configuration
- **Per-platform display verbosity** configuration ([#8006](https://github.com/NousResearch/hermes-agent/pull/8006))
- **Component-separated logging** with session context and filtering ([#7991](https://github.com/NousResearch/hermes-agent/pull/7991))
- **`network.force_ipv4`** config to fix IPv6 timeout issues ([#8196](https://github.com/NousResearch/hermes-agent/pull/8196))
- **Standardize message whitespace and JSON formatting** ([#7988](https://github.com/NousResearch/hermes-agent/pull/7988))
- **Rebrand OpenClaw → Hermes** during migration ([#8210](https://github.com/NousResearch/hermes-agent/pull/8210))
- Config.yaml takes priority over env vars for auxiliary settings ([#7889](https://github.com/NousResearch/hermes-agent/pull/7889))
- Harden setup provider flows + live OpenRouter catalog refresh ([#7078](https://github.com/NousResearch/hermes-agent/pull/7078))
- Normalize reasoning effort ordering across all surfaces ([#6804](https://github.com/NousResearch/hermes-agent/pull/6804))
- Remove dead `LLM_MODEL` env var + migration to clear stale entries ([#6543](https://github.com/NousResearch/hermes-agent/pull/6543))
- Remove `/prompt` slash command — prefix expansion footgun ([#6752](https://github.com/NousResearch/hermes-agent/pull/6752))
- `HERMES_HOME_MODE` env var to override permissions — @ygd58 ([#6993](https://github.com/NousResearch/hermes-agent/pull/6993))
- Fall back to default model when model config is empty ([#8303](https://github.com/NousResearch/hermes-agent/pull/8303))
- Warn when compression model context is too small ([#7894](https://github.com/NousResearch/hermes-agent/pull/7894))
-
---
-
-## 🔧 Tool System
-
-### Environments & Execution
- **Unified spawn-per-call execution layer** for environments ([#6343](https://github.com/NousResearch/hermes-agent/pull/6343))
- **Unified file sync** with mtime tracking, deletion, and transactional state ([#7087](https://github.com/NousResearch/hermes-agent/pull/7087))
- **Persistent sandbox envs** survive between turns ([#6412](https://github.com/NousResearch/hermes-agent/pull/6412))
- **Bulk file sync** via tar pipe for SSH/Modal backends — @alt-glitch ([#8014](https://github.com/NousResearch/hermes-agent/pull/8014))
- **Daytona** — bulk upload, config bridge, silent disk cap ([#7538](https://github.com/NousResearch/hermes-agent/pull/7538))
- Foreground timeout cap to prevent session deadlocks ([#7082](https://github.com/NousResearch/hermes-agent/pull/7082))
- Guard invalid command values ([#6417](https://github.com/NousResearch/hermes-agent/pull/6417))
-
-### MCP
- **`hermes mcp add --env` and `--preset`** support ([#7970](https://github.com/NousResearch/hermes-agent/pull/7970))
- Combine `content` and `structuredContent` when both present ([#7118](https://github.com/NousResearch/hermes-agent/pull/7118))
- MCP tool name deconfliction fixes ([#7654](https://github.com/NousResearch/hermes-agent/pull/7654))
-
-### Browser
- Browser hardening — dead code removal, caching, scroll perf, security, thread safety ([#7354](https://github.com/NousResearch/hermes-agent/pull/7354))
- `/browser connect` auto-launch uses dedicated Chrome profile dir ([#6821](https://github.com/NousResearch/hermes-agent/pull/6821))
- Reap orphaned browser sessions on startup ([#7931](https://github.com/NousResearch/hermes-agent/pull/7931))
-
-### Voice & Vision
- **Voxtral TTS provider** (Mistral AI) ([#7653](https://github.com/NousResearch/hermes-agent/pull/7653))
- **TTS speed support** for Edge TTS, OpenAI TTS, MiniMax ([#8666](https://github.com/NousResearch/hermes-agent/pull/8666))
- **Vision auto-resize** for oversized images, raise limit to 20 MB, retry-on-failure ([#7883](https://github.com/NousResearch/hermes-agent/pull/7883), [#7902](https://github.com/NousResearch/hermes-agent/pull/7902))
- STT provider-model mismatch fix (whisper-1 vs faster-whisper) ([#7113](https://github.com/NousResearch/hermes-agent/pull/7113))
-
-### Other Tools
- **`hermes dump`** command for setup summary ([#6550](https://github.com/NousResearch/hermes-agent/pull/6550))
- TODO store enforces ID uniqueness during replace operations ([#7986](https://github.com/NousResearch/hermes-agent/pull/7986))
- List all available toolsets in `delegate_task` schema description ([#8231](https://github.com/NousResearch/hermes-agent/pull/8231))
- API server: tool progress as custom SSE event to prevent model corruption ([#7500](https://github.com/NousResearch/hermes-agent/pull/7500))
- API server: share one Docker container across all conversations ([#7127](https://github.com/NousResearch/hermes-agent/pull/7127))
-
---
-
-## 🧩 Skills Ecosystem
-
- **Centralized skills index + tree cache** — eliminates rate-limit failures on install ([#8575](https://github.com/NousResearch/hermes-agent/pull/8575))
- **More aggressive skill loading instructions** in system prompt (v3) ([#8209](https://github.com/NousResearch/hermes-agent/pull/8209), [#8286](https://github.com/NousResearch/hermes-agent/pull/8286))
- **Google Workspace skill** migrated to GWS CLI backend ([#6788](https://github.com/NousResearch/hermes-agent/pull/6788))
- **Creative divergence strategies** skill — @SHL0MS ([#6882](https://github.com/NousResearch/hermes-agent/pull/6882))
- **Creative ideation** — constraint-driven project generation — @SHL0MS ([#7555](https://github.com/NousResearch/hermes-agent/pull/7555))
- Parallelize skills browse/search to prevent hanging ([#7301](https://github.com/NousResearch/hermes-agent/pull/7301))
- Read name from SKILL.md frontmatter in skills_sync ([#7623](https://github.com/NousResearch/hermes-agent/pull/7623))
-
---
-
-## 🔒 Security & Reliability
-
-### Security Hardening
- **Twilio webhook signature validation** — SMS RCE fix ([#7933](https://github.com/NousResearch/hermes-agent/pull/7933))
- **Shell injection neutralization** in `_write_to_sandbox` via path quoting ([#7940](https://github.com/NousResearch/hermes-agent/pull/7940))
- **Git argument injection** and path traversal prevention in checkpoint manager ([#7944](https://github.com/NousResearch/hermes-agent/pull/7944))
- **SSRF redirect bypass** in Slack image uploads + base.py cache helpers ([#7151](https://github.com/NousResearch/hermes-agent/pull/7151))
- **Path traversal, credential gate, DANGEROUS_PATTERNS gaps** ([#7156](https://github.com/NousResearch/hermes-agent/pull/7156))
- **API bind guard** — enforce `API_SERVER_KEY` for non-loopback binding ([#7455](https://github.com/NousResearch/hermes-agent/pull/7455))
- **Approval button authorization** — require auth for session continuation — @Cafexss ([#6930](https://github.com/NousResearch/hermes-agent/pull/6930))
- Path boundary enforcement in skill manager operations ([#7156](https://github.com/NousResearch/hermes-agent/pull/7156))
- DingTalk/API webhook URL origin validation, header injection rejection ([#7455](https://github.com/NousResearch/hermes-agent/pull/7455))
-
-### Reliability
- **Contextual error diagnostics** for invalid API responses ([#8565](https://github.com/NousResearch/hermes-agent/pull/8565))
- **Prevent 400 format errors** from triggering compression loop on Codex ([#6751](https://github.com/NousResearch/hermes-agent/pull/6751))
- **Don't halve context_length** on output-cap-too-large errors — @KUSH42 ([#6664](https://github.com/NousResearch/hermes-agent/pull/6664))
- **Recover primary client** on OpenAI transport errors ([#7108](https://github.com/NousResearch/hermes-agent/pull/7108))
- **Credential pool rotation** on billing-classified 400s ([#7112](https://github.com/NousResearch/hermes-agent/pull/7112))
- **Auto-increase stream read timeout** for local LLM providers ([#6967](https://github.com/NousResearch/hermes-agent/pull/6967))
- **Fall back to default certs** when CA bundle path doesn't exist ([#7352](https://github.com/NousResearch/hermes-agent/pull/7352))
- **Disambiguate usage-limit patterns** in error classifier — @sprmn24 ([#6836](https://github.com/NousResearch/hermes-agent/pull/6836))
- Harden cron script timeout and provider recovery ([#7079](https://github.com/NousResearch/hermes-agent/pull/7079))
- Gateway interrupt detection resilient to monitor task failures ([#8208](https://github.com/NousResearch/hermes-agent/pull/8208))
- Prevent unwanted session auto-reset after graceful gateway restarts ([#8299](https://github.com/NousResearch/hermes-agent/pull/8299))
- Prevent duplicate update prompt spam in gateway watcher ([#8343](https://github.com/NousResearch/hermes-agent/pull/8343))
- Deduplicate reasoning items in Responses API input ([#7946](https://github.com/NousResearch/hermes-agent/pull/7946))
-
-### Infrastructure
- **Multi-arch Docker image** — amd64 + arm64 ([#6124](https://github.com/NousResearch/hermes-agent/pull/6124))
- **Docker runs as non-root user** with virtualenv — @benbarclay contributing ([#8226](https://github.com/NousResearch/hermes-agent/pull/8226))
- **Use `uv`** for Docker dependency resolution to fix resolution-too-deep ([#6965](https://github.com/NousResearch/hermes-agent/pull/6965))
- **Container-aware Nix CLI** — auto-route into managed container — @alt-glitch ([#7543](https://github.com/NousResearch/hermes-agent/pull/7543))
- **Nix shared-state permission model** for interactive CLI users — @alt-glitch ([#6796](https://github.com/NousResearch/hermes-agent/pull/6796))
- **Per-profile subprocess HOME isolation** ([#7357](https://github.com/NousResearch/hermes-agent/pull/7357))
- Profile paths fixed in Docker — profiles go to mounted volume ([#7170](https://github.com/NousResearch/hermes-agent/pull/7170))
- Docker container gateway pathway hardened ([#8614](https://github.com/NousResearch/hermes-agent/pull/8614))
- Enable unbuffered stdout for live Docker logs ([#6749](https://github.com/NousResearch/hermes-agent/pull/6749))
- Install procps in Docker image — @HiddenPuppy ([#7032](https://github.com/NousResearch/hermes-agent/pull/7032))
- Shallow git clone for faster installation — @sosyz ([#8396](https://github.com/NousResearch/hermes-agent/pull/8396))
- `hermes update` always reset on stash conflict ([#7010](https://github.com/NousResearch/hermes-agent/pull/7010))
- Write update exit code before gateway restart (cgroup kill race) ([#8288](https://github.com/NousResearch/hermes-agent/pull/8288))
- Nix: `setupSecrets` optional, tirith runtime dep — @devorun, @ethernet8023 ([#6261](https://github.com/NousResearch/hermes-agent/pull/6261), [#6721](https://github.com/NousResearch/hermes-agent/pull/6721))
- launchd stop uses `bootout` so `KeepAlive` doesn't respawn ([#7119](https://github.com/NousResearch/hermes-agent/pull/7119))
-
---
-
-## 🐛 Notable Bug Fixes
-
- Fix: `/model` switch not persisting across gateway messages ([#7081](https://github.com/NousResearch/hermes-agent/pull/7081))
- Fix: session-scoped gateway model overrides ignored — @Hygaard ([#7662](https://github.com/NousResearch/hermes-agent/pull/7662))
- Fix: compaction model context length ignoring config — 3 related issues ([#8258](https://github.com/NousResearch/hermes-agent/pull/8258), [#8107](https://github.com/NousResearch/hermes-agent/pull/8107))
- Fix: OpenCode.ai context window resolved to 128K instead of 1M ([#6472](https://github.com/NousResearch/hermes-agent/pull/6472))
- Fix: Codex fallback auth-store lookup — @cherifya ([#6462](https://github.com/NousResearch/hermes-agent/pull/6462))
- Fix: duplicate completion notifications when process killed ([#7124](https://github.com/NousResearch/hermes-agent/pull/7124))
- Fix: agent daemon thread prevents orphan CLI processes on tab close ([#8557](https://github.com/NousResearch/hermes-agent/pull/8557))
- Fix: stale image attachment on text paste and voice input ([#7077](https://github.com/NousResearch/hermes-agent/pull/7077))
- Fix: DM thread session seeding causing cross-thread contamination ([#7084](https://github.com/NousResearch/hermes-agent/pull/7084))
- Fix: OpenClaw migration shows dry-run preview before executing ([#6769](https://github.com/NousResearch/hermes-agent/pull/6769))
- Fix: auth errors misclassified as retryable — @kuishou68 ([#7027](https://github.com/NousResearch/hermes-agent/pull/7027))
- Fix: Copilot-Integration-Id header missing ([#7083](https://github.com/NousResearch/hermes-agent/pull/7083))
- Fix: ACP session capabilities — @luyao618 ([#6985](https://github.com/NousResearch/hermes-agent/pull/6985))
- Fix: ACP PromptResponse usage from top-level fields ([#7086](https://github.com/NousResearch/hermes-agent/pull/7086))
- Fix: several failing/flaky tests on main — @dsocolobsky ([#6777](https://github.com/NousResearch/hermes-agent/pull/6777))
- Fix: backup marker filenames — @sprmn24 ([#8600](https://github.com/NousResearch/hermes-agent/pull/8600))
- Fix: `NoneType` in fast_mode check — @0xbyt4 ([#7350](https://github.com/NousResearch/hermes-agent/pull/7350))
- Fix: missing imports in uninstall.py — @JiayuuWang ([#7034](https://github.com/NousResearch/hermes-agent/pull/7034))
-
---
-
-## 📚 Documentation
-
- Platform adapter developer guide + WeCom Callback docs ([#7969](https://github.com/NousResearch/hermes-agent/pull/7969))
- Cron troubleshooting guide ([#7122](https://github.com/NousResearch/hermes-agent/pull/7122))
- Streaming timeout auto-detection for local LLMs ([#6990](https://github.com/NousResearch/hermes-agent/pull/6990))
- Tool-use enforcement documentation expanded ([#7984](https://github.com/NousResearch/hermes-agent/pull/7984))
- BlueBubbles pairing instructions ([#6548](https://github.com/NousResearch/hermes-agent/pull/6548))
- Telegram proxy support section ([#6348](https://github.com/NousResearch/hermes-agent/pull/6348))
- `hermes dump` and `hermes logs` CLI reference ([#6552](https://github.com/NousResearch/hermes-agent/pull/6552))
- `tool_progress_overrides` configuration reference ([#6364](https://github.com/NousResearch/hermes-agent/pull/6364))
- Compression model context length warning docs ([#7879](https://github.com/NousResearch/hermes-agent/pull/7879))
-
---
-
-## 👥 Contributors
-
-**269 merged PRs** from **24 contributors** across **487 commits**.
-
-### Community Contributors
- **@alt-glitch** (6 PRs) — Nix container-aware CLI, shared-state permissions, Matrix SQLite crypto store, bulk SSH/Modal file sync, Matrix mautrix compat
- **@SHL0MS** (2 PRs) — Creative divergence strategies skill, creative ideation skill
- **@sprmn24** (2 PRs) — Error classifier disambiguation, backup marker fix
- **@nicoloboschi** — Hindsight memory plugin feature parity
- **@Hygaard** — Session-scoped gateway model override fix
- **@jarvis-phw** — Discord allowed_channels whitelist
- **@Kathie-yu** — Honcho initOnSessionStart for tools mode
- **@hermes-agent-dhabibi** — Discord forum channel topic inheritance
- **@kira-ariaki** — Discord .log attachments and size limit
- **@cherifya** — Codex fallback auth-store lookup
- **@Cafexss** — Security: auth for session continuation
- **@KUSH42** — Compaction context_length fix
- **@kuishou68** — Auth error retryable classification fix
- **@luyao618** — ACP session capabilities
- **@ygd58** — HERMES_HOME_MODE env var override
- **@0xbyt4** — Fast mode NoneType fix
- **@JiayuuWang** — CLI uninstall import fix
- **@HiddenPuppy** — Docker procps installation
- **@dsocolobsky** — Test suite fixes
- **@bobashopcashier** (1 PR) — Graceful gateway drain before restart (salvaged into #7503 from #7290)
- **@benbarclay** — Docker image tag simplification
- **@sosyz** — Shallow git clone for faster install
- **@devorun** — Nix setupSecrets optional
- **@ethernet8023** — Nix tirith runtime dep
-
---
-
-**Full Changelog**: [v2026.4.8...v2026.4.13](https://github.com/NousResearch/hermes-agent/compare/v2026.4.8...v2026.4.13)
@@ -1,84 +0,0 @@
-# Hermes Agent Security Policy
-
-This document outlines the security protocols, trust model, and deployment hardening guidelines for the **Hermes Agent** project.
-
-## 1. Vulnerability Reporting
-
-Hermes Agent does **not** operate a bug bounty program. Security issues should be reported via [GitHub Security Advisories (GHSA)](https://github.com/NousResearch/hermes-agent/security/advisories/new) or by emailing **security@nousresearch.com**. Do not open public issues for security vulnerabilities.
-
-### Required Submission Details
- **Title & Severity:** Concise description and CVSS score/rating.
- **Affected Component:** Exact file path and line range (e.g., `tools/approval.py:120-145`).
- **Environment:** Output of `hermes version`, commit SHA, OS, and Python version.
- **Reproduction:** Step-by-step Proof-of-Concept (PoC) against `main` or the latest release.
- **Impact:** Explanation of what trust boundary was crossed.
-
---
-
-## 2. Trust Model
-
-The core assumption is that Hermes is a **personal agent** with one trusted operator.
-
-### Operator & Session Trust
- **Single Tenant:** The system protects the operator from LLM actions, not from malicious co-tenants. Multi-user isolation must happen at the OS/host level.
- **Gateway Security:** Authorized callers (Telegram, Discord, Slack, etc.) receive equal trust. Session keys are used for routing, not as authorization boundaries.
- **Execution:** Defaults to `terminal.backend: local` (direct host execution). Container isolation (Docker, Modal, Daytona) is opt-in for sandboxing.
-
-### Dangerous Command Approval
-The approval system (`tools/approval.py`) is a core security boundary. Terminal commands, file operations, and other potentially destructive actions are gated behind explicit user confirmation before execution. The approval mode is configurable via `approvals.mode` in `config.yaml`:
- `"on"` (default) — prompts the user to approve dangerous commands.
- `"auto"` — auto-approves after a configurable delay.
- `"off"` — disables the gate entirely (break-glass; see Section 3).
-
-### Output Redaction
-`agent/redact.py` strips secret-like patterns (API keys, tokens, credentials) from all display output before it reaches the terminal or gateway platform. This prevents accidental credential leakage in chat logs, tool previews, and response text. Redaction operates on the display layer only — underlying values remain intact for internal agent operations.
-
-### Skills vs. MCP Servers
- **Installed Skills:** High trust. Equivalent to local host code; skills can read environment variables and run arbitrary commands.
- **MCP Servers:** Lower trust. MCP subprocesses receive a filtered environment (`_build_safe_env()` in `tools/mcp_tool.py`) — only safe baseline variables (`PATH`, `HOME`, `XDG_*`) plus variables explicitly declared in the server's `env` config block are passed through. Host credentials are stripped by default. Additionally, packages invoked via `npx`/`uvx` are checked against the OSV malware database before spawning.
-
-### Code Execution Sandbox
-The `execute_code` tool (`tools/code_execution_tool.py`) runs LLM-generated Python scripts in a child process with API keys and tokens stripped from the environment to prevent credential exfiltration. Only environment variables explicitly declared by loaded skills (via `env_passthrough`) or by the user in `config.yaml` (`terminal.env_passthrough`) are passed through. The child accesses Hermes tools via RPC, not direct API calls.
-
-### Subagents
- **No recursive delegation:** The `delegate_task` tool is disabled for child agents.
- **Depth limit:** `MAX_DEPTH = 2` — parent (depth 0) can spawn a child (depth 1); grandchildren are rejected.
- **Memory isolation:** Subagents run with `skip_memory=True` and do not have access to the parent's persistent memory provider. The parent receives only the task prompt and final response as an observation.
-
---
-
-## 3. Out of Scope (Non-Vulnerabilities)
-
-The following scenarios are **not** considered security breaches:
- **Prompt Injection:** Unless it results in a concrete bypass of the approval system, toolset restrictions, or container sandbox.
- **Public Exposure:** Deploying the gateway to the public internet without external authentication or network protection.
- **Trusted State Access:** Reports that require pre-existing write access to `~/.hermes/`, `.env`, or `config.yaml` (these are operator-owned files).
- **Default Behavior:** Host-level command execution when `terminal.backend` is set to `local` — this is the documented default, not a vulnerability.
- **Configuration Trade-offs:** Intentional break-glass settings such as `approvals.mode: "off"` or `terminal.backend: local` in production.
- **Tool-level read/access restrictions:** The agent has unrestricted shell access via the `terminal` tool by design. Reports that a specific tool (e.g., `read_file`) can access a resource are not vulnerabilities if the same access is available through `terminal`. Tool-level deny lists only constitute a meaningful security boundary when paired with equivalent restrictions on the terminal side (as with write operations, where `WRITE_DENIED_PATHS` is paired with the dangerous command approval system).
-
---
-
-## 4. Deployment Hardening & Best Practices
-
-### Filesystem & Network
- **Production sandboxing:** Use container backends (`docker`, `modal`, `daytona`) instead of `local` for untrusted workloads.
- **File permissions:** Run as non-root (the Docker image uses UID 10000); protect credentials with `chmod 600 ~/.hermes/.env` on local installs.
- **Network exposure:** Do not expose the gateway or API server to the public internet without VPN, Tailscale, or firewall protection. SSRF protection is enabled by default across all gateway platform adapters (Telegram, Discord, Slack, Matrix, Mattermost, etc.) with redirect validation. Note: the local terminal backend does not apply SSRF filtering, as it operates within the trusted operator's environment.
-
-### Skills & Supply Chain
- **Skill installation:** Review Skills Guard reports (`tools/skills_guard.py`) before installing third-party skills. The audit log at `~/.hermes/skills/.hub/audit.log` tracks every install and removal.
- **MCP safety:** OSV malware checking runs automatically for `npx`/`uvx` packages before MCP server processes are spawned.
- **CI/CD:** GitHub Actions are pinned to full commit SHAs. The `supply-chain-audit.yml` workflow blocks PRs containing `.pth` files or suspicious `base64`+`exec` patterns.
-
-### Credential Storage
- API keys and tokens belong exclusively in `~/.hermes/.env` — never in `config.yaml` or checked into version control.
- The credential pool system (`agent/credential_pool.py`) handles key rotation and fallback. Credentials are resolved from environment variables, not stored in plaintext databases.
-
---
-
-## 5. Disclosure Process
-
- **Coordinated Disclosure:** 90-day window or until a fix is released, whichever comes first.
- **Communication:** All updates occur via the GHSA thread or email correspondence with security@nousresearch.com.
- **Credits:** Reporters are credited in release notes unless anonymity is requested.
@@ -49,7 +49,6 @@ def make_tool_progress_cb(
    session_id: str,
    loop: asyncio.AbstractEventLoop,
    tool_call_ids: Dict[str, Deque[str]],
-    tool_call_meta: Dict[str, Dict[str, Any]],
 ) -> Callable:
    """Create a ``tool_progress_callback`` for AIAgent.

@@ -85,16 +84,6 @@ def make_tool_progress_cb(
            tool_call_ids[name] = queue
        queue.append(tc_id)

-        snapshot = None
-        if name in {"write_file", "patch", "skill_manage"}:
-            try:
-                from agent.display import capture_local_edit_snapshot
-
-                snapshot = capture_local_edit_snapshot(name, args)
-            except Exception:
-                logger.debug("Failed to capture ACP edit snapshot for %s", name, exc_info=True)
-        tool_call_meta[tc_id] = {"args": args, "snapshot": snapshot}
-
        update = build_tool_start(tc_id, name, args)
        _send_update(conn, session_id, loop, update)

@@ -130,7 +119,6 @@ def make_step_cb(
    session_id: str,
    loop: asyncio.AbstractEventLoop,
    tool_call_ids: Dict[str, Deque[str]],
-    tool_call_meta: Dict[str, Dict[str, Any]],
 ) -> Callable:
    """Create a ``step_callback`` for AIAgent.

@@ -144,12 +132,10 @@ def make_step_cb(
            for tool_info in prev_tools:
                tool_name = None
                result = None
-                function_args = None

                if isinstance(tool_info, dict):
                    tool_name = tool_info.get("name") or tool_info.get("function_name")
                    result = tool_info.get("result") or tool_info.get("output")
-                    function_args = tool_info.get("arguments") or tool_info.get("args")
                elif isinstance(tool_info, str):
                    tool_name = tool_info

@@ -159,13 +145,8 @@ def make_step_cb(
                    tool_call_ids[tool_name] = queue
                if tool_name and queue:
                    tc_id = queue.popleft()
-                    meta = tool_call_meta.pop(tc_id, {})
                    update = build_tool_complete(
-                        tc_id,
-                        tool_name,
-                        result=str(result) if result is not None else None,
-                        function_args=function_args or meta.get("args"),
-                        snapshot=meta.get("snapshot"),
+                        tc_id, tool_name, result=str(result) if result is not None else None
                    )
                    _send_update(conn, session_id, loop, update)
                    if not queue:
@@ -26,7 +26,6 @@ from acp.schema import (
    McpServerHttp,
    McpServerSse,
    McpServerStdio,
-    ModelInfo,
    NewSessionResponse,
    PromptResponse,
    ResumeSessionResponse,
@@ -37,8 +36,6 @@ from acp.schema import (
    SessionCapabilities,
    SessionForkCapabilities,
    SessionListCapabilities,
-    SessionModelState,
-    SessionResumeCapabilities,
    SessionInfo,
    TextContentBlock,
    UnstructuredCommandInput,
@@ -149,98 +146,6 @@ class HermesACPAgent(acp.Agent):
        self._conn = conn
        logger.info("ACP client connected")

-    @staticmethod
-    def _encode_model_choice(provider: str | None, model: str | None) -> str:
-        """Encode a model selection so ACP clients can keep provider context."""
-        raw_model = str(model or "").strip()
-        if not raw_model:
-            return ""
-        raw_provider = str(provider or "").strip().lower()
-        if not raw_provider:
-            return raw_model
-        return f"{raw_provider}:{raw_model}"
-
-    def _build_model_state(self, state: SessionState) -> SessionModelState | None:
-        """Return the ACP model selector payload for editors like Zed."""
-        model = str(state.model or getattr(state.agent, "model", "") or "").strip()
-        provider = getattr(state.agent, "provider", None) or detect_provider() or "openrouter"
-
-        try:
-            from hermes_cli.models import curated_models_for_provider, normalize_provider, provider_label
-
-            normalized_provider = normalize_provider(provider)
-            provider_name = provider_label(normalized_provider)
-            available_models: list[ModelInfo] = []
-            seen_ids: set[str] = set()
-
-            for model_id, description in curated_models_for_provider(normalized_provider):
-                rendered_model = str(model_id or "").strip()
-                if not rendered_model:
-                    continue
-                choice_id = self._encode_model_choice(normalized_provider, rendered_model)
-                if choice_id in seen_ids:
-                    continue
-                desc_parts = [f"Provider: {provider_name}"]
-                if description:
-                    desc_parts.append(str(description).strip())
-                if rendered_model == model:
-                    desc_parts.append("current")
-                available_models.append(
-                    ModelInfo(
-                        model_id=choice_id,
-                        name=rendered_model,
-                        description=" • ".join(part for part in desc_parts if part),
-                    )
-                )
-                seen_ids.add(choice_id)
-
-            current_model_id = self._encode_model_choice(normalized_provider, model)
-            if current_model_id and current_model_id not in seen_ids:
-                available_models.insert(
-                    0,
-                    ModelInfo(
-                        model_id=current_model_id,
-                        name=model,
-                        description=f"Provider: {provider_name} • current",
-                    ),
-                )
-
-            if available_models:
-                return SessionModelState(
-                    available_models=available_models,
-                    current_model_id=current_model_id or available_models[0].model_id,
-                )
-        except Exception:
-            logger.debug("Could not build ACP model state", exc_info=True)
-
-        if not model:
-            return None
-
-        fallback_choice = self._encode_model_choice(provider, model)
-        return SessionModelState(
-            available_models=[ModelInfo(model_id=fallback_choice, name=model)],
-            current_model_id=fallback_choice,
-        )
-
-    @staticmethod
-    def _resolve_model_selection(raw_model: str, current_provider: str) -> tuple[str, str]:
-        """Resolve ``provider:model`` input into the provider and normalized model id."""
-        target_provider = current_provider
-        new_model = raw_model.strip()
-
-        try:
-            from hermes_cli.models import detect_provider_for_model, parse_model_input
-
-            target_provider, new_model = parse_model_input(new_model, current_provider)
-            if target_provider == current_provider:
-                detected = detect_provider_for_model(new_model, current_provider)
-                if detected:
-                    target_provider, new_model = detected
-        except Exception:
-            logger.debug("Provider detection failed, using model as-is", exc_info=True)
-
-        return target_provider, new_model
-
    async def _register_session_mcp_servers(
        self,
        state: SessionState,
@@ -340,11 +245,9 @@ class HermesACPAgent(acp.Agent):
            protocol_version=acp.PROTOCOL_VERSION,
            agent_info=Implementation(name="hermes-agent", version=HERMES_VERSION),
            agent_capabilities=AgentCapabilities(
-                load_session=True,
                session_capabilities=SessionCapabilities(
                    fork=SessionForkCapabilities(),
                    list=SessionListCapabilities(),
-                    resume=SessionResumeCapabilities(),
                ),
            ),
            auth_methods=auth_methods,
@@ -367,10 +270,7 @@ class HermesACPAgent(acp.Agent):
        await self._register_session_mcp_servers(state, mcp_servers)
        logger.info("New session %s (cwd=%s)", state.session_id, cwd)
        self._schedule_available_commands_update(state.session_id)
-        return NewSessionResponse(
-            session_id=state.session_id,
-            models=self._build_model_state(state),
-        )
+        return NewSessionResponse(session_id=state.session_id)

    async def load_session(
        self,
@@ -386,7 +286,7 @@ class HermesACPAgent(acp.Agent):
        await self._register_session_mcp_servers(state, mcp_servers)
        logger.info("Loaded session %s", session_id)
        self._schedule_available_commands_update(session_id)
-        return LoadSessionResponse(models=self._build_model_state(state))
+        return LoadSessionResponse()

    async def resume_session(
        self,
@@ -402,7 +302,7 @@ class HermesACPAgent(acp.Agent):
        await self._register_session_mcp_servers(state, mcp_servers)
        logger.info("Resumed session %s", state.session_id)
        self._schedule_available_commands_update(state.session_id)
-        return ResumeSessionResponse(models=self._build_model_state(state))
+        return ResumeSessionResponse()

    async def cancel(self, session_id: str, **kwargs: Any) -> None:
        state = self.session_manager.get_session(session_id)
@@ -437,20 +337,11 @@ class HermesACPAgent(acp.Agent):
        cwd: str | None = None,
        **kwargs: Any,
    ) -> ListSessionsResponse:
-        infos = self.session_manager.list_sessions(cwd=cwd)
-        sessions = []
-        for s in infos:
-            updated_at = s.get("updated_at")
-            if updated_at is not None and not isinstance(updated_at, str):
-                updated_at = str(updated_at)
-            sessions.append(
-                SessionInfo(
-                    session_id=s["session_id"],
-                    cwd=s["cwd"],
-                    title=s.get("title"),
-                    updated_at=updated_at,
-                )
-            )
+        infos = self.session_manager.list_sessions()
+        sessions = [
+            SessionInfo(session_id=s["session_id"], cwd=s["cwd"])
+            for s in infos
+        ]
        return ListSessionsResponse(sessions=sessions)

    # ---- Prompt (core) ------------------------------------------------------
@@ -495,13 +386,12 @@ class HermesACPAgent(acp.Agent):
            state.cancel_event.clear()

        tool_call_ids: dict[str, Deque[str]] = defaultdict(deque)
-        tool_call_meta: dict[str, dict[str, Any]] = {}
        previous_approval_cb = None

        if conn:
-            tool_progress_cb = make_tool_progress_cb(conn, session_id, loop, tool_call_ids, tool_call_meta)
+            tool_progress_cb = make_tool_progress_cb(conn, session_id, loop, tool_call_ids)
            thinking_cb = make_thinking_cb(conn, session_id, loop)
-            step_cb = make_step_cb(conn, session_id, loop, tool_call_ids, tool_call_meta)
+            step_cb = make_step_cb(conn, session_id, loop, tool_call_ids)
            message_cb = make_message_cb(conn, session_id, loop)
            approval_cb = make_approval_callback(conn.request_permission, loop, session_id)
        else:
@@ -556,31 +446,19 @@ class HermesACPAgent(acp.Agent):
            self.session_manager.save_session(session_id)

        final_response = result.get("final_response", "")
-        if final_response:
-            try:
-                from agent.title_generator import maybe_auto_title
-
-                maybe_auto_title(
-                    self.session_manager._get_db(),
-                    session_id,
-                    user_text,
-                    final_response,
-                    state.history,
-                )
-            except Exception:
-                logger.debug("Failed to auto-title ACP session %s", session_id, exc_info=True)
        if final_response and conn:
            update = acp.update_agent_message_text(final_response)
            await conn.session_update(session_id, update)

        usage = None
-        if any(result.get(key) is not None for key in ("prompt_tokens", "completion_tokens", "total_tokens")):
+        usage_data = result.get("usage")
+        if usage_data and isinstance(usage_data, dict):
            usage = Usage(
-                input_tokens=result.get("prompt_tokens", 0),
-                output_tokens=result.get("completion_tokens", 0),
-                total_tokens=result.get("total_tokens", 0),
-                thought_tokens=result.get("reasoning_tokens"),
-                cached_read_tokens=result.get("cache_read_tokens"),
+                input_tokens=usage_data.get("prompt_tokens", 0),
+                output_tokens=usage_data.get("completion_tokens", 0),
+                total_tokens=usage_data.get("total_tokens", 0),
+                thought_tokens=usage_data.get("reasoning_tokens"),
+                cached_read_tokens=usage_data.get("cached_tokens"),
            )

        stop_reason = "cancelled" if state.cancel_event and state.cancel_event.is_set() else "end_turn"
@@ -676,15 +554,27 @@ class HermesACPAgent(acp.Agent):
            provider = getattr(state.agent, "provider", None) or "auto"
            return f"Current model: {model}\nProvider: {provider}"

+        new_model = args.strip()
+        target_provider = None
        current_provider = getattr(state.agent, "provider", None) or "openrouter"
-        target_provider, new_model = self._resolve_model_selection(args, current_provider)
+
+        # Auto-detect provider for the requested model
+        try:
+            from hermes_cli.models import parse_model_input, detect_provider_for_model
+            target_provider, new_model = parse_model_input(new_model, current_provider)
+            if target_provider == current_provider:
+                detected = detect_provider_for_model(new_model, current_provider)
+                if detected:
+                    target_provider, new_model = detected
+        except Exception:
+            logger.debug("Provider detection failed, using model as-is", exc_info=True)

        state.model = new_model
        state.agent = self.session_manager._make_agent(
            session_id=state.session_id,
            cwd=state.cwd,
            model=new_model,
-            requested_provider=target_provider,
+            requested_provider=target_provider or current_provider,
        )
        self.session_manager.save_session(state.session_id)
        provider_label = getattr(state.agent, "provider", None) or target_provider or current_provider
@@ -786,30 +676,20 @@ class HermesACPAgent(acp.Agent):
        """Switch the model for a session (called by ACP protocol)."""
        state = self.session_manager.get_session(session_id)
        if state:
+            state.model = model_id
            current_provider = getattr(state.agent, "provider", None)
-            requested_provider, resolved_model = self._resolve_model_selection(
-                model_id,
-                current_provider or "openrouter",
-            )
-            state.model = resolved_model
-            provider_changed = bool(current_provider and requested_provider != current_provider)
-            current_base_url = None if provider_changed else getattr(state.agent, "base_url", None)
-            current_api_mode = None if provider_changed else getattr(state.agent, "api_mode", None)
+            current_base_url = getattr(state.agent, "base_url", None)
+            current_api_mode = getattr(state.agent, "api_mode", None)
            state.agent = self.session_manager._make_agent(
                session_id=session_id,
                cwd=state.cwd,
-                model=resolved_model,
-                requested_provider=requested_provider,
+                model=model_id,
+                requested_provider=current_provider,
                base_url=current_base_url,
                api_mode=current_api_mode,
            )
            self.session_manager.save_session(session_id)
-            logger.info(
-                "Session %s: model switched to %s via provider %s",
-                session_id,
-                resolved_model,
-                requested_provider,
-            )
+            logger.info("Session %s: model switched to %s", session_id, model_id)
            return SetSessionModelResponse()
        logger.warning("Session %s: model switch requested for missing session", session_id)
        return None
@@ -13,12 +13,8 @@ from hermes_constants import get_hermes_home
 import copy
 import json
 import logging
-import os
-import re
 import sys
-import time
 import uuid
-from datetime import datetime, timezone
 from dataclasses import dataclass, field
 from threading import Lock
 from typing import Any, Dict, List, Optional
@@ -26,64 +22,6 @@ from typing import Any, Dict, List, Optional
 logger = logging.getLogger(__name__)


-def _normalize_cwd_for_compare(cwd: str | None) -> str:
-    raw = str(cwd or ".").strip()
-    if not raw:
-        raw = "."
-    expanded = os.path.expanduser(raw)
-
-    # Normalize Windows drive paths into the equivalent WSL mount form so
-    # ACP history filters match the same workspace across Windows and WSL.
-    match = re.match(r"^([A-Za-z]):[\\/](.*)$", expanded)
-    if match:
-        drive = match.group(1).lower()
-        tail = match.group(2).replace("\\", "/")
-        expanded = f"/mnt/{drive}/{tail}"
-    elif re.match(r"^/mnt/[A-Za-z]/", expanded):
-        expanded = f"/mnt/{expanded[5].lower()}/{expanded[7:]}"
-
-    return os.path.normpath(expanded)
-
-
-def _build_session_title(title: Any, preview: Any, cwd: str | None) -> str:
-    explicit = str(title or "").strip()
-    if explicit:
-        return explicit
-    preview_text = str(preview or "").strip()
-    if preview_text:
-        return preview_text
-    leaf = os.path.basename(str(cwd or "").rstrip("/\\"))
-    return leaf or "New thread"
-
-
-def _format_updated_at(value: Any) -> str | None:
-    if value is None:
-        return None
-    if isinstance(value, str) and value.strip():
-        return value
-    try:
-        return datetime.fromtimestamp(float(value), tz=timezone.utc).isoformat()
-    except Exception:
-        return None
-
-
-def _updated_at_sort_key(value: Any) -> float:
-    if value is None:
-        return float("-inf")
-    if isinstance(value, (int, float)):
-        return float(value)
-    raw = str(value).strip()
-    if not raw:
-        return float("-inf")
-    try:
-        return datetime.fromisoformat(raw.replace("Z", "+00:00")).timestamp()
-    except Exception:
-        try:
-            return float(raw)
-        except Exception:
-            return float("-inf")
-
-
 def _acp_stderr_print(*args, **kwargs) -> None:
    """Best-effort human-readable output sink for ACP stdio sessions.

@@ -224,78 +162,47 @@ class SessionManager:
        logger.info("Forked ACP session %s -> %s", session_id, new_id)
        return state

-    def list_sessions(self, cwd: str | None = None) -> List[Dict[str, Any]]:
+    def list_sessions(self) -> List[Dict[str, Any]]:
        """Return lightweight info dicts for all sessions (memory + database)."""
-        normalized_cwd = _normalize_cwd_for_compare(cwd) if cwd else None
-        db = self._get_db()
-        persisted_rows: dict[str, dict[str, Any]] = {}
-
-        if db is not None:
-            try:
-                for row in db.list_sessions_rich(source="acp", limit=1000):
-                    persisted_rows[str(row["id"])] = dict(row)
-            except Exception:
-                logger.debug("Failed to load ACP sessions from DB", exc_info=True)
-
        # Collect in-memory sessions first.
        with self._lock:
            seen_ids = set(self._sessions.keys())
-            results = []
-            for s in self._sessions.values():
-                history_len = len(s.history)
-                if history_len <= 0:
-                    continue
-                if normalized_cwd and _normalize_cwd_for_compare(s.cwd) != normalized_cwd:
-                    continue
-                persisted = persisted_rows.get(s.session_id, {})
-                preview = next(
-                    (
-                        str(msg.get("content") or "").strip()
-                        for msg in s.history
-                        if msg.get("role") == "user" and str(msg.get("content") or "").strip()
-                    ),
-                    persisted.get("preview") or "",
-                )
-                results.append(
-                    {
-                        "session_id": s.session_id,
-                        "cwd": s.cwd,
-                        "model": s.model,
-                        "history_len": history_len,
-                        "title": _build_session_title(persisted.get("title"), preview, s.cwd),
-                        "updated_at": _format_updated_at(
-                            persisted.get("last_active") or persisted.get("started_at") or time.time()
-                        ),
-                    }
-                )
+            results = [
+                {
+                    "session_id": s.session_id,
+                    "cwd": s.cwd,
+                    "model": s.model,
+                    "history_len": len(s.history),
+                }
+                for s in self._sessions.values()
+            ]

        # Merge any persisted sessions not currently in memory.
-        for sid, row in persisted_rows.items():
-            if sid in seen_ids:
-                continue
-            message_count = int(row.get("message_count") or 0)
-            if message_count <= 0:
-                continue
-            # Extract cwd from model_config JSON.
-            session_cwd = "."
-            mc = row.get("model_config")
-            if mc:
-                try:
-                    session_cwd = json.loads(mc).get("cwd", ".")
-                except (json.JSONDecodeError, TypeError):
-                    pass
-            if normalized_cwd and _normalize_cwd_for_compare(session_cwd) != normalized_cwd:
-                continue
-            results.append({
-                "session_id": sid,
-                "cwd": session_cwd,
-                "model": row.get("model") or "",
-                "history_len": message_count,
-                "title": _build_session_title(row.get("title"), row.get("preview"), session_cwd),
-                "updated_at": _format_updated_at(row.get("last_active") or row.get("started_at")),
-            })
+        db = self._get_db()
+        if db is not None:
+            try:
+                rows = db.search_sessions(source="acp", limit=1000)
+                for row in rows:
+                    sid = row["id"]
+                    if sid in seen_ids:
+                        continue
+                    # Extract cwd from model_config JSON.
+                    cwd = "."
+                    mc = row.get("model_config")
+                    if mc:
+                        try:
+                            cwd = json.loads(mc).get("cwd", ".")
+                        except (json.JSONDecodeError, TypeError):
+                            pass
+                    results.append({
+                        "session_id": sid,
+                        "cwd": cwd,
+                        "model": row.get("model") or "",
+                        "history_len": row.get("message_count") or 0,
+                    })
+            except Exception:
+                logger.debug("Failed to list ACP sessions from DB", exc_info=True)

-        results.sort(key=lambda item: _updated_at_sort_key(item.get("updated_at")), reverse=True)
        return results

    def update_cwd(self, session_id: str, cwd: str) -> Optional[SessionState]:
@@ -2,7 +2,6 @@

 from __future__ import annotations

-import json
 import uuid
 from typing import Any, Dict, List, Optional

@@ -97,170 +96,6 @@ def build_tool_title(tool_name: str, args: Dict[str, Any]) -> str:
    return tool_name


-def _build_patch_mode_content(patch_text: str) -> List[Any]:
-    """Parse V4A patch mode input into ACP diff blocks when possible."""
-    if not patch_text:
-        return [acp.tool_content(acp.text_block(""))]
-
-    try:
-        from tools.patch_parser import OperationType, parse_v4a_patch
-
-        operations, error = parse_v4a_patch(patch_text)
-        if error or not operations:
-            return [acp.tool_content(acp.text_block(patch_text))]
-
-        content: List[Any] = []
-        for op in operations:
-            if op.operation == OperationType.UPDATE:
-                old_chunks: list[str] = []
-                new_chunks: list[str] = []
-                for hunk in op.hunks:
-                    old_lines = [line.content for line in hunk.lines if line.prefix in (" ", "-")]
-                    new_lines = [line.content for line in hunk.lines if line.prefix in (" ", "+")]
-                    if old_lines or new_lines:
-                        old_chunks.append("\n".join(old_lines))
-                        new_chunks.append("\n".join(new_lines))
-
-                old_text = "\n...\n".join(chunk for chunk in old_chunks if chunk)
-                new_text = "\n...\n".join(chunk for chunk in new_chunks if chunk)
-                if old_text or new_text:
-                    content.append(
-                        acp.tool_diff_content(
-                            path=op.file_path,
-                            old_text=old_text or None,
-                            new_text=new_text or "",
-                        )
-                    )
-                continue
-
-            if op.operation == OperationType.ADD:
-                added_lines = [line.content for hunk in op.hunks for line in hunk.lines if line.prefix == "+"]
-                content.append(
-                    acp.tool_diff_content(
-                        path=op.file_path,
-                        new_text="\n".join(added_lines),
-                    )
-                )
-                continue
-
-            if op.operation == OperationType.DELETE:
-                content.append(
-                    acp.tool_diff_content(
-                        path=op.file_path,
-                        old_text=f"Delete file: {op.file_path}",
-                        new_text="",
-                    )
-                )
-                continue
-
-            if op.operation == OperationType.MOVE:
-                content.append(
-                    acp.tool_content(acp.text_block(f"Move file: {op.file_path} -> {op.new_path}"))
-                )
-
-        return content or [acp.tool_content(acp.text_block(patch_text))]
-    except Exception:
-        return [acp.tool_content(acp.text_block(patch_text))]
-
-
-def _strip_diff_prefix(path: str) -> str:
-    raw = str(path or "").strip()
-    if raw.startswith(("a/", "b/")):
-        return raw[2:]
-    return raw
-
-
-def _parse_unified_diff_content(diff_text: str) -> List[Any]:
-    """Convert unified diff text into ACP diff content blocks."""
-    if not diff_text:
-        return []
-
-    content: List[Any] = []
-    current_old_path: Optional[str] = None
-    current_new_path: Optional[str] = None
-    old_lines: list[str] = []
-    new_lines: list[str] = []
-
-    def _flush() -> None:
-        nonlocal current_old_path, current_new_path, old_lines, new_lines
-        if current_old_path is None and current_new_path is None:
-            return
-        path = current_new_path if current_new_path and current_new_path != "/dev/null" else current_old_path
-        if not path or path == "/dev/null":
-            current_old_path = None
-            current_new_path = None
-            old_lines = []
-            new_lines = []
-            return
-        content.append(
-            acp.tool_diff_content(
-                path=_strip_diff_prefix(path),
-                old_text="\n".join(old_lines) if old_lines else None,
-                new_text="\n".join(new_lines),
-            )
-        )
-        current_old_path = None
-        current_new_path = None
-        old_lines = []
-        new_lines = []
-
-    for line in diff_text.splitlines():
-        if line.startswith("--- "):
-            _flush()
-            current_old_path = line[4:].strip()
-            continue
-        if line.startswith("+++ "):
-            current_new_path = line[4:].strip()
-            continue
-        if line.startswith("@@"):
-            continue
-        if current_old_path is None and current_new_path is None:
-            continue
-        if line.startswith("+"):
-            new_lines.append(line[1:])
-        elif line.startswith("-"):
-            old_lines.append(line[1:])
-        elif line.startswith(" "):
-            shared = line[1:]
-            old_lines.append(shared)
-            new_lines.append(shared)
-
-    _flush()
-    return content
-
-
-def _build_tool_complete_content(
-    tool_name: str,
-    result: Optional[str],
-    *,
-    function_args: Optional[Dict[str, Any]] = None,
-    snapshot: Any = None,
-) -> List[Any]:
-    """Build structured ACP completion content, falling back to plain text."""
-    display_result = result or ""
-    if len(display_result) > 5000:
-        display_result = display_result[:4900] + f"\n... ({len(result)} chars total, truncated)"
-
-    if tool_name in {"write_file", "patch", "skill_manage"}:
-        try:
-            from agent.display import extract_edit_diff
-
-            diff_text = extract_edit_diff(
-                tool_name,
-                result,
-                function_args=function_args,
-                snapshot=snapshot,
-            )
-            if isinstance(diff_text, str) and diff_text.strip():
-                diff_content = _parse_unified_diff_content(diff_text)
-                if diff_content:
-                    return diff_content
-        except Exception:
-            pass
-
-    return [acp.tool_content(acp.text_block(display_result))]
-
-
 # ---------------------------------------------------------------------------
 # Build ACP content objects for tool-call events
 # ---------------------------------------------------------------------------
@@ -284,8 +119,9 @@ def build_tool_start(
            new = arguments.get("new_string", "")
            content = [acp.tool_diff_content(path=path, new_text=new, old_text=old)]
        else:
+            # Patch mode — show the patch content as text
            patch_text = arguments.get("patch", "")
-            content = _build_patch_mode_content(patch_text)
+            content = [acp.tool_content(acp.text_block(patch_text))]
        return acp.start_tool_call(
            tool_call_id, title, kind=kind, content=content, locations=locations,
            raw_input=arguments,
@@ -342,17 +178,16 @@ def build_tool_complete(
    tool_call_id: str,
    tool_name: str,
    result: Optional[str] = None,
-    function_args: Optional[Dict[str, Any]] = None,
-    snapshot: Any = None,
 ) -> ToolCallProgress:
    """Create a ToolCallUpdate (progress) event for a completed tool call."""
    kind = get_tool_kind(tool_name)
-    content = _build_tool_complete_content(
-        tool_name,
-        result,
-        function_args=function_args,
-        snapshot=snapshot,
-    )
+
+    # Truncate very large results for the UI
+    display_result = result or ""
+    if len(display_result) > 5000:
+        display_result = display_result[:4900] + f"\n... ({len(result)} chars total, truncated)"
+
+    content = [acp.tool_content(acp.text_block(display_result))]
    return acp.update_tool_call(
        tool_call_id,
        kind=kind,
@@ -28,45 +28,19 @@ except ImportError:
 logger = logging.getLogger(__name__)

 THINKING_BUDGET = {"xhigh": 32000, "high": 16000, "medium": 8000, "low": 4000}
-# Hermes effort → Anthropic adaptive-thinking effort (output_config.effort).
-# Anthropic exposes 5 levels on 4.7+: low, medium, high, xhigh, max.
-# Opus/Sonnet 4.6 only expose 4 levels: low, medium, high, max — no xhigh.
-# We preserve xhigh as xhigh on 4.7+ (the recommended default for coding/
-# agentic work) and downgrade it to max on pre-4.7 adaptive models (which
-# is the strongest level they accept).  "minimal" is a legacy alias that
-# maps to low on every model.  See:
-# https://platform.claude.com/docs/en/about-claude/models/migration-guide
 ADAPTIVE_EFFORT_MAP = {
-    "max":     "max",
-    "xhigh":   "xhigh",
-    "high":    "high",
-    "medium":  "medium",
-    "low":     "low",
+    "xhigh": "max",
+    "high": "high",
+    "medium": "medium",
+    "low": "low",
    "minimal": "low",
 }

-# Models that accept the "xhigh" output_config.effort level.  Opus 4.7 added
-# xhigh as a distinct level between high and max; older adaptive-thinking
-# models (4.6) reject it with a 400.  Keep this substring list in sync with
-# the Anthropic migration guide as new model families ship.
-_XHIGH_EFFORT_SUBSTRINGS = ("4-7", "4.7")
-
-# Models where extended thinking is deprecated/removed (4.6+ behavior: adaptive
-# is the only supported mode; 4.7 additionally forbids manual thinking entirely
-# and drops temperature/top_p/top_k).
-_ADAPTIVE_THINKING_SUBSTRINGS = ("4-6", "4.6", "4-7", "4.7")
-
-# Models where temperature/top_p/top_k return 400 if set to non-default values.
-# This is the Opus 4.7 contract; future 4.x+ models are expected to follow it.
-_NO_SAMPLING_PARAMS_SUBSTRINGS = ("4-7", "4.7")
-
 # ── Max output token limits per Anthropic model ───────────────────────
 # Source: Anthropic docs + Cline model catalog.  Anthropic's API requires
 # max_tokens as a mandatory field.  Previously we hardcoded 16384, which
 # starves thinking-enabled models (thinking tokens count toward the limit).
 _ANTHROPIC_OUTPUT_LIMITS = {
-    # Claude 4.7
-    "claude-opus-4-7":   128_000,
    # Claude 4.6
    "claude-opus-4-6":   128_000,
    "claude-sonnet-4-6":  64_000,
@@ -86,8 +60,6 @@ _ANTHROPIC_OUTPUT_LIMITS = {
    "claude-3-opus":       4_096,
    "claude-3-sonnet":     4_096,
    "claude-3-haiku":      4_096,
-    # Third-party Anthropic-compatible providers
-    "minimax":            131_072,
 }

 # For any model not in the table, assume the highest current limit.
@@ -102,11 +74,8 @@ def _get_anthropic_max_output(model: str) -> int:
    model IDs (claude-sonnet-4-5-20250929) and variant suffixes (:1m, :fast)
    resolve correctly.  Longest-prefix match wins to avoid e.g. "claude-3-5"
    matching before "claude-3-5-sonnet".
-
-    Normalizes dots to hyphens so that model names like
-    ``anthropic/claude-opus-4.6`` match the ``claude-opus-4-6`` table key.
    """
-    m = model.lower().replace(".", "-")
+    m = model.lower()
    best_key = ""
    best_val = _ANTHROPIC_DEFAULT_OUTPUT_LIMIT
    for key, val in _ANTHROPIC_OUTPUT_LIMITS.items():
@@ -117,50 +86,15 @@ def _get_anthropic_max_output(model: str) -> int:


 def _supports_adaptive_thinking(model: str) -> bool:
-    """Return True for Claude 4.6+ models that support adaptive thinking."""
-    return any(v in model for v in _ADAPTIVE_THINKING_SUBSTRINGS)
+    """Return True for Claude 4.6 models that support adaptive thinking."""
+    return any(v in model for v in ("4-6", "4.6"))


-def _supports_xhigh_effort(model: str) -> bool:
-    """Return True for models that accept the 'xhigh' adaptive effort level.
-
-    Opus 4.7 introduced xhigh as a distinct level between high and max.
-    Pre-4.7 adaptive models (Opus/Sonnet 4.6) only accept low/medium/high/max
-    and reject xhigh with an HTTP 400. Callers should downgrade xhigh→max
-    when this returns False.
-    """
-    return any(v in model for v in _XHIGH_EFFORT_SUBSTRINGS)
-
-
-def _forbids_sampling_params(model: str) -> bool:
-    """Return True for models that 400 on any non-default temperature/top_p/top_k.
-
-    Opus 4.7 explicitly rejects sampling parameters; later Claude releases are
-    expected to follow suit.  Callers should omit these fields entirely rather
-    than passing zero/default values (the API rejects anything non-null).
-    """
-    return any(v in model for v in _NO_SAMPLING_PARAMS_SUBSTRINGS)
-
-
-# Beta headers for enhanced features (sent with ALL auth types).
-# As of Opus 4.7 (2026-04-16), both of these are GA on Claude 4.6+ — the
-# beta headers are still accepted (harmless no-op) but not required. Kept
-# here so older Claude (4.5, 4.1) + third-party Anthropic-compat endpoints
-# that still gate on the headers continue to get the enhanced features.
-# Migration guide: remove these if you no longer support ≤4.5 models.
+# Beta headers for enhanced features (sent with ALL auth types)
 _COMMON_BETAS = [
    "interleaved-thinking-2025-05-14",
    "fine-grained-tool-streaming-2025-05-14",
 ]
-# MiniMax's Anthropic-compatible endpoints fail tool-use requests when
-# the fine-grained tool streaming beta is present.  Omit it so tool calls
-# fall back to the provider's default response path.
-_TOOL_STREAMING_BETA = "fine-grained-tool-streaming-2025-05-14"
-
-# Fast mode beta — enables the ``speed: "fast"`` request parameter for
-# significantly higher output token throughput on Opus 4.6 (~2.5x).
-# See https://platform.claude.com/docs/en/build-with-claude/fast-mode
-_FAST_MODE_BETA = "fast-mode-2026-02-01"

 # Additional beta headers required for OAuth/subscription auth.
 # Matches what Claude Code (and pi-ai / OpenCode) send.
@@ -215,27 +149,18 @@ def _get_claude_code_version() -> str:


 def _is_oauth_token(key: str) -> bool:
-    """Check if the key is an Anthropic OAuth/setup token.
+    """Check if the key is an OAuth/setup token (not a regular Console API key).

-    Positively identifies Anthropic OAuth tokens by their key format:
-    - ``sk-ant-`` prefix (but NOT ``sk-ant-api``) → setup tokens, managed keys
-    - ``eyJ`` prefix → JWTs from the Anthropic OAuth flow
-
-    Non-Anthropic keys (MiniMax, Alibaba, etc.) don't match either pattern
-    and correctly return False.
+    Regular API keys start with 'sk-ant-api'. Everything else (setup-tokens
+    starting with 'sk-ant-oat', managed keys, JWTs, etc.) needs Bearer auth.
    """
    if not key:
        return False
-    # Regular Anthropic Console API keys — x-api-key auth, never OAuth
+    # Regular Console API keys use x-api-key header
    if key.startswith("sk-ant-api"):
        return False
-    # Anthropic-issued tokens (setup-tokens sk-ant-oat-*, managed keys)
-    if key.startswith("sk-ant-"):
-        return True
-    # JWTs from Anthropic OAuth flow
-    if key.startswith("eyJ"):
-        return True
-    return False
+    # Everything else (setup-tokens, managed keys, JWTs) uses Bearer auth
+    return True


 def _normalize_base_url_text(base_url) -> str:
@@ -279,19 +204,6 @@ def _requires_bearer_auth(base_url: str | None) -> bool:
    return normalized.startswith(("https://api.minimax.io/anthropic", "https://api.minimaxi.com/anthropic"))


-def _common_betas_for_base_url(base_url: str | None) -> list[str]:
-    """Return the beta headers that are safe for the configured endpoint.
-
-    MiniMax's Anthropic-compatible endpoints (Bearer-auth) reject requests
-    that include Anthropic's ``fine-grained-tool-streaming`` beta — every
-    tool-use message triggers a connection error.  Strip that beta for
-    Bearer-auth endpoints while keeping all other betas intact.
-    """
-    if _requires_bearer_auth(base_url):
-        return [b for b in _COMMON_BETAS if b != _TOOL_STREAMING_BETA]
-    return _COMMON_BETAS
-
-
 def build_anthropic_client(api_key: str, base_url: str = None):
    """Create an Anthropic client, auto-detecting setup-tokens vs API keys.

@@ -310,7 +222,6 @@ def build_anthropic_client(api_key: str, base_url: str = None):
    }
    if normalized_base_url:
        kwargs["base_url"] = normalized_base_url
-    common_betas = _common_betas_for_base_url(normalized_base_url)

    if _requires_bearer_auth(normalized_base_url):
        # Some Anthropic-compatible providers (e.g. MiniMax) expect the API key in
@@ -320,21 +231,21 @@ def build_anthropic_client(api_key: str, base_url: str = None):
        # not use Anthropic's sk-ant-api prefix and would otherwise be misread as
        # Anthropic OAuth/setup tokens.
        kwargs["auth_token"] = api_key
-        if common_betas:
-            kwargs["default_headers"] = {"anthropic-beta": ",".join(common_betas)}
+        if _COMMON_BETAS:
+            kwargs["default_headers"] = {"anthropic-beta": ",".join(_COMMON_BETAS)}
    elif _is_third_party_anthropic_endpoint(base_url):
        # Third-party proxies (Azure AI Foundry, AWS Bedrock, etc.) use their
        # own API keys with x-api-key auth. Skip OAuth detection — their keys
        # don't follow Anthropic's sk-ant-* prefix convention and would be
        # misclassified as OAuth tokens.
        kwargs["api_key"] = api_key
-        if common_betas:
-            kwargs["default_headers"] = {"anthropic-beta": ",".join(common_betas)}
+        if _COMMON_BETAS:
+            kwargs["default_headers"] = {"anthropic-beta": ",".join(_COMMON_BETAS)}
    elif _is_oauth_token(api_key):
        # OAuth access token / setup-token → Bearer auth + Claude Code identity.
        # Anthropic routes OAuth requests based on user-agent and headers;
        # without Claude Code's fingerprint, requests get intermittent 500s.
-        all_betas = common_betas + _OAUTH_ONLY_BETAS
+        all_betas = _COMMON_BETAS + _OAUTH_ONLY_BETAS
        kwargs["auth_token"] = api_key
        kwargs["default_headers"] = {
            "anthropic-beta": ",".join(all_betas),
@@ -344,39 +255,12 @@ def build_anthropic_client(api_key: str, base_url: str = None):
    else:
        # Regular API key → x-api-key header + common betas
        kwargs["api_key"] = api_key
-        if common_betas:
-            kwargs["default_headers"] = {"anthropic-beta": ",".join(common_betas)}
+        if _COMMON_BETAS:
+            kwargs["default_headers"] = {"anthropic-beta": ",".join(_COMMON_BETAS)}

    return _anthropic_sdk.Anthropic(**kwargs)


-def build_anthropic_bedrock_client(region: str):
-    """Create an AnthropicBedrock client for Bedrock Claude models.
-
-    Uses the Anthropic SDK's native Bedrock adapter, which provides full
-    Claude feature parity: prompt caching, thinking budgets, adaptive
-    thinking, fast mode — features not available via the Converse API.
-
-    Auth uses the boto3 default credential chain (IAM roles, SSO, env vars).
-    """
-    if _anthropic_sdk is None:
-        raise ImportError(
-            "The 'anthropic' package is required for the Bedrock provider. "
-            "Install it with: pip install 'anthropic>=0.39.0'"
-        )
-    if not hasattr(_anthropic_sdk, "AnthropicBedrock"):
-        raise ImportError(
-            "anthropic.AnthropicBedrock not available. "
-            "Upgrade with: pip install 'anthropic>=0.39.0'"
-        )
-    from httpx import Timeout
-
-    return _anthropic_sdk.AnthropicBedrock(
-        aws_region=region,
-        timeout=Timeout(timeout=900.0, connect=10.0),
-    )
-
-
 def read_claude_code_credentials() -> Optional[Dict[str, Any]]:
    """Read refreshable Claude Code OAuth credentials from ~/.claude/.credentials.json.

@@ -601,6 +485,35 @@ def _prefer_refreshable_claude_code_token(env_token: str, creds: Optional[Dict[s
    return None


+def get_anthropic_token_source(token: Optional[str] = None) -> str:
+    """Best-effort source classification for an Anthropic credential token."""
+    token = (token or "").strip()
+    if not token:
+        return "none"
+
+    env_token = os.getenv("ANTHROPIC_TOKEN", "").strip()
+    if env_token and env_token == token:
+        return "anthropic_token_env"
+
+    cc_env_token = os.getenv("CLAUDE_CODE_OAUTH_TOKEN", "").strip()
+    if cc_env_token and cc_env_token == token:
+        return "claude_code_oauth_token_env"
+
+    creds = read_claude_code_credentials()
+    if creds and creds.get("accessToken") == token:
+        return str(creds.get("source") or "claude_code_credentials")
+
+    managed_key = read_claude_managed_key()
+    if managed_key and managed_key == token:
+        return "claude_json_primary_api_key"
+
+    api_key = os.getenv("ANTHROPIC_API_KEY", "").strip()
+    if api_key and api_key == token:
+        return "anthropic_api_key_env"
+
+    return "unknown"
+
+
 def resolve_anthropic_token() -> Optional[str]:
    """Resolve an Anthropic token from all available sources.

@@ -807,6 +720,21 @@ def run_hermes_oauth_login_pure() -> Optional[Dict[str, Any]]:
    }


+def _save_hermes_oauth_credentials(access_token: str, refresh_token: str, expires_at_ms: int) -> None:
+    """Save OAuth credentials to ~/.hermes/.anthropic_oauth.json."""
+    data = {
+        "accessToken": access_token,
+        "refreshToken": refresh_token,
+        "expiresAt": expires_at_ms,
+    }
+    try:
+        _HERMES_OAUTH_FILE.parent.mkdir(parents=True, exist_ok=True)
+        _HERMES_OAUTH_FILE.write_text(json.dumps(data, indent=2), encoding="utf-8")
+        _HERMES_OAUTH_FILE.chmod(0o600)
+    except (OSError, IOError) as e:
+        logger.debug("Failed to save Hermes OAuth credentials: %s", e)
+
+
 def read_hermes_oauth_credentials() -> Optional[Dict[str, Any]]:
    """Read Hermes-managed OAuth credentials from ~/.hermes/.anthropic_oauth.json."""
    if _HERMES_OAUTH_FILE.exists():
@@ -855,6 +783,39 @@ def _sanitize_tool_id(tool_id: str) -> str:
    return sanitized or "tool_0"


+def _convert_openai_image_part_to_anthropic(part: Dict[str, Any]) -> Optional[Dict[str, Any]]:
+    """Convert an OpenAI-style image block to Anthropic's image source format."""
+    image_data = part.get("image_url", {})
+    url = image_data.get("url", "") if isinstance(image_data, dict) else str(image_data)
+    if not isinstance(url, str) or not url.strip():
+        return None
+    url = url.strip()
+
+    if url.startswith("data:"):
+        header, sep, data = url.partition(",")
+        if sep and ";base64" in header:
+            media_type = header[5:].split(";", 1)[0] or "image/png"
+            return {
+                "type": "image",
+                "source": {
+                    "type": "base64",
+                    "media_type": media_type,
+                    "data": data,
+                },
+            }
+
+    if url.startswith(("http://", "https://")):
+        return {
+            "type": "image",
+            "source": {
+                "type": "url",
+                "url": url,
+            },
+        }
+
+    return None
+
+
 def convert_tools_to_anthropic(tools: List[Dict]) -> List[Dict]:
    """Convert OpenAI tool definitions to Anthropic format."""
    if not tools:
@@ -1274,31 +1235,13 @@ def build_anthropic_kwargs(
    preserve_dots: bool = False,
    context_length: Optional[int] = None,
    base_url: str | None = None,
-    fast_mode: bool = False,
 ) -> Dict[str, Any]:
    """Build kwargs for anthropic.messages.create().

-    Naming note — two distinct concepts, easily confused:
-      max_tokens     = OUTPUT token cap for a single response.
-                       Anthropic's API calls this "max_tokens" but it only
-                       limits the *output*.  Anthropic's own native SDK
-                       renamed it "max_output_tokens" for clarity.
-      context_length = TOTAL context window (input tokens + output tokens).
-                       The API enforces: input_tokens + max_tokens ≤ context_length.
-                       Stored on the ContextCompressor; reduced on overflow errors.
-
-    When *max_tokens* is None the model's native output ceiling is used
-    (e.g. 128K for Opus 4.6, 64K for Sonnet 4.6).
-
-    When *context_length* is provided and the model's native output ceiling
-    exceeds it (e.g. a local endpoint with an 8K window), the output cap is
-    clamped to context_length − 1.  This only kicks in for unusually small
-    context windows; for full-size models the native output cap is always
-    smaller than the context window so no clamping happens.
-    NOTE: this clamping does not account for prompt size — if the prompt is
-    large, Anthropic may still reject the request.  The caller must detect
-    "max_tokens too large given prompt" errors and retry with a smaller cap
-    (see parse_available_output_tokens_from_error + _ephemeral_max_output_tokens).
+    When *max_tokens* is None, the model's native output limit is used
+    (e.g. 128K for Opus 4.6, 64K for Sonnet 4.6).  If *context_length*
+    is provided, the effective limit is clamped so it doesn't exceed
+    the context window.

    When *is_oauth* is True, applies Claude Code compatibility transforms:
    system prompt prefix, tool name prefixing, and prompt sanitization.
@@ -1308,24 +1251,15 @@ def build_anthropic_kwargs(

    When *base_url* points to a third-party Anthropic-compatible endpoint,
    thinking block signatures are stripped (they are Anthropic-proprietary).
-
-    When *fast_mode* is True, adds ``extra_body["speed"] = "fast"`` and the
-    fast-mode beta header for ~2.5x faster output throughput on Opus 4.6.
-    Currently only supported on native Anthropic endpoints (not third-party
-    compatible ones).
    """
    system, anthropic_messages = convert_messages_to_anthropic(messages, base_url=base_url)
    anthropic_tools = convert_tools_to_anthropic(tools) if tools else []

    model = normalize_model_name(model, preserve_dots=preserve_dots)
-    # effective_max_tokens = output cap for this call (≠ total context window)
    effective_max_tokens = max_tokens or _get_anthropic_max_output(model)

-    # Clamp output cap to fit inside the total context window.
-    # Only matters for small custom endpoints where context_length < native
-    # output ceiling.  For standard Anthropic models context_length (e.g.
-    # 200K) is always larger than the output ceiling (e.g. 128K), so this
-    # branch is not taken.
+    # Clamp to context window if the user set a lower context_length
+    # (e.g. custom endpoint with limited capacity).
    if context_length and effective_max_tokens > context_length:
        effective_max_tokens = max(context_length - 1, 1)

@@ -1393,31 +1327,17 @@ def build_anthropic_kwargs(
            kwargs["tool_choice"] = {"type": "tool", "name": tool_choice}

    # Map reasoning_config to Anthropic's thinking parameter.
-    # Claude 4.6+ models use adaptive thinking + output_config.effort.
+    # Claude 4.6 models use adaptive thinking + output_config.effort.
    # Older models use manual thinking with budget_tokens.
-    # MiniMax Anthropic-compat endpoints support thinking (manual mode only,
-    # not adaptive).  Haiku does NOT support extended thinking — skip entirely.
-    #
-    # On 4.7+ the `thinking.display` field defaults to "omitted", which
-    # silently hides reasoning text that Hermes surfaces in its CLI. We
-    # request "summarized" so the reasoning blocks stay populated — matching
-    # 4.6 behavior and preserving the activity-feed UX during long tool runs.
+    # Haiku and MiniMax models do NOT support extended thinking — skip entirely.
    if reasoning_config and isinstance(reasoning_config, dict):
-        if reasoning_config.get("enabled") is not False and "haiku" not in model.lower():
+        if reasoning_config.get("enabled") is not False and "haiku" not in model.lower() and "minimax" not in model.lower():
            effort = str(reasoning_config.get("effort", "medium")).lower()
            budget = THINKING_BUDGET.get(effort, 8000)
            if _supports_adaptive_thinking(model):
-                kwargs["thinking"] = {
-                    "type": "adaptive",
-                    "display": "summarized",
-                }
-                adaptive_effort = ADAPTIVE_EFFORT_MAP.get(effort, "medium")
-                # Downgrade xhigh→max on models that don't list xhigh as a
-                # supported level (Opus/Sonnet 4.6). Opus 4.7+ keeps xhigh.
-                if adaptive_effort == "xhigh" and not _supports_xhigh_effort(model):
-                    adaptive_effort = "max"
+                kwargs["thinking"] = {"type": "adaptive"}
                kwargs["output_config"] = {
-                    "effort": adaptive_effort,
+                    "effort": ADAPTIVE_EFFORT_MAP.get(effort, "medium")
                }
            else:
                kwargs["thinking"] = {"type": "enabled", "budget_tokens": budget}
@@ -1425,29 +1345,6 @@ def build_anthropic_kwargs(
                kwargs["temperature"] = 1
                kwargs["max_tokens"] = max(effective_max_tokens, budget + 4096)

-    # ── Strip sampling params on 4.7+ ─────────────────────────────────
-    # Opus 4.7 rejects any non-default temperature/top_p/top_k with a 400.
-    # Callers (auxiliary_client, flush_memories, etc.) may set these for
-    # older models; drop them here as a safety net so upstream 4.6 → 4.7
-    # migrations don't require coordinated edits everywhere.
-    if _forbids_sampling_params(model):
-        for _sampling_key in ("temperature", "top_p", "top_k"):
-            kwargs.pop(_sampling_key, None)
-
-    # ── Fast mode (Opus 4.6 only) ────────────────────────────────────
-    # Adds extra_body.speed="fast" + the fast-mode beta header for ~2.5x
-    # output speed. Only for native Anthropic endpoints — third-party
-    # providers would reject the unknown beta header and speed parameter.
-    if fast_mode and not _is_third_party_anthropic_endpoint(base_url):
-        kwargs.setdefault("extra_body", {})["speed"] = "fast"
-        # Build extra_headers with ALL applicable betas (the per-request
-        # extra_headers override the client-level anthropic-beta header).
-        betas = list(_common_betas_for_base_url(base_url))
-        if is_oauth:
-            betas.extend(_OAUTH_ONLY_BETAS)
-        betas.append(_FAST_MODE_BETA)
-        kwargs["extra_headers"] = {"anthropic-beta": ",".join(betas)}
-
    return kwargs


@@ -1491,20 +1388,12 @@ def normalize_anthropic_response(
                )
            )

-    # Map Anthropic stop_reason to OpenAI finish_reason.
-    # Newer stop reasons added in Claude 4.5+ / 4.7:
-    #   - refusal: the model declined to answer (cyber safeguards, CSAM, etc.)
-    #   - model_context_window_exceeded: hit context limit (not max_tokens)
-    # Both need distinct handling upstream — a refusal should surface to the
-    # user with a clear message, and a context-window overflow should trigger
-    # compression/truncation rather than be treated as normal end-of-turn.
+    # Map Anthropic stop_reason to OpenAI finish_reason
    stop_reason_map = {
        "end_turn": "stop",
        "tool_use": "tool_calls",
        "max_tokens": "length",
        "stop_sequence": "stop",
-        "refusal": "content_filter",
-        "model_context_window_exceeded": "length",
    }
    finish_reason = stop_reason_map.get(response.stop_reason, "stop")

@@ -1517,4 +1406,4 @@ def normalize_anthropic_response(
            reasoning_details=reasoning_details or None,
        ),
        finish_reason,
-    )
+    )
@@ -0,0 +1,114 @@
+"""BuiltinMemoryProvider — wraps MEMORY.md / USER.md as a MemoryProvider.
+
+Always registered as the first provider. Cannot be disabled or removed.
+This is the existing Hermes memory system exposed through the provider
+interface for compatibility with the MemoryManager.
+
+The actual storage logic lives in tools/memory_tool.py (MemoryStore).
+This provider is a thin adapter that delegates to MemoryStore and
+exposes the memory tool schema.
+"""
+
+from __future__ import annotations
+
+import json
+import logging
+from typing import Any, Dict, List
+
+from agent.memory_provider import MemoryProvider
+from tools.registry import tool_error
+
+logger = logging.getLogger(__name__)
+
+
+class BuiltinMemoryProvider(MemoryProvider):
+    """Built-in file-backed memory (MEMORY.md + USER.md).
+
+    Always active, never disabled by other providers. The `memory` tool
+    is handled by run_agent.py's agent-level tool interception (not through
+    the normal registry), so get_tool_schemas() returns an empty list —
+    the memory tool is already wired separately.
+    """
+
+    def __init__(
+        self,
+        memory_store=None,
+        memory_enabled: bool = False,
+        user_profile_enabled: bool = False,
+    ):
+        self._store = memory_store
+        self._memory_enabled = memory_enabled
+        self._user_profile_enabled = user_profile_enabled
+
+    @property
+    def name(self) -> str:
+        return "builtin"
+
+    def is_available(self) -> bool:
+        """Built-in memory is always available."""
+        return True
+
+    def initialize(self, session_id: str, **kwargs) -> None:
+        """Load memory from disk if not already loaded."""
+        if self._store is not None:
+            self._store.load_from_disk()
+
+    def system_prompt_block(self) -> str:
+        """Return MEMORY.md and USER.md content for the system prompt.
+
+        Uses the frozen snapshot captured at load time. This ensures the
+        system prompt stays stable throughout a session (preserving the
+        prompt cache), even though the live entries may change via tool calls.
+        """
+        if not self._store:
+            return ""
+
+        parts = []
+        if self._memory_enabled:
+            mem_block = self._store.format_for_system_prompt("memory")
+            if mem_block:
+                parts.append(mem_block)
+        if self._user_profile_enabled:
+            user_block = self._store.format_for_system_prompt("user")
+            if user_block:
+                parts.append(user_block)
+
+        return "\n\n".join(parts)
+
+    def prefetch(self, query: str, *, session_id: str = "") -> str:
+        """Built-in memory doesn't do query-based recall — it's injected via system_prompt_block."""
+        return ""
+
+    def sync_turn(self, user_content: str, assistant_content: str, *, session_id: str = "") -> None:
+        """Built-in memory doesn't auto-sync turns — writes happen via the memory tool."""
+
+    def get_tool_schemas(self) -> List[Dict[str, Any]]:
+        """Return empty list.
+
+        The `memory` tool is an agent-level intercepted tool, handled
+        specially in run_agent.py before normal tool dispatch. It's not
+        part of the standard tool registry. We don't duplicate it here.
+        """
+        return []
+
+    def handle_tool_call(self, tool_name: str, args: Dict[str, Any], **kwargs) -> str:
+        """Not used — the memory tool is intercepted in run_agent.py."""
+        return tool_error("Built-in memory tool is handled by the agent loop")
+
+    def shutdown(self) -> None:
+        """No cleanup needed — files are saved on every write."""
+
+    # -- Property access for backward compatibility --------------------------
+
+    @property
+    def store(self):
+        """Access the underlying MemoryStore for legacy code paths."""
+        return self._store
+
+    @property
+    def memory_enabled(self) -> bool:
+        return self._memory_enabled
+
+    @property
+    def user_profile_enabled(self) -> bool:
+        return self._user_profile_enabled
@@ -1,184 +0,0 @@
-"""Abstract base class for pluggable context engines.
-
-A context engine controls how conversation context is managed when
-approaching the model's token limit. The built-in ContextCompressor
-is the default implementation. Third-party engines (e.g. LCM) can
-replace it via the plugin system or by being placed in the
-``plugins/context_engine/<name>/`` directory.
-
-Selection is config-driven: ``context.engine`` in config.yaml.
-Default is ``"compressor"`` (the built-in). Only one engine is active.
-
-The engine is responsible for:
-  - Deciding when compaction should fire
-  - Performing compaction (summarization, DAG construction, etc.)
-  - Optionally exposing tools the agent can call (e.g. lcm_grep)
-  - Tracking token usage from API responses
-
-Lifecycle:
-  1. Engine is instantiated and registered (plugin register() or default)
-  2. on_session_start() called when a conversation begins
-  3. update_from_response() called after each API response with usage data
-  4. should_compress() checked after each turn
-  5. compress() called when should_compress() returns True
-  6. on_session_end() called at real session boundaries (CLI exit, /reset,
-     gateway session expiry) — NOT per-turn
-"""
-
-from abc import ABC, abstractmethod
-from typing import Any, Dict, List
-
-
-class ContextEngine(ABC):
-    """Base class all context engines must implement."""
-
-    # -- Identity ----------------------------------------------------------
-
-    @property
-    @abstractmethod
-    def name(self) -> str:
-        """Short identifier (e.g. 'compressor', 'lcm')."""
-
-    # -- Token state (read by run_agent.py for display/logging) ------------
-    #
-    # Engines MUST maintain these. run_agent.py reads them directly.
-
-    last_prompt_tokens: int = 0
-    last_completion_tokens: int = 0
-    last_total_tokens: int = 0
-    threshold_tokens: int = 0
-    context_length: int = 0
-    compression_count: int = 0
-
-    # -- Compaction parameters (read by run_agent.py for preflight) --------
-    #
-    # These control the preflight compression check.  Subclasses may
-    # override via __init__ or property; defaults are sensible for most
-    # engines.
-
-    threshold_percent: float = 0.75
-    protect_first_n: int = 3
-    protect_last_n: int = 6
-
-    # -- Core interface ----------------------------------------------------
-
-    @abstractmethod
-    def update_from_response(self, usage: Dict[str, Any]) -> None:
-        """Update tracked token usage from an API response.
-
-        Called after every LLM call with the usage dict from the response.
-        """
-
-    @abstractmethod
-    def should_compress(self, prompt_tokens: int = None) -> bool:
-        """Return True if compaction should fire this turn."""
-
-    @abstractmethod
-    def compress(
-        self,
-        messages: List[Dict[str, Any]],
-        current_tokens: int = None,
-    ) -> List[Dict[str, Any]]:
-        """Compact the message list and return the new message list.
-
-        This is the main entry point. The engine receives the full message
-        list and returns a (possibly shorter) list that fits within the
-        context budget. The implementation is free to summarize, build a
-        DAG, or do anything else — as long as the returned list is a valid
-        OpenAI-format message sequence.
-        """
-
-    # -- Optional: pre-flight check ----------------------------------------
-
-    def should_compress_preflight(self, messages: List[Dict[str, Any]]) -> bool:
-        """Quick rough check before the API call (no real token count yet).
-
-        Default returns False (skip pre-flight). Override if your engine
-        can do a cheap estimate.
-        """
-        return False
-
-    # -- Optional: session lifecycle ---------------------------------------
-
-    def on_session_start(self, session_id: str, **kwargs) -> None:
-        """Called when a new conversation session begins.
-
-        Use this to load persisted state (DAG, store) for the session.
-        kwargs may include hermes_home, platform, model, etc.
-        """
-
-    def on_session_end(self, session_id: str, messages: List[Dict[str, Any]]) -> None:
-        """Called at real session boundaries (CLI exit, /reset, gateway expiry).
-
-        Use this to flush state, close DB connections, etc.
-        NOT called per-turn — only when the session truly ends.
-        """
-
-    def on_session_reset(self) -> None:
-        """Called on /new or /reset. Reset per-session state.
-
-        Default resets compression_count and token tracking.
-        """
-        self.last_prompt_tokens = 0
-        self.last_completion_tokens = 0
-        self.last_total_tokens = 0
-        self.compression_count = 0
-
-    # -- Optional: tools ---------------------------------------------------
-
-    def get_tool_schemas(self) -> List[Dict[str, Any]]:
-        """Return tool schemas this engine provides to the agent.
-
-        Default returns empty list (no tools). LCM would return schemas
-        for lcm_grep, lcm_describe, lcm_expand here.
-        """
-        return []
-
-    def handle_tool_call(self, name: str, args: Dict[str, Any], **kwargs) -> str:
-        """Handle a tool call from the agent.
-
-        Only called for tool names returned by get_tool_schemas().
-        Must return a JSON string.
-
-        kwargs may include:
-          messages: the current in-memory message list (for live ingestion)
-        """
-        import json
-        return json.dumps({"error": f"Unknown context engine tool: {name}"})
-
-    # -- Optional: status / display ----------------------------------------
-
-    def get_status(self) -> Dict[str, Any]:
-        """Return status dict for display/logging.
-
-        Default returns the standard fields run_agent.py expects.
-        """
-        return {
-            "last_prompt_tokens": self.last_prompt_tokens,
-            "threshold_tokens": self.threshold_tokens,
-            "context_length": self.context_length,
-            "usage_percent": (
-                min(100, self.last_prompt_tokens / self.context_length * 100)
-                if self.context_length else 0
-            ),
-            "compression_count": self.compression_count,
-        }
-
-    # -- Optional: model switch support ------------------------------------
-
-    def update_model(
-        self,
-        model: str,
-        context_length: int,
-        base_url: str = "",
-        api_key: str = "",
-        provider: str = "",
-    ) -> None:
-        """Called when the user switches models or on fallback activation.
-
-        Default updates context_length and recalculates threshold_tokens
-        from threshold_percent. Override if your engine needs more
-        (e.g. recalculate DAG budgets, switch summary models).
-        """
-        self.context_length = context_length
-        self.threshold_tokens = int(context_length * self.threshold_percent)
@@ -13,9 +13,8 @@ from typing import Awaitable, Callable

 from agent.model_metadata import estimate_tokens_rough

-_QUOTED_REFERENCE_VALUE = r'(?:`[^`\n]+`|"[^"\n]+"|\'[^\'\n]+\')'
 REFERENCE_PATTERN = re.compile(
-    rf"(?<![\w/])@(?:(?P<simple>diff|staged)\b|(?P<kind>file|folder|git|url):(?P<value>{_QUOTED_REFERENCE_VALUE}(?::\d+(?:-\d+)?)?|\S+))"
+    r"(?<![\w/])@(?:(?P<simple>diff|staged)\b|(?P<kind>file|folder|git|url):(?P<value>\S+))"
 )
 TRAILING_PUNCTUATION = ",.;!?"
 _SENSITIVE_HOME_DIRS = (".ssh", ".aws", ".gnupg", ".kube", ".docker", ".azure", ".config/gh")
@@ -82,10 +81,14 @@ def parse_context_references(message: str) -> list[ContextReference]:
        value = _strip_trailing_punctuation(match.group("value") or "")
        line_start = None
        line_end = None
-        target = _strip_reference_wrappers(value)
+        target = value

        if kind == "file":
-            target, line_start, line_end = _parse_file_reference_value(value)
+            range_match = re.match(r"^(?P<path>.+?):(?P<start>\d+)(?:-(?P<end>\d+))?$", value)
+            if range_match:
+                target = range_match.group("path")
+                line_start = int(range_match.group("start"))
+                line_end = int(range_match.group("end") or range_match.group("start"))

        refs.append(
            ContextReference(
@@ -372,38 +375,6 @@ def _strip_trailing_punctuation(value: str) -> str:
    return stripped


-def _strip_reference_wrappers(value: str) -> str:
-    if len(value) >= 2 and value[0] == value[-1] and value[0] in "`\"'":
-        return value[1:-1]
-    return value
-
-
-def _parse_file_reference_value(value: str) -> tuple[str, int | None, int | None]:
-    quoted_match = re.match(
-        r'^(?P<quote>`|"|\')(?P<path>.+?)(?P=quote)(?::(?P<start>\d+)(?:-(?P<end>\d+))?)?$',
-        value,
-    )
-    if quoted_match:
-        line_start = quoted_match.group("start")
-        line_end = quoted_match.group("end")
-        return (
-            quoted_match.group("path"),
-            int(line_start) if line_start is not None else None,
-            int(line_end or line_start) if line_start is not None else None,
-        )
-
-    range_match = re.match(r"^(?P<path>.+?):(?P<start>\d+)(?:-(?P<end>\d+))?$", value)
-    if range_match:
-        line_start = int(range_match.group("start"))
-        return (
-            range_match.group("path"),
-            line_start,
-            int(range_match.group("end") or range_match.group("start")),
-        )
-
-    return _strip_reference_wrappers(value), None, None
-
-
 def _remove_reference_tokens(message: str, refs: list[ContextReference]) -> str:
    pieces: list[str] = []
    cursor = 0
@@ -313,25 +313,9 @@ class CopilotACPClient:
            tools=tools,
            tool_choice=tool_choice,
        )
-        # Normalise timeout: run_agent.py may pass an httpx.Timeout object
-        # (used natively by the OpenAI SDK) rather than a plain float.
-        if timeout is None:
-            _effective_timeout = _DEFAULT_TIMEOUT_SECONDS
-        elif isinstance(timeout, (int, float)):
-            _effective_timeout = float(timeout)
-        else:
-            # httpx.Timeout or similar — pick the largest component so the
-            # subprocess has enough wall-clock time for the full response.
-            _candidates = [
-                getattr(timeout, attr, None)
-                for attr in ("read", "write", "connect", "pool", "timeout")
-            ]
-            _numeric = [float(v) for v in _candidates if isinstance(v, (int, float))]
-            _effective_timeout = max(_numeric) if _numeric else _DEFAULT_TIMEOUT_SECONDS
-
        response_text, reasoning_text = self._run_prompt(
            prompt_text,
-            timeout_seconds=_effective_timeout,
+            timeout_seconds=float(timeout or _DEFAULT_TIMEOUT_SECONDS),
        )

        tool_calls, cleaned_text = _extract_tool_calls_from_text(response_text)
@@ -19,15 +19,12 @@ from hermes_cli.auth import (
    CODEX_ACCESS_TOKEN_REFRESH_SKEW_SECONDS,
    DEFAULT_AGENT_KEY_MIN_TTL_SECONDS,
    PROVIDER_REGISTRY,
-    _auth_store_lock,
    _codex_access_token_is_expiring,
    _decode_jwt_claims,
+    _import_codex_cli_tokens,
    _load_auth_store,
    _load_provider_state,
-    _resolve_kimi_base_url,
    _resolve_zai_base_url,
-    _save_auth_store,
-    _save_provider_state,
    read_credential_pool,
    write_credential_pool,
 )
@@ -67,10 +64,10 @@ SUPPORTED_POOL_STRATEGIES = {
 }

 # Cooldown before retrying an exhausted credential.
-# 429 (rate-limited) and 402 (billing/quota) both cool down after 1 hour.
-# Provider-supplied reset_at timestamps override these defaults.
+# 429 (rate-limited) cools down faster since quotas reset frequently.
+# 402 (billing/quota) and other codes use a longer default.
 EXHAUSTED_TTL_429_SECONDS = 60 * 60          # 1 hour
-EXHAUSTED_TTL_DEFAULT_SECONDS = 60 * 60      # 1 hour
+EXHAUSTED_TTL_DEFAULT_SECONDS = 24 * 60 * 60 # 24 hours

 # Pool key prefix for custom OpenAI-compatible endpoints.
 # Custom endpoints all share provider='custom' but are keyed by their
@@ -286,14 +283,6 @@ def _iter_custom_providers(config: Optional[dict] = None):
        return
    custom_providers = config.get("custom_providers")
    if not isinstance(custom_providers, list):
-        # Fall back to the v12+ providers dict via the compatibility layer
-        try:
-            from hermes_cli.config import get_compatible_custom_providers
-
-            custom_providers = get_compatible_custom_providers(config)
-        except Exception:
-            return
-    if not custom_providers:
        return
    for entry in custom_providers:
        if not isinstance(entry, dict):
@@ -455,66 +444,38 @@ class CredentialPool:
            logger.debug("Failed to sync from credentials file: %s", exc)
        return entry

-    def _sync_device_code_entry_to_auth_store(self, entry: PooledCredential) -> None:
-        """Write refreshed pool entry tokens back to auth.json providers.
+    def _sync_codex_entry_from_cli(self, entry: PooledCredential) -> PooledCredential:
+        """Sync an openai-codex pool entry from ~/.codex/auth.json if tokens differ.

-        After a pool-level refresh, the pool entry has fresh tokens but
-        auth.json's ``providers.<id>`` still holds the pre-refresh state.
-        On the next ``load_pool()``, ``_seed_from_singletons()`` reads that
-        stale state and can overwrite the fresh pool entry — potentially
-        re-seeding a consumed single-use refresh token.
-
-        Applies to any OAuth provider whose singleton lives in auth.json
-        (currently Nous and OpenAI Codex).
+        OpenAI OAuth refresh tokens are single-use and rotate on every refresh.
+        When the Codex CLI (or another Hermes profile) refreshes its token,
+        the pool entry's refresh_token becomes stale.  This method detects that
+        by comparing against ~/.codex/auth.json and syncing the fresh pair.
        """
-        if entry.source != "device_code":
-            return
+        if self.provider != "openai-codex":
+            return entry
        try:
-            with _auth_store_lock():
-                auth_store = _load_auth_store()
-                if self.provider == "nous":
-                    state = _load_provider_state(auth_store, "nous")
-                    if state is None:
-                        return
-                    state["access_token"] = entry.access_token
-                    if entry.refresh_token:
-                        state["refresh_token"] = entry.refresh_token
-                    if entry.expires_at:
-                        state["expires_at"] = entry.expires_at
-                    if entry.agent_key:
-                        state["agent_key"] = entry.agent_key
-                    if entry.agent_key_expires_at:
-                        state["agent_key_expires_at"] = entry.agent_key_expires_at
-                    for extra_key in ("obtained_at", "expires_in", "agent_key_id",
-                                      "agent_key_expires_in", "agent_key_reused",
-                                      "agent_key_obtained_at"):
-                        val = entry.extra.get(extra_key)
-                        if val is not None:
-                            state[extra_key] = val
-                    if entry.inference_base_url:
-                        state["inference_base_url"] = entry.inference_base_url
-                    _save_provider_state(auth_store, "nous", state)
-
-                elif self.provider == "openai-codex":
-                    state = _load_provider_state(auth_store, "openai-codex")
-                    if not isinstance(state, dict):
-                        return
-                    tokens = state.get("tokens")
-                    if not isinstance(tokens, dict):
-                        return
-                    tokens["access_token"] = entry.access_token
-                    if entry.refresh_token:
-                        tokens["refresh_token"] = entry.refresh_token
-                    if entry.last_refresh:
-                        state["last_refresh"] = entry.last_refresh
-                    _save_provider_state(auth_store, "openai-codex", state)
-
-                else:
-                    return
-
-                _save_auth_store(auth_store)
+            cli_tokens = _import_codex_cli_tokens()
+            if not cli_tokens:
+                return entry
+            cli_refresh = cli_tokens.get("refresh_token", "")
+            cli_access = cli_tokens.get("access_token", "")
+            if cli_refresh and cli_refresh != entry.refresh_token:
+                logger.debug("Pool entry %s: syncing tokens from ~/.codex/auth.json (refresh token changed)", entry.id)
+                updated = replace(
+                    entry,
+                    access_token=cli_access,
+                    refresh_token=cli_refresh,
+                    last_status=None,
+                    last_status_at=None,
+                    last_error_code=None,
+                )
+                self._replace_entry(entry, updated)
+                self._persist()
+                return updated
        except Exception as exc:
-            logger.debug("Failed to sync %s pool entry back to auth store: %s", self.provider, exc)
+            logger.debug("Failed to sync from ~/.codex/auth.json: %s", exc)
+        return entry

    def _refresh_entry(self, entry: PooledCredential, *, force: bool) -> Optional[PooledCredential]:
        if entry.auth_type != AUTH_TYPE_OAUTH or not entry.refresh_token:
@@ -649,10 +610,6 @@ class CredentialPool:
        )
        self._replace_entry(entry, updated)
        self._persist()
-        # Sync refreshed tokens back to auth.json providers so that
-        # _seed_from_singletons() on the next load_pool() sees fresh state
-        # instead of re-seeding stale/consumed tokens.
-        self._sync_device_code_entry_to_auth_store(updated)
        return updated

    def _entry_needs_refresh(self, entry: PooledCredential) -> bool:
@@ -674,6 +631,17 @@ class CredentialPool:
            return False
        return False

+    def mark_used(self, entry_id: Optional[str] = None) -> None:
+        """Increment request_count for tracking. Used by least_used strategy."""
+        target_id = entry_id or self._current_id
+        if not target_id:
+            return
+        with self._lock:
+            for idx, entry in enumerate(self._entries):
+                if entry.id == target_id:
+                    self._entries[idx] = replace(entry, request_count=entry.request_count + 1)
+                    return
+
    def select(self) -> Optional[PooledCredential]:
        with self._lock:
            return self._select_unlocked()
@@ -698,6 +666,16 @@ class CredentialPool:
                if synced is not entry:
                    entry = synced
                    cleared_any = True
+            # For openai-codex entries, sync from ~/.codex/auth.json before
+            # any status/refresh checks.  This picks up tokens refreshed by
+            # the Codex CLI or another Hermes profile.
+            if (self.provider == "openai-codex"
+                    and entry.last_status == STATUS_EXHAUSTED
+                    and entry.refresh_token):
+                synced = self._sync_codex_entry_from_cli(entry)
+                if synced is not entry:
+                    entry = synced
+                    cleared_any = True
            if entry.last_status == STATUS_EXHAUSTED:
                exhausted_until = _exhausted_until(entry)
                if exhausted_until is not None and now < exhausted_until:
@@ -825,6 +803,11 @@ class CredentialPool:
            else:
                self._active_leases[credential_id] = count - 1

+    def active_lease_count(self, credential_id: str) -> int:
+        """Return the number of active leases for a credential."""
+        with self._lock:
+            return self._active_leases.get(credential_id, 0)
+
    def try_refresh_current(self) -> Optional[PooledCredential]:
        with self._lock:
            return self._try_refresh_current_unlocked()
@@ -984,17 +967,6 @@ def _seed_from_singletons(provider: str, entries: List[PooledCredential]) -> Tup
    auth_store = _load_auth_store()

    if provider == "anthropic":
-        # Only auto-discover external credentials (Claude Code, Hermes PKCE)
-        # when the user has explicitly configured anthropic as their provider.
-        # Without this gate, auxiliary client fallback chains silently read
-        # ~/.claude/.credentials.json without user consent.  See PR #4210.
-        try:
-            from hermes_cli.auth import is_provider_explicitly_configured
-            if not is_provider_explicitly_configured("anthropic"):
-                return changed, active_sources
-        except ImportError:
-            pass
-
        from agent.anthropic_adapter import read_claude_code_credentials, read_hermes_oauth_credentials

        for source_name, creds in (
@@ -1002,13 +974,6 @@ def _seed_from_singletons(provider: str, entries: List[PooledCredential]) -> Tup
            ("claude_code", read_claude_code_credentials()),
        ):
            if creds and creds.get("accessToken"):
-                # Check if user explicitly removed this source
-                try:
-                    from hermes_cli.auth import is_source_suppressed
-                    if is_source_suppressed(provider, source_name):
-                        continue
-                except ImportError:
-                    pass
                active_sources.add(source_name)
                changed |= _upsert_entry(
                    entries,
@@ -1028,14 +993,6 @@ def _seed_from_singletons(provider: str, entries: List[PooledCredential]) -> Tup
        state = _load_provider_state(auth_store, "nous")
        if state:
            active_sources.add("device_code")
-            # Prefer a user-supplied label embedded in the singleton state
-            # (set by persist_nous_credentials(label=...) when the user ran
-            # `hermes auth add nous --label <name>`).  Fall back to the
-            # auto-derived token fingerprint for logins that didn't supply one.
-            custom_label = str(state.get("label") or "").strip()
-            seeded_label = custom_label or label_from_token(
-                state.get("access_token", ""), "device_code"
-            )
            changed |= _upsert_entry(
                entries,
                provider,
@@ -1054,87 +1011,13 @@ def _seed_from_singletons(provider: str, entries: List[PooledCredential]) -> Tup
                    "agent_key": state.get("agent_key"),
                    "agent_key_expires_at": state.get("agent_key_expires_at"),
                    "tls": state.get("tls") if isinstance(state.get("tls"), dict) else None,
-                    "label": seeded_label,
+                    "label": label_from_token(state.get("access_token", ""), "device_code"),
                },
            )

-    elif provider == "copilot":
-        # Copilot tokens are resolved dynamically via `gh auth token` or
-        # env vars (COPILOT_GITHUB_TOKEN / GH_TOKEN).  They don't live in
-        # the auth store or credential pool, so we resolve them here.
-        try:
-            from hermes_cli.copilot_auth import resolve_copilot_token
-            token, source = resolve_copilot_token()
-            if token:
-                source_name = "gh_cli" if "gh" in source.lower() else f"env:{source}"
-                active_sources.add(source_name)
-                pconfig = PROVIDER_REGISTRY.get(provider)
-                changed |= _upsert_entry(
-                    entries,
-                    provider,
-                    source_name,
-                    {
-                        "source": source_name,
-                        "auth_type": AUTH_TYPE_API_KEY,
-                        "access_token": token,
-                        "base_url": pconfig.inference_base_url if pconfig else "",
-                        "label": source,
-                    },
-                )
-        except Exception as exc:
-            logger.debug("Copilot token seed failed: %s", exc)
-
-    elif provider == "qwen-oauth":
-        # Qwen OAuth tokens live in ~/.qwen/oauth_creds.json, written by
-        # the Qwen CLI (`qwen auth qwen-oauth`).  They aren't in the
-        # Hermes auth store or env vars, so resolve them here.
-        # Use refresh_if_expiring=False to avoid network calls during
-        # pool loading / provider discovery.
-        try:
-            from hermes_cli.auth import resolve_qwen_runtime_credentials
-            creds = resolve_qwen_runtime_credentials(refresh_if_expiring=False)
-            token = creds.get("api_key", "")
-            if token:
-                source_name = creds.get("source", "qwen-cli")
-                active_sources.add(source_name)
-                changed |= _upsert_entry(
-                    entries,
-                    provider,
-                    source_name,
-                    {
-                        "source": source_name,
-                        "auth_type": AUTH_TYPE_OAUTH,
-                        "access_token": token,
-                        "expires_at_ms": creds.get("expires_at_ms"),
-                        "base_url": creds.get("base_url", ""),
-                        "label": creds.get("auth_file", source_name),
-                    },
-                )
-        except Exception as exc:
-            logger.debug("Qwen OAuth token seed failed: %s", exc)
-
    elif provider == "openai-codex":
-        # Respect user suppression — `hermes auth remove openai-codex` marks
-        # the device_code source as suppressed so it won't be re-seeded from
-        # the Hermes auth store.  Without this gate the removal is instantly
-        # undone on the next load_pool() call.
-        codex_suppressed = False
-        try:
-            from hermes_cli.auth import is_source_suppressed
-            codex_suppressed = is_source_suppressed(provider, "device_code")
-        except ImportError:
-            pass
-        if codex_suppressed:
-            return changed, active_sources
-
        state = _load_provider_state(auth_store, "openai-codex")
        tokens = state.get("tokens") if isinstance(state, dict) else None
-        # Hermes owns its own Codex auth state — we do NOT auto-import from
-        # ~/.codex/auth.json at pool-load time.  OAuth refresh tokens are
-        # single-use, so sharing them with Codex CLI / VS Code causes
-        # refresh_token_reused race failures.  Users who want to adopt
-        # existing Codex CLI credentials get a one-time, explicit prompt
-        # via `hermes auth openai-codex`.
        if isinstance(tokens, dict) and tokens.get("access_token"):
            active_sources.add("device_code")
            changed |= _upsert_entry(
@@ -1201,9 +1084,7 @@ def _seed_from_env(provider: str, entries: List[PooledCredential]) -> Tuple[bool
        active_sources.add(source)
        auth_type = AUTH_TYPE_OAUTH if provider == "anthropic" and not token.startswith("sk-ant-api") else AUTH_TYPE_API_KEY
        base_url = env_url or pconfig.inference_base_url
-        if provider == "kimi-coding":
-            base_url = _resolve_kimi_base_url(token, pconfig.inference_base_url, env_url)
-        elif provider == "zai":
+        if provider == "zai":
            base_url = _resolve_zai_base_url(token, pconfig.inference_base_url, env_url)
        changed |= _upsert_entry(
            entries,
@@ -4,6 +4,7 @@ Pure display functions and classes with no AIAgent dependency.
 Used by AIAgent._execute_tool_calls for CLI feedback.
 """

+import json
 import logging
 import os
 import sys
@@ -13,8 +14,6 @@ from dataclasses import dataclass, field
 from difflib import unified_diff
 from pathlib import Path

-from utils import safe_json_loads
-
 # ANSI escape codes for coloring tool failure indicators
 _RED = "\033[31m"
 _RESET = "\033[0m"
@@ -22,67 +21,11 @@ _RESET = "\033[0m"
 logger = logging.getLogger(__name__)

 _ANSI_RESET = "\033[0m"
-
-# Diff colors — resolved lazily from the skin engine so they adapt
-# to light/dark themes.  Falls back to sensible defaults on import
-# failure.  We cache after first resolution for performance.
-_diff_colors_cached: dict[str, str] | None = None
-
-
-def _diff_ansi() -> dict[str, str]:
-    """Return ANSI escapes for diff display, resolved from the active skin."""
-    global _diff_colors_cached
-    if _diff_colors_cached is not None:
-        return _diff_colors_cached
-
-    # Defaults that work on dark terminals
-    dim = "\033[38;2;150;150;150m"
-    file_c = "\033[38;2;180;160;255m"
-    hunk = "\033[38;2;120;120;140m"
-    minus = "\033[38;2;255;255;255;48;2;120;20;20m"
-    plus = "\033[38;2;255;255;255;48;2;20;90;20m"
-
-    try:
-        from hermes_cli.skin_engine import get_active_skin
-        skin = get_active_skin()
-
-        def _hex_fg(key: str, fallback_rgb: tuple[int, int, int]) -> str:
-            h = skin.get_color(key, "")
-            if h and len(h) == 7 and h[0] == "#":
-                r, g, b = int(h[1:3], 16), int(h[3:5], 16), int(h[5:7], 16)
-                return f"\033[38;2;{r};{g};{b}m"
-            r, g, b = fallback_rgb
-            return f"\033[38;2;{r};{g};{b}m"
-
-        dim = _hex_fg("banner_dim", (150, 150, 150))
-        file_c = _hex_fg("session_label", (180, 160, 255))
-        hunk = _hex_fg("session_border", (120, 120, 140))
-        # minus/plus use background colors — derive from ui_error/ui_ok
-        err_h = skin.get_color("ui_error", "#ef5350")
-        ok_h = skin.get_color("ui_ok", "#4caf50")
-        if err_h and len(err_h) == 7:
-            er, eg, eb = int(err_h[1:3], 16), int(err_h[3:5], 16), int(err_h[5:7], 16)
-            # Use a dark tinted version as background
-            minus = f"\033[38;2;255;255;255;48;2;{max(er//2,20)};{max(eg//4,10)};{max(eb//4,10)}m"
-        if ok_h and len(ok_h) == 7:
-            or_, og, ob = int(ok_h[1:3], 16), int(ok_h[3:5], 16), int(ok_h[5:7], 16)
-            plus = f"\033[38;2;255;255;255;48;2;{max(or_//4,10)};{max(og//2,20)};{max(ob//4,10)}m"
-    except Exception:
-        pass
-
-    _diff_colors_cached = {
-        "dim": dim, "file": file_c, "hunk": hunk,
-        "minus": minus, "plus": plus,
-    }
-    return _diff_colors_cached
-
-
-# Module-level helpers — each call resolves from the active skin lazily.
-def _diff_dim():   return _diff_ansi()["dim"]
-def _diff_file():  return _diff_ansi()["file"]
-def _diff_hunk():  return _diff_ansi()["hunk"]
-def _diff_minus(): return _diff_ansi()["minus"]
-def _diff_plus():  return _diff_ansi()["plus"]
+_ANSI_DIM = "\033[38;2;150;150;150m"
+_ANSI_FILE = "\033[38;2;180;160;255m"
+_ANSI_HUNK = "\033[38;2;120;120;140m"
+_ANSI_MINUS = "\033[38;2;255;255;255;48;2;120;20;20m"
+_ANSI_PLUS = "\033[38;2;255;255;255;48;2;20;90;20m"
 _MAX_INLINE_DIFF_FILES = 6
 _MAX_INLINE_DIFF_LINES = 80

@@ -124,6 +67,26 @@ def _get_skin():
        return None


+def get_skin_faces(key: str, default: list) -> list:
+    """Get spinner face list from active skin, falling back to default."""
+    skin = _get_skin()
+    if skin:
+        faces = skin.get_spinner_list(key)
+        if faces:
+            return faces
+    return default
+
+
+def get_skin_verbs() -> list:
+    """Get thinking verbs from active skin."""
+    skin = _get_skin()
+    if skin:
+        verbs = skin.get_spinner_list("thinking_verbs")
+        if verbs:
+            return verbs
+    return KawaiiSpinner.THINKING_VERBS
+
+
 def get_skin_tool_prefix() -> str:
    """Get tool output prefix character from active skin."""
    skin = _get_skin()
@@ -367,8 +330,9 @@ def _result_succeeded(result: str | None) -> bool:
    """Conservatively detect whether a tool result represents success."""
    if not result:
        return False
-    data = safe_json_loads(result)
-    if data is None:
+    try:
+        data = json.loads(result)
+    except (json.JSONDecodeError, TypeError):
        return False
    if not isinstance(data, dict):
        return False
@@ -417,7 +381,10 @@ def extract_edit_diff(
 ) -> str | None:
    """Extract a unified diff from a file-edit tool result."""
    if tool_name == "patch" and result:
-        data = safe_json_loads(result)
+        try:
+            data = json.loads(result)
+        except (json.JSONDecodeError, TypeError):
+            data = None
        if isinstance(data, dict):
            diff = data.get("diff")
            if isinstance(diff, str) and diff.strip():
@@ -456,19 +423,19 @@ def _render_inline_unified_diff(diff: str) -> list[str]:
        if raw_line.startswith("+++ "):
            to_file = raw_line[4:].strip()
            if from_file or to_file:
-                rendered.append(f"{_diff_file()}{from_file or 'a/?'} → {to_file or 'b/?'}{_ANSI_RESET}")
+                rendered.append(f"{_ANSI_FILE}{from_file or 'a/?'} → {to_file or 'b/?'}{_ANSI_RESET}")
            continue
        if raw_line.startswith("@@"):
-            rendered.append(f"{_diff_hunk()}{raw_line}{_ANSI_RESET}")
+            rendered.append(f"{_ANSI_HUNK}{raw_line}{_ANSI_RESET}")
            continue
        if raw_line.startswith("-"):
-            rendered.append(f"{_diff_minus()}{raw_line}{_ANSI_RESET}")
+            rendered.append(f"{_ANSI_MINUS}{raw_line}{_ANSI_RESET}")
            continue
        if raw_line.startswith("+"):
-            rendered.append(f"{_diff_plus()}{raw_line}{_ANSI_RESET}")
+            rendered.append(f"{_ANSI_PLUS}{raw_line}{_ANSI_RESET}")
            continue
        if raw_line.startswith(" "):
-            rendered.append(f"{_diff_dim()}{raw_line}{_ANSI_RESET}")
+            rendered.append(f"{_ANSI_DIM}{raw_line}{_ANSI_RESET}")
            continue
        if raw_line:
            rendered.append(raw_line)
@@ -534,7 +501,7 @@ def _summarize_rendered_diff_sections(
        summary = f"… omitted {omitted_lines} diff line(s)"
        if omitted_files:
            summary += f" across {omitted_files} additional file(s)/section(s)"
-        rendered.append(f"{_diff_hunk()}{summary}{_ANSI_RESET}")
+        rendered.append(f"{_ANSI_HUNK}{summary}{_ANSI_RESET}")

    return rendered

@@ -600,45 +567,6 @@ class KawaiiSpinner:
        "analyzing", "computing", "synthesizing", "formulating", "brainstorming",
    ]

-    @classmethod
-    def get_waiting_faces(cls) -> list:
-        """Return waiting faces from the active skin, falling back to KAWAII_WAITING."""
-        try:
-            skin = _get_skin()
-            if skin:
-                faces = skin.spinner.get("waiting_faces", [])
-                if faces:
-                    return faces
-        except Exception:
-            pass
-        return cls.KAWAII_WAITING
-
-    @classmethod
-    def get_thinking_faces(cls) -> list:
-        """Return thinking faces from the active skin, falling back to KAWAII_THINKING."""
-        try:
-            skin = _get_skin()
-            if skin:
-                faces = skin.spinner.get("thinking_faces", [])
-                if faces:
-                    return faces
-        except Exception:
-            pass
-        return cls.KAWAII_THINKING
-
-    @classmethod
-    def get_thinking_verbs(cls) -> list:
-        """Return thinking verbs from the active skin, falling back to THINKING_VERBS."""
-        try:
-            skin = _get_skin()
-            if skin:
-                verbs = skin.spinner.get("thinking_verbs", [])
-                if verbs:
-                    return verbs
-        except Exception:
-            pass
-        return cls.THINKING_VERBS
-
    def __init__(self, message: str = "", spinner_type: str = 'dots', print_fn=None):
        self.message = message
        self.spinner_frames = self.SPINNERS.get(spinner_type, self.SPINNERS['dots'])
@@ -795,6 +723,46 @@ class KawaiiSpinner:
        return False


+# =========================================================================
+# Kawaii face arrays (used by AIAgent._execute_tool_calls for spinner text)
+# =========================================================================
+
+KAWAII_SEARCH = [
+    "♪(´ε` )", "(｡◕‿◕｡)", "ヾ(＾∇＾)", "(◕ᴗ◕✿)", "( ˘▽˘)っ",
+    "٩(◕‿◕｡)۶", "(✿◠‿◠)", "♪～(´ε｀ )", "(ノ´ヮ`)ノ*:・゚✧", "＼(◎o◎)／",
+]
+KAWAII_READ = [
+    "φ(゜▽゜*)♪", "( ˘▽˘)っ", "(⌐■_■)", "٩(｡•́‿•̀｡)۶", "(◕‿◕✿)",
+    "ヾ(＠⌒ー⌒＠)ノ", "(✧ω✧)", "♪(๑ᴖ◡ᴖ๑)♪", "(≧◡≦)", "( ´ ▽ ` )ノ",
+]
+KAWAII_TERMINAL = [
+    "ヽ(>∀<☆)ノ", "(ノ°∀°)ノ", "٩(^ᴗ^)۶", "ヾ(⌐■_■)ノ♪", "(•̀ᴗ•́)و",
+    "┗(＾0＾)┓", "(｀・ω・´)", "＼(￣▽￣)／", "(ง •̀_•́)ง", "ヽ(´▽`)/",
+]
+KAWAII_BROWSER = [
+    "(ノ°∀°)ノ", "(☞゚ヮ゚)☞", "( ͡° ͜ʖ ͡°)", "┌( ಠ_ಠ)┘", "(⊙_⊙)？",
+    "ヾ(•ω•`)o", "(￣ω￣)", "( ˇωˇ )", "(ᵔᴥᵔ)", "＼(◎o◎)／",
+]
+KAWAII_CREATE = [
+    "✧*。٩(ˊᗜˋ*)و✧", "(ﾉ◕ヮ◕)ﾉ*:・ﾟ✧", "ヽ(>∀<☆)ノ", "٩(♡ε♡)۶", "(◕‿◕)♡",
+    "✿◕ ‿ ◕✿", "(*≧▽≦)", "ヾ(＾-＾)ノ", "(☆▽☆)", "°˖✧◝(⁰▿⁰)◜✧˖°",
+]
+KAWAII_SKILL = [
+    "ヾ(＠⌒ー⌒＠)ノ", "(๑˃ᴗ˂)ﻭ", "٩(◕‿◕｡)۶", "(✿╹◡╹)", "ヽ(・∀・)ノ",
+    "(ノ´ヮ`)ノ*:・ﾟ✧", "♪(๑ᴖ◡ᴖ๑)♪", "(◠‿◠)", "٩(ˊᗜˋ*)و", "(＾▽＾)",
+    "ヾ(＾∇＾)", "(★ω★)/", "٩(｡•́‿•̀｡)۶", "(◕ᴗ◕✿)", "＼(◎o◎)／",
+    "(✧ω✧)", "ヽ(>∀<☆)ノ", "( ˘▽˘)っ", "(≧◡≦) ♡", "ヾ(￣▽￣)",
+]
+KAWAII_THINK = [
+    "(っ°Д°;)っ", "(；′⌒`)", "(・_・ヾ", "( ´_ゝ`)", "(￣ヘ￣)",
+    "(。-`ω´-)", "( ˘︹˘ )", "(¬_¬)", "ヽ(ー_ー )ノ", "(；一_一)",
+]
+KAWAII_GENERIC = [
+    "♪(´ε` )", "(◕‿◕✿)", "ヾ(＾∇＾)", "٩(◕‿◕｡)۶", "(✿◠‿◠)",
+    "(ノ´ヮ`)ノ*:・ﾟ✧", "ヽ(>∀<☆)ノ", "(☆▽☆)", "( ˘▽˘)っ", "(≧◡≦)",
+]
+
+
 # =========================================================================
 # Cute tool message (completion line that replaces the spinner)
 # =========================================================================
@@ -810,19 +778,23 @@ def _detect_tool_failure(tool_name: str, result: str | None) -> tuple[bool, str]
        return False, ""

    if tool_name == "terminal":
-        data = safe_json_loads(result)
-        if isinstance(data, dict):
+        try:
+            data = json.loads(result)
            exit_code = data.get("exit_code")
            if exit_code is not None and exit_code != 0:
                return True, f" [exit {exit_code}]"
+        except (json.JSONDecodeError, TypeError, AttributeError):
+            logger.debug("Could not parse terminal result as JSON for exit code check")
        return False, ""

    # Memory-specific: distinguish "full" from real errors
    if tool_name == "memory":
-        data = safe_json_loads(result)
-        if isinstance(data, dict):
+        try:
+            data = json.loads(result)
            if data.get("success") is False and "exceed the limit" in data.get("error", ""):
                return True, " [full]"
+        except (json.JSONDecodeError, TypeError, AttributeError):
+            logger.debug("Could not parse memory result as JSON for capacity check")

    # Generic heuristic for non-terminal tools
    lower = result[:500].lower()
@@ -993,4 +965,100 @@ def get_cute_tool_message(
 # Honcho session line (one-liner with clickable OSC 8 hyperlink)
 # =========================================================================

+_DIM = "\033[2m"
+_SKY_BLUE = "\033[38;5;117m"
+_ANSI_RESET = "\033[0m"

+
+def honcho_session_url(workspace: str, session_name: str) -> str:
+    """Build a Honcho app URL for a session."""
+    from urllib.parse import quote
+    return (
+        f"https://app.honcho.dev/explore"
+        f"?workspace={quote(workspace, safe='')}"
+        f"&view=sessions"
+        f"&session={quote(session_name, safe='')}"
+    )
+
+
+def _osc8_link(url: str, text: str) -> str:
+    """OSC 8 terminal hyperlink (clickable in iTerm2, Ghostty, WezTerm, etc.)."""
+    return f"\033]8;;{url}\033\\{text}\033]8;;\033\\"
+
+
+# =========================================================================
+# Context pressure display (CLI user-facing warnings)
+# =========================================================================
+
+# ANSI color codes for context pressure tiers
+_CYAN = "\033[36m"
+_YELLOW = "\033[33m"
+_BOLD = "\033[1m"
+_DIM_ANSI = "\033[2m"
+
+# Bar characters
+_BAR_FILLED = "▰"
+_BAR_EMPTY = "▱"
+_BAR_WIDTH = 20
+
+
+def format_context_pressure(
+    compaction_progress: float,
+    threshold_tokens: int,
+    threshold_percent: float,
+    compression_enabled: bool = True,
+) -> str:
+    """Build a formatted context pressure line for CLI display.
+
+    The bar and percentage show progress toward the compaction threshold,
+    NOT the raw context window.  100% = compaction fires.
+
+    Args:
+        compaction_progress: How close to compaction (0.0–1.0, 1.0 = fires).
+        threshold_tokens: Compaction threshold in tokens.
+        threshold_percent: Compaction threshold as a fraction of context window.
+        compression_enabled: Whether auto-compression is active.
+    """
+    pct_int = min(int(compaction_progress * 100), 100)
+    filled = min(int(compaction_progress * _BAR_WIDTH), _BAR_WIDTH)
+    bar = _BAR_FILLED * filled + _BAR_EMPTY * (_BAR_WIDTH - filled)
+
+    threshold_k = f"{threshold_tokens // 1000}k" if threshold_tokens >= 1000 else str(threshold_tokens)
+    threshold_pct_int = int(threshold_percent * 100)
+
+    color = f"{_BOLD}{_YELLOW}"
+    icon = "⚠"
+    if compression_enabled:
+        hint = "compaction approaching"
+    else:
+        hint = "no auto-compaction"
+
+    return (
+        f"  {color}{icon} context {bar} {pct_int}% to compaction{_ANSI_RESET}"
+        f"  {_DIM_ANSI}{threshold_k} threshold ({threshold_pct_int}%) · {hint}{_ANSI_RESET}"
+    )
+
+
+def format_context_pressure_gateway(
+    compaction_progress: float,
+    threshold_percent: float,
+    compression_enabled: bool = True,
+) -> str:
+    """Build a plain-text context pressure notification for messaging platforms.
+
+    No ANSI — just Unicode and plain text suitable for Telegram/Discord/etc.
+    The percentage shows progress toward the compaction threshold.
+    """
+    pct_int = min(int(compaction_progress * 100), 100)
+    filled = min(int(compaction_progress * _BAR_WIDTH), _BAR_WIDTH)
+    bar = _BAR_FILLED * filled + _BAR_EMPTY * (_BAR_WIDTH - filled)
+
+    threshold_pct_int = int(threshold_percent * 100)
+
+    icon = "⚠️"
+    if compression_enabled:
+        hint = f"Context compaction approaching (threshold: {threshold_pct_int}% of window)."
+    else:
+        hint = "Auto-compaction is disabled — context may be truncated."
+
+    return f"{icon} Context: {bar} {pct_int}% to compaction\n{hint}"
@@ -1,829 +0,0 @@
-"""API error classification for smart failover and recovery.
-
-Provides a structured taxonomy of API errors and a priority-ordered
-classification pipeline that determines the correct recovery action
-(retry, rotate credential, fallback to another provider, compress
-context, or abort).
-
-Replaces scattered inline string-matching with a centralized classifier
-that the main retry loop in run_agent.py consults for every API failure.
-"""
-
-from __future__ import annotations
-
-import enum
-import logging
-from dataclasses import dataclass, field
-from typing import Any, Dict, Optional
-
-logger = logging.getLogger(__name__)
-
-
-# ── Error taxonomy ──────────────────────────────────────────────────────
-
-class FailoverReason(enum.Enum):
-    """Why an API call failed — determines recovery strategy."""
-
-    # Authentication / authorization
-    auth = "auth"                        # Transient auth (401/403) — refresh/rotate
-    auth_permanent = "auth_permanent"    # Auth failed after refresh — abort
-
-    # Billing / quota
-    billing = "billing"                  # 402 or confirmed credit exhaustion — rotate immediately
-    rate_limit = "rate_limit"            # 429 or quota-based throttling — backoff then rotate
-
-    # Server-side
-    overloaded = "overloaded"            # 503/529 — provider overloaded, backoff
-    server_error = "server_error"        # 500/502 — internal server error, retry
-
-    # Transport
-    timeout = "timeout"                  # Connection/read timeout — rebuild client + retry
-
-    # Context / payload
-    context_overflow = "context_overflow"  # Context too large — compress, not failover
-    payload_too_large = "payload_too_large"  # 413 — compress payload
-
-    # Model
-    model_not_found = "model_not_found"  # 404 or invalid model — fallback to different model
-
-    # Request format
-    format_error = "format_error"        # 400 bad request — abort or strip + retry
-
-    # Provider-specific
-    thinking_signature = "thinking_signature"  # Anthropic thinking block sig invalid
-    long_context_tier = "long_context_tier"    # Anthropic "extra usage" tier gate
-
-    # Catch-all
-    unknown = "unknown"                  # Unclassifiable — retry with backoff
-
-
-# ── Classification result ───────────────────────────────────────────────
-
-@dataclass
-class ClassifiedError:
-    """Structured classification of an API error with recovery hints."""
-
-    reason: FailoverReason
-    status_code: Optional[int] = None
-    provider: Optional[str] = None
-    model: Optional[str] = None
-    message: str = ""
-    error_context: Dict[str, Any] = field(default_factory=dict)
-
-    # Recovery action hints — the retry loop checks these instead of
-    # re-classifying the error itself.
-    retryable: bool = True
-    should_compress: bool = False
-    should_rotate_credential: bool = False
-    should_fallback: bool = False
-
-    @property
-    def is_auth(self) -> bool:
-        return self.reason in (FailoverReason.auth, FailoverReason.auth_permanent)
-
-
-
-# ── Provider-specific patterns ──────────────────────────────────────────
-
-# Patterns that indicate billing exhaustion (not transient rate limit)
-_BILLING_PATTERNS = [
-    "insufficient credits",
-    "insufficient_quota",
-    "credit balance",
-    "credits have been exhausted",
-    "top up your credits",
-    "payment required",
-    "billing hard limit",
-    "exceeded your current quota",
-    "account is deactivated",
-    "plan does not include",
-]
-
-# Patterns that indicate rate limiting (transient, will resolve)
-_RATE_LIMIT_PATTERNS = [
-    "rate limit",
-    "rate_limit",
-    "too many requests",
-    "throttled",
-    "requests per minute",
-    "tokens per minute",
-    "requests per day",
-    "try again in",
-    "please retry after",
-    "resource_exhausted",
-    "rate increased too quickly",  # Alibaba/DashScope throttling
-    # AWS Bedrock throttling
-    "throttlingexception",
-    "too many concurrent requests",
-    "servicequotaexceededexception",
-]
-
-# Usage-limit patterns that need disambiguation (could be billing OR rate_limit)
-_USAGE_LIMIT_PATTERNS = [
-    "usage limit",
-    "quota",
-    "limit exceeded",
-    "key limit exceeded",
-]
-
-# Patterns confirming usage limit is transient (not billing)
-_USAGE_LIMIT_TRANSIENT_SIGNALS = [
-    "try again",
-    "retry",
-    "resets at",
-    "reset in",
-    "wait",
-    "requests remaining",
-    "periodic",
-    "window",
-]
-
-# Payload-too-large patterns detected from message text (no status_code attr).
-# Proxies and some backends embed the HTTP status in the error message.
-_PAYLOAD_TOO_LARGE_PATTERNS = [
-    "request entity too large",
-    "payload too large",
-    "error code: 413",
-]
-
-# Context overflow patterns
-_CONTEXT_OVERFLOW_PATTERNS = [
-    "context length",
-    "context size",
-    "maximum context",
-    "token limit",
-    "too many tokens",
-    "reduce the length",
-    "exceeds the limit",
-    "context window",
-    "prompt is too long",
-    "prompt exceeds max length",
-    "max_tokens",
-    "maximum number of tokens",
-    # vLLM / local inference server patterns
-    "exceeds the max_model_len",
-    "max_model_len",
-    "prompt length",             # "engine prompt length X exceeds"
-    "input is too long",
-    "maximum model length",
-    # Ollama patterns
-    "context length exceeded",
-    "truncating input",
-    # llama.cpp / llama-server patterns
-    "slot context",              # "slot context: N tokens, prompt N tokens"
-    "n_ctx_slot",
-    # Chinese error messages (some providers return these)
-    "超过最大长度",
-    "上下文长度",
-    # AWS Bedrock Converse API error patterns
-    "input is too long",
-    "max input token",
-    "input token",
-    "exceeds the maximum number of input tokens",
-]
-
-# Model not found patterns
-_MODEL_NOT_FOUND_PATTERNS = [
-    "is not a valid model",
-    "invalid model",
-    "model not found",
-    "model_not_found",
-    "does not exist",
-    "no such model",
-    "unknown model",
-    "unsupported model",
-]
-
-# Auth patterns (non-status-code signals)
-_AUTH_PATTERNS = [
-    "invalid api key",
-    "invalid_api_key",
-    "authentication",
-    "unauthorized",
-    "forbidden",
-    "invalid token",
-    "token expired",
-    "token revoked",
-    "access denied",
-]
-
-# Anthropic thinking block signature patterns
-_THINKING_SIG_PATTERNS = [
-    "signature",  # Combined with "thinking" check
-]
-
-# Transport error type names
-_TRANSPORT_ERROR_TYPES = frozenset({
-    "ReadTimeout", "ConnectTimeout", "PoolTimeout",
-    "ConnectError", "RemoteProtocolError",
-    "ConnectionError", "ConnectionResetError",
-    "ConnectionAbortedError", "BrokenPipeError",
-    "TimeoutError", "ReadError",
-    "ServerDisconnectedError",
-    # OpenAI SDK errors (not subclasses of Python builtins)
-    "APIConnectionError",
-    "APITimeoutError",
-})
-
-# Server disconnect patterns (no status code, but transport-level)
-_SERVER_DISCONNECT_PATTERNS = [
-    "server disconnected",
-    "peer closed connection",
-    "connection reset by peer",
-    "connection was closed",
-    "network connection lost",
-    "unexpected eof",
-    "incomplete chunked read",
-]
-
-
-# ── Classification pipeline ─────────────────────────────────────────────
-
-def classify_api_error(
-    error: Exception,
-    *,
-    provider: str = "",
-    model: str = "",
-    approx_tokens: int = 0,
-    context_length: int = 200000,
-    num_messages: int = 0,
-) -> ClassifiedError:
-    """Classify an API error into a structured recovery recommendation.
-
-    Priority-ordered pipeline:
-      1. Special-case provider-specific patterns (thinking sigs, tier gates)
-      2. HTTP status code + message-aware refinement
-      3. Error code classification (from body)
-      4. Message pattern matching (billing vs rate_limit vs context vs auth)
-      5. Transport error heuristics
-      6. Server disconnect + large session → context overflow
-      7. Fallback: unknown (retryable with backoff)
-
-    Args:
-        error: The exception from the API call.
-        provider: Current provider name (e.g. "openrouter", "anthropic").
-        model: Current model slug.
-        approx_tokens: Approximate token count of the current context.
-        context_length: Maximum context length for the current model.
-
-    Returns:
-        ClassifiedError with reason and recovery action hints.
-    """
-    status_code = _extract_status_code(error)
-    error_type = type(error).__name__
-    body = _extract_error_body(error)
-    error_code = _extract_error_code(body)
-
-    # Build a comprehensive error message string for pattern matching.
-    # str(error) alone may not include the body message (e.g. OpenAI SDK's
-    # APIStatusError.__str__ returns the first arg, not the body).  Append
-    # the body message so patterns like "try again" in 402 disambiguation
-    # are detected even when only present in the structured body.
-    #
-    # Also extract metadata.raw — OpenRouter wraps upstream provider errors
-    # inside {"error": {"message": "Provider returned error", "metadata":
-    # {"raw": "<actual error JSON>"}}} and the real error message (e.g.
-    # "context length exceeded") is only in the inner JSON.
-    _raw_msg = str(error).lower()
-    _body_msg = ""
-    _metadata_msg = ""
-    if isinstance(body, dict):
-        _err_obj = body.get("error", {})
-        if isinstance(_err_obj, dict):
-            _body_msg = (_err_obj.get("message") or "").lower()
-            # Parse metadata.raw for wrapped provider errors
-            _metadata = _err_obj.get("metadata", {})
-            if isinstance(_metadata, dict):
-                _raw_json = _metadata.get("raw") or ""
-                if isinstance(_raw_json, str) and _raw_json.strip():
-                    try:
-                        import json
-                        _inner = json.loads(_raw_json)
-                        if isinstance(_inner, dict):
-                            _inner_err = _inner.get("error", {})
-                            if isinstance(_inner_err, dict):
-                                _metadata_msg = (_inner_err.get("message") or "").lower()
-                    except (json.JSONDecodeError, TypeError):
-                        pass
-        if not _body_msg:
-            _body_msg = (body.get("message") or "").lower()
-    # Combine all message sources for pattern matching
-    parts = [_raw_msg]
-    if _body_msg and _body_msg not in _raw_msg:
-        parts.append(_body_msg)
-    if _metadata_msg and _metadata_msg not in _raw_msg and _metadata_msg not in _body_msg:
-        parts.append(_metadata_msg)
-    error_msg = " ".join(parts)
-    provider_lower = (provider or "").strip().lower()
-    model_lower = (model or "").strip().lower()
-
-    def _result(reason: FailoverReason, **overrides) -> ClassifiedError:
-        defaults = {
-            "reason": reason,
-            "status_code": status_code,
-            "provider": provider,
-            "model": model,
-            "message": _extract_message(error, body),
-        }
-        defaults.update(overrides)
-        return ClassifiedError(**defaults)
-
-    # ── 1. Provider-specific patterns (highest priority) ────────────
-
-    # Anthropic thinking block signature invalid (400).
-    # Don't gate on provider — OpenRouter proxies Anthropic errors, so the
-    # provider may be "openrouter" even though the error is Anthropic-specific.
-    # The message pattern ("signature" + "thinking") is unique enough.
-    if (
-        status_code == 400
-        and "signature" in error_msg
-        and "thinking" in error_msg
-    ):
-        return _result(
-            FailoverReason.thinking_signature,
-            retryable=True,
-            should_compress=False,
-        )
-
-    # Anthropic long-context tier gate (429 "extra usage" + "long context")
-    if (
-        status_code == 429
-        and "extra usage" in error_msg
-        and "long context" in error_msg
-    ):
-        return _result(
-            FailoverReason.long_context_tier,
-            retryable=True,
-            should_compress=True,
-        )
-
-    # ── 2. HTTP status code classification ──────────────────────────
-
-    if status_code is not None:
-        classified = _classify_by_status(
-            status_code, error_msg, error_code, body,
-            provider=provider_lower, model=model_lower,
-            approx_tokens=approx_tokens, context_length=context_length,
-            num_messages=num_messages,
-            result_fn=_result,
-        )
-        if classified is not None:
-            return classified
-
-    # ── 3. Error code classification ────────────────────────────────
-
-    if error_code:
-        classified = _classify_by_error_code(error_code, error_msg, _result)
-        if classified is not None:
-            return classified
-
-    # ── 4. Message pattern matching (no status code) ────────────────
-
-    classified = _classify_by_message(
-        error_msg, error_type,
-        approx_tokens=approx_tokens,
-        context_length=context_length,
-        result_fn=_result,
-    )
-    if classified is not None:
-        return classified
-
-    # ── 5. Server disconnect + large session → context overflow ─────
-    # Must come BEFORE generic transport error catch — a disconnect on
-    # a large session is more likely context overflow than a transient
-    # transport hiccup.  Without this ordering, RemoteProtocolError
-    # always maps to timeout regardless of session size.
-
-    is_disconnect = any(p in error_msg for p in _SERVER_DISCONNECT_PATTERNS)
-    if is_disconnect and not status_code:
-        is_large = approx_tokens > context_length * 0.6 or approx_tokens > 120000 or num_messages > 200
-        if is_large:
-            return _result(
-                FailoverReason.context_overflow,
-                retryable=True,
-                should_compress=True,
-            )
-        return _result(FailoverReason.timeout, retryable=True)
-
-    # ── 6. Transport / timeout heuristics ───────────────────────────
-
-    if error_type in _TRANSPORT_ERROR_TYPES or isinstance(error, (TimeoutError, ConnectionError, OSError)):
-        return _result(FailoverReason.timeout, retryable=True)
-
-    # ── 7. Fallback: unknown ────────────────────────────────────────
-
-    return _result(FailoverReason.unknown, retryable=True)
-
-
-# ── Status code classification ──────────────────────────────────────────
-
-def _classify_by_status(
-    status_code: int,
-    error_msg: str,
-    error_code: str,
-    body: dict,
-    *,
-    provider: str,
-    model: str,
-    approx_tokens: int,
-    context_length: int,
-    num_messages: int = 0,
-    result_fn,
-) -> Optional[ClassifiedError]:
-    """Classify based on HTTP status code with message-aware refinement."""
-
-    if status_code == 401:
-        # Not retryable on its own — credential pool rotation and
-        # provider-specific refresh (Codex, Anthropic, Nous) run before
-        # the retryability check in run_agent.py.  If those succeed, the
-        # loop `continue`s.  If they fail, retryable=False ensures we
-        # hit the client-error abort path (which tries fallback first).
-        return result_fn(
-            FailoverReason.auth,
-            retryable=False,
-            should_rotate_credential=True,
-            should_fallback=True,
-        )
-
-    if status_code == 403:
-        # OpenRouter 403 "key limit exceeded" is actually billing
-        if "key limit exceeded" in error_msg or "spending limit" in error_msg:
-            return result_fn(
-                FailoverReason.billing,
-                retryable=False,
-                should_rotate_credential=True,
-                should_fallback=True,
-            )
-        return result_fn(
-            FailoverReason.auth,
-            retryable=False,
-            should_fallback=True,
-        )
-
-    if status_code == 402:
-        return _classify_402(error_msg, result_fn)
-
-    if status_code == 404:
-        if any(p in error_msg for p in _MODEL_NOT_FOUND_PATTERNS):
-            return result_fn(
-                FailoverReason.model_not_found,
-                retryable=False,
-                should_fallback=True,
-            )
-        # Generic 404 — could be model or endpoint
-        return result_fn(
-            FailoverReason.model_not_found,
-            retryable=False,
-            should_fallback=True,
-        )
-
-    if status_code == 413:
-        return result_fn(
-            FailoverReason.payload_too_large,
-            retryable=True,
-            should_compress=True,
-        )
-
-    if status_code == 429:
-        # Already checked long_context_tier above; this is a normal rate limit
-        return result_fn(
-            FailoverReason.rate_limit,
-            retryable=True,
-            should_rotate_credential=True,
-            should_fallback=True,
-        )
-
-    if status_code == 400:
-        return _classify_400(
-            error_msg, error_code, body,
-            provider=provider, model=model,
-            approx_tokens=approx_tokens,
-            context_length=context_length,
-            num_messages=num_messages,
-            result_fn=result_fn,
-        )
-
-    if status_code in (500, 502):
-        return result_fn(FailoverReason.server_error, retryable=True)
-
-    if status_code in (503, 529):
-        return result_fn(FailoverReason.overloaded, retryable=True)
-
-    # Other 4xx — non-retryable
-    if 400 <= status_code < 500:
-        return result_fn(
-            FailoverReason.format_error,
-            retryable=False,
-            should_fallback=True,
-        )
-
-    # Other 5xx — retryable
-    if 500 <= status_code < 600:
-        return result_fn(FailoverReason.server_error, retryable=True)
-
-    return None
-
-
-def _classify_402(error_msg: str, result_fn) -> ClassifiedError:
-    """Disambiguate 402: billing exhaustion vs transient usage limit.
-
-    The key insight from OpenClaw: some 402s are transient rate limits
-    disguised as payment errors.  "Usage limit, try again in 5 minutes"
-    is NOT a billing problem — it's a periodic quota that resets.
-    """
-    # Check for transient usage-limit signals first
-    has_usage_limit = any(p in error_msg for p in _USAGE_LIMIT_PATTERNS)
-    has_transient_signal = any(p in error_msg for p in _USAGE_LIMIT_TRANSIENT_SIGNALS)
-
-    if has_usage_limit and has_transient_signal:
-        # Transient quota — treat as rate limit, not billing
-        return result_fn(
-            FailoverReason.rate_limit,
-            retryable=True,
-            should_rotate_credential=True,
-            should_fallback=True,
-        )
-
-    # Confirmed billing exhaustion
-    return result_fn(
-        FailoverReason.billing,
-        retryable=False,
-        should_rotate_credential=True,
-        should_fallback=True,
-    )
-
-
-def _classify_400(
-    error_msg: str,
-    error_code: str,
-    body: dict,
-    *,
-    provider: str,
-    model: str,
-    approx_tokens: int,
-    context_length: int,
-    num_messages: int = 0,
-    result_fn,
-) -> ClassifiedError:
-    """Classify 400 Bad Request — context overflow, format error, or generic."""
-
-    # Context overflow from 400
-    if any(p in error_msg for p in _CONTEXT_OVERFLOW_PATTERNS):
-        return result_fn(
-            FailoverReason.context_overflow,
-            retryable=True,
-            should_compress=True,
-        )
-
-    # Some providers return model-not-found as 400 instead of 404 (e.g. OpenRouter).
-    if any(p in error_msg for p in _MODEL_NOT_FOUND_PATTERNS):
-        return result_fn(
-            FailoverReason.model_not_found,
-            retryable=False,
-            should_fallback=True,
-        )
-
-    # Some providers return rate limit / billing errors as 400 instead of 429/402.
-    # Check these patterns before falling through to format_error.
-    if any(p in error_msg for p in _RATE_LIMIT_PATTERNS):
-        return result_fn(
-            FailoverReason.rate_limit,
-            retryable=True,
-            should_rotate_credential=True,
-            should_fallback=True,
-        )
-    if any(p in error_msg for p in _BILLING_PATTERNS):
-        return result_fn(
-            FailoverReason.billing,
-            retryable=False,
-            should_rotate_credential=True,
-            should_fallback=True,
-        )
-
-    # Generic 400 + large session → probable context overflow
-    # Anthropic sometimes returns a bare "Error" message when context is too large
-    err_body_msg = ""
-    if isinstance(body, dict):
-        err_obj = body.get("error", {})
-        if isinstance(err_obj, dict):
-            err_body_msg = (err_obj.get("message") or "").strip().lower()
-        # Responses API (and some providers) use flat body: {"message": "..."}
-        if not err_body_msg:
-            err_body_msg = (body.get("message") or "").strip().lower()
-    is_generic = len(err_body_msg) < 30 or err_body_msg in ("error", "")
-    is_large = approx_tokens > context_length * 0.4 or approx_tokens > 80000 or num_messages > 80
-
-    if is_generic and is_large:
-        return result_fn(
-            FailoverReason.context_overflow,
-            retryable=True,
-            should_compress=True,
-        )
-
-    # Non-retryable format error
-    return result_fn(
-        FailoverReason.format_error,
-        retryable=False,
-        should_fallback=True,
-    )
-
-
-# ── Error code classification ───────────────────────────────────────────
-
-def _classify_by_error_code(
-    error_code: str, error_msg: str, result_fn,
-) -> Optional[ClassifiedError]:
-    """Classify by structured error codes from the response body."""
-    code_lower = error_code.lower()
-
-    if code_lower in ("resource_exhausted", "throttled", "rate_limit_exceeded"):
-        return result_fn(
-            FailoverReason.rate_limit,
-            retryable=True,
-            should_rotate_credential=True,
-        )
-
-    if code_lower in ("insufficient_quota", "billing_not_active", "payment_required"):
-        return result_fn(
-            FailoverReason.billing,
-            retryable=False,
-            should_rotate_credential=True,
-            should_fallback=True,
-        )
-
-    if code_lower in ("model_not_found", "model_not_available", "invalid_model"):
-        return result_fn(
-            FailoverReason.model_not_found,
-            retryable=False,
-            should_fallback=True,
-        )
-
-    if code_lower in ("context_length_exceeded", "max_tokens_exceeded"):
-        return result_fn(
-            FailoverReason.context_overflow,
-            retryable=True,
-            should_compress=True,
-        )
-
-    return None
-
-
-# ── Message pattern classification ──────────────────────────────────────
-
-def _classify_by_message(
-    error_msg: str,
-    error_type: str,
-    *,
-    approx_tokens: int,
-    context_length: int,
-    result_fn,
-) -> Optional[ClassifiedError]:
-    """Classify based on error message patterns when no status code is available."""
-
-    # Payload-too-large patterns (from message text when no status_code)
-    if any(p in error_msg for p in _PAYLOAD_TOO_LARGE_PATTERNS):
-        return result_fn(
-            FailoverReason.payload_too_large,
-            retryable=True,
-            should_compress=True,
-        )
-
-    # Usage-limit patterns need the same disambiguation as 402: some providers
-    # surface "usage limit" errors without an HTTP status code.  A transient
-    # signal ("try again", "resets at", …) means it's a periodic quota, not
-    # billing exhaustion.
-    has_usage_limit = any(p in error_msg for p in _USAGE_LIMIT_PATTERNS)
-    if has_usage_limit:
-        has_transient_signal = any(p in error_msg for p in _USAGE_LIMIT_TRANSIENT_SIGNALS)
-        if has_transient_signal:
-            return result_fn(
-                FailoverReason.rate_limit,
-                retryable=True,
-                should_rotate_credential=True,
-                should_fallback=True,
-            )
-        return result_fn(
-            FailoverReason.billing,
-            retryable=False,
-            should_rotate_credential=True,
-            should_fallback=True,
-        )
-
-    # Billing patterns
-    if any(p in error_msg for p in _BILLING_PATTERNS):
-        return result_fn(
-            FailoverReason.billing,
-            retryable=False,
-            should_rotate_credential=True,
-            should_fallback=True,
-        )
-
-    # Rate limit patterns
-    if any(p in error_msg for p in _RATE_LIMIT_PATTERNS):
-        return result_fn(
-            FailoverReason.rate_limit,
-            retryable=True,
-            should_rotate_credential=True,
-            should_fallback=True,
-        )
-
-    # Context overflow patterns
-    if any(p in error_msg for p in _CONTEXT_OVERFLOW_PATTERNS):
-        return result_fn(
-            FailoverReason.context_overflow,
-            retryable=True,
-            should_compress=True,
-        )
-
-    # Auth patterns
-    # Auth errors should NOT be retried directly — the credential is invalid and
-    # retrying with the same key will always fail.  Set retryable=False so the
-    # caller triggers credential rotation (should_rotate_credential=True) or
-    # provider fallback rather than an immediate retry loop.
-    if any(p in error_msg for p in _AUTH_PATTERNS):
-        return result_fn(
-            FailoverReason.auth,
-            retryable=False,
-            should_rotate_credential=True,
-            should_fallback=True,
-        )
-
-    # Model not found patterns
-    if any(p in error_msg for p in _MODEL_NOT_FOUND_PATTERNS):
-        return result_fn(
-            FailoverReason.model_not_found,
-            retryable=False,
-            should_fallback=True,
-        )
-
-    return None
-
-
-# ── Helpers ─────────────────────────────────────────────────────────────
-
-def _extract_status_code(error: Exception) -> Optional[int]:
-    """Walk the error and its cause chain to find an HTTP status code."""
-    current = error
-    for _ in range(5):  # Max depth to prevent infinite loops
-        code = getattr(current, "status_code", None)
-        if isinstance(code, int):
-            return code
-        # Some SDKs use .status instead of .status_code
-        code = getattr(current, "status", None)
-        if isinstance(code, int) and 100 <= code < 600:
-            return code
-        # Walk cause chain
-        cause = getattr(current, "__cause__", None) or getattr(current, "__context__", None)
-        if cause is None or cause is current:
-            break
-        current = cause
-    return None
-
-
-def _extract_error_body(error: Exception) -> dict:
-    """Extract the structured error body from an SDK exception."""
-    body = getattr(error, "body", None)
-    if isinstance(body, dict):
-        return body
-    # Some errors have .response.json()
-    response = getattr(error, "response", None)
-    if response is not None:
-        try:
-            json_body = response.json()
-            if isinstance(json_body, dict):
-                return json_body
-        except Exception:
-            pass
-    return {}
-
-
-def _extract_error_code(body: dict) -> str:
-    """Extract an error code string from the response body."""
-    if not body:
-        return ""
-    error_obj = body.get("error", {})
-    if isinstance(error_obj, dict):
-        code = error_obj.get("code") or error_obj.get("type") or ""
-        if isinstance(code, str) and code.strip():
-            return code.strip()
-    # Top-level code
-    code = body.get("code") or body.get("error_code") or ""
-    if isinstance(code, (str, int)):
-        return str(code).strip()
-    return ""
-
-
-def _extract_message(error: Exception, body: dict) -> str:
-    """Extract the most informative error message."""
-    # Try structured body first
-    if body:
-        error_obj = body.get("error", {})
-        if isinstance(error_obj, dict):
-            msg = error_obj.get("message", "")
-            if isinstance(msg, str) and msg.strip():
-                return msg.strip()[:500]
-        msg = body.get("message", "")
-        if isinstance(msg, str) and msg.strip():
-            return msg.strip()[:500]
-    # Fallback to str(error)
-    return str(error)[:500]
@@ -1,895 +0,0 @@
-"""OpenAI-compatible facade that talks to Google's Cloud Code Assist backend.
-
-This adapter lets Hermes use the ``google-gemini-cli`` provider as if it were
-a standard OpenAI-shaped chat completion endpoint, while the underlying HTTP
-traffic goes to ``cloudcode-pa.googleapis.com/v1internal:{generateContent,
-streamGenerateContent}`` with a Bearer access token obtained via OAuth PKCE.
-
-Architecture
------------
- ``GeminiCloudCodeClient`` exposes ``.chat.completions.create(**kwargs)``
-  mirroring the subset of the OpenAI SDK that ``run_agent.py`` uses.
- Incoming OpenAI ``messages[]`` / ``tools[]`` / ``tool_choice`` are translated
-  to Gemini's native ``contents[]`` / ``tools[].functionDeclarations`` /
-  ``toolConfig`` / ``systemInstruction`` shape.
- The request body is wrapped ``{project, model, user_prompt_id, request}``
-  per Code Assist API expectations.
- Responses (``candidates[].content.parts[]``) are converted back to
-  OpenAI ``choices[0].message`` shape with ``content`` + ``tool_calls``.
- Streaming uses SSE (``?alt=sse``) and yields OpenAI-shaped delta chunks.
-
-Attribution
-----------
-Translation semantics follow jenslys/opencode-gemini-auth (MIT) and the public
-Gemini API docs. Request envelope shape
-(``{project, model, user_prompt_id, request}``) is documented nowhere; it is
-reverse-engineered from the opencode-gemini-auth and clawdbot implementations.
-"""
-
-from __future__ import annotations
-
-import json
-import logging
-import os
-import time
-import uuid
-from types import SimpleNamespace
-from typing import Any, Dict, Iterator, List, Optional
-
-import httpx
-
-from agent import google_oauth
-from agent.google_code_assist import (
-    CODE_ASSIST_ENDPOINT,
-    FREE_TIER_ID,
-    CodeAssistError,
-    ProjectContext,
-    resolve_project_context,
-)
-
-logger = logging.getLogger(__name__)
-
-
-# =============================================================================
-# Request translation: OpenAI → Gemini
-# =============================================================================
-
-_ROLE_MAP_OPENAI_TO_GEMINI = {
-    "user": "user",
-    "assistant": "model",
-    "system": "user",   # handled separately via systemInstruction
-    "tool": "user",     # functionResponse is wrapped in a user-role turn
-    "function": "user",
-}
-
-
-def _coerce_content_to_text(content: Any) -> str:
-    """OpenAI content may be str or a list of parts; reduce to plain text."""
-    if content is None:
-        return ""
-    if isinstance(content, str):
-        return content
-    if isinstance(content, list):
-        pieces: List[str] = []
-        for p in content:
-            if isinstance(p, str):
-                pieces.append(p)
-            elif isinstance(p, dict):
-                if p.get("type") == "text" and isinstance(p.get("text"), str):
-                    pieces.append(p["text"])
-                # Multimodal (image_url, etc.) — stub for now; log and skip
-                elif p.get("type") in ("image_url", "input_audio"):
-                    logger.debug("Dropping multimodal part (not yet supported): %s", p.get("type"))
-        return "\n".join(pieces)
-    return str(content)
-
-
-def _translate_tool_call_to_gemini(tool_call: Dict[str, Any]) -> Dict[str, Any]:
-    """OpenAI tool_call -> Gemini functionCall part."""
-    fn = tool_call.get("function") or {}
-    args_raw = fn.get("arguments", "")
-    try:
-        args = json.loads(args_raw) if isinstance(args_raw, str) and args_raw else {}
-    except json.JSONDecodeError:
-        args = {"_raw": args_raw}
-    if not isinstance(args, dict):
-        args = {"_value": args}
-    return {
-        "functionCall": {
-            "name": fn.get("name") or "",
-            "args": args,
-        },
-        # Sentinel signature — matches opencode-gemini-auth's approach.
-        # Without this, Code Assist rejects function calls that originated
-        # outside its own chain.
-        "thoughtSignature": "skip_thought_signature_validator",
-    }
-
-
-def _translate_tool_result_to_gemini(message: Dict[str, Any]) -> Dict[str, Any]:
-    """OpenAI tool-role message -> Gemini functionResponse part.
-
-    The function name isn't in the OpenAI tool message directly; it must be
-    passed via the assistant message that issued the call. For simplicity we
-    look up ``name`` on the message (OpenAI SDK copies it there) or on the
-    ``tool_call_id`` cross-reference.
-    """
-    name = str(message.get("name") or message.get("tool_call_id") or "tool")
-    content = _coerce_content_to_text(message.get("content"))
-    # Gemini expects the response as a dict under `response`. We wrap plain
-    # text in {"output": "..."}.
-    try:
-        parsed = json.loads(content) if content.strip().startswith(("{", "[")) else None
-    except json.JSONDecodeError:
-        parsed = None
-    response = parsed if isinstance(parsed, dict) else {"output": content}
-    return {
-        "functionResponse": {
-            "name": name,
-            "response": response,
-        },
-    }
-
-
-def _build_gemini_contents(
-    messages: List[Dict[str, Any]],
-) -> tuple[List[Dict[str, Any]], Optional[Dict[str, Any]]]:
-    """Convert OpenAI messages[] to Gemini contents[] + systemInstruction."""
-    system_text_parts: List[str] = []
-    contents: List[Dict[str, Any]] = []
-
-    for msg in messages:
-        if not isinstance(msg, dict):
-            continue
-        role = str(msg.get("role") or "user")
-
-        if role == "system":
-            system_text_parts.append(_coerce_content_to_text(msg.get("content")))
-            continue
-
-        # Tool result message — emit a user-role turn with functionResponse
-        if role == "tool" or role == "function":
-            contents.append({
-                "role": "user",
-                "parts": [_translate_tool_result_to_gemini(msg)],
-            })
-            continue
-
-        gemini_role = _ROLE_MAP_OPENAI_TO_GEMINI.get(role, "user")
-        parts: List[Dict[str, Any]] = []
-
-        text = _coerce_content_to_text(msg.get("content"))
-        if text:
-            parts.append({"text": text})
-
-        # Assistant messages can carry tool_calls
-        tool_calls = msg.get("tool_calls") or []
-        if isinstance(tool_calls, list):
-            for tc in tool_calls:
-                if isinstance(tc, dict):
-                    parts.append(_translate_tool_call_to_gemini(tc))
-
-        if not parts:
-            # Gemini rejects empty parts; skip the turn entirely
-            continue
-
-        contents.append({"role": gemini_role, "parts": parts})
-
-    system_instruction: Optional[Dict[str, Any]] = None
-    joined_system = "\n".join(p for p in system_text_parts if p).strip()
-    if joined_system:
-        system_instruction = {
-            "role": "system",
-            "parts": [{"text": joined_system}],
-        }
-
-    return contents, system_instruction
-
-
-def _translate_tools_to_gemini(tools: Any) -> List[Dict[str, Any]]:
-    """OpenAI tools[] -> Gemini tools[].functionDeclarations[]."""
-    if not isinstance(tools, list) or not tools:
-        return []
-    declarations: List[Dict[str, Any]] = []
-    for t in tools:
-        if not isinstance(t, dict):
-            continue
-        fn = t.get("function") or {}
-        if not isinstance(fn, dict):
-            continue
-        name = fn.get("name")
-        if not name:
-            continue
-        decl = {"name": str(name)}
-        if fn.get("description"):
-            decl["description"] = str(fn["description"])
-        params = fn.get("parameters")
-        if isinstance(params, dict):
-            decl["parameters"] = params
-        declarations.append(decl)
-    if not declarations:
-        return []
-    return [{"functionDeclarations": declarations}]
-
-
-def _translate_tool_choice_to_gemini(tool_choice: Any) -> Optional[Dict[str, Any]]:
-    """OpenAI tool_choice -> Gemini toolConfig.functionCallingConfig."""
-    if tool_choice is None:
-        return None
-    if isinstance(tool_choice, str):
-        if tool_choice == "auto":
-            return {"functionCallingConfig": {"mode": "AUTO"}}
-        if tool_choice == "required":
-            return {"functionCallingConfig": {"mode": "ANY"}}
-        if tool_choice == "none":
-            return {"functionCallingConfig": {"mode": "NONE"}}
-    if isinstance(tool_choice, dict):
-        fn = tool_choice.get("function") or {}
-        name = fn.get("name")
-        if name:
-            return {
-                "functionCallingConfig": {
-                    "mode": "ANY",
-                    "allowedFunctionNames": [str(name)],
-                },
-            }
-    return None
-
-
-def _normalize_thinking_config(config: Any) -> Optional[Dict[str, Any]]:
-    """Accept thinkingBudget / thinkingLevel / includeThoughts (+ snake_case)."""
-    if not isinstance(config, dict) or not config:
-        return None
-    budget = config.get("thinkingBudget", config.get("thinking_budget"))
-    level = config.get("thinkingLevel", config.get("thinking_level"))
-    include = config.get("includeThoughts", config.get("include_thoughts"))
-    normalized: Dict[str, Any] = {}
-    if isinstance(budget, (int, float)):
-        normalized["thinkingBudget"] = int(budget)
-    if isinstance(level, str) and level.strip():
-        normalized["thinkingLevel"] = level.strip().lower()
-    if isinstance(include, bool):
-        normalized["includeThoughts"] = include
-    return normalized or None
-
-
-def build_gemini_request(
-    *,
-    messages: List[Dict[str, Any]],
-    tools: Any = None,
-    tool_choice: Any = None,
-    temperature: Optional[float] = None,
-    max_tokens: Optional[int] = None,
-    top_p: Optional[float] = None,
-    stop: Any = None,
-    thinking_config: Any = None,
-) -> Dict[str, Any]:
-    """Build the inner Gemini request body (goes inside ``request`` wrapper)."""
-    contents, system_instruction = _build_gemini_contents(messages)
-
-    body: Dict[str, Any] = {"contents": contents}
-    if system_instruction is not None:
-        body["systemInstruction"] = system_instruction
-
-    gemini_tools = _translate_tools_to_gemini(tools)
-    if gemini_tools:
-        body["tools"] = gemini_tools
-    tool_cfg = _translate_tool_choice_to_gemini(tool_choice)
-    if tool_cfg is not None:
-        body["toolConfig"] = tool_cfg
-
-    generation_config: Dict[str, Any] = {}
-    if isinstance(temperature, (int, float)):
-        generation_config["temperature"] = float(temperature)
-    if isinstance(max_tokens, int) and max_tokens > 0:
-        generation_config["maxOutputTokens"] = max_tokens
-    if isinstance(top_p, (int, float)):
-        generation_config["topP"] = float(top_p)
-    if isinstance(stop, str) and stop:
-        generation_config["stopSequences"] = [stop]
-    elif isinstance(stop, list) and stop:
-        generation_config["stopSequences"] = [str(s) for s in stop if s]
-    normalized_thinking = _normalize_thinking_config(thinking_config)
-    if normalized_thinking:
-        generation_config["thinkingConfig"] = normalized_thinking
-    if generation_config:
-        body["generationConfig"] = generation_config
-
-    return body
-
-
-def wrap_code_assist_request(
-    *,
-    project_id: str,
-    model: str,
-    inner_request: Dict[str, Any],
-    user_prompt_id: Optional[str] = None,
-) -> Dict[str, Any]:
-    """Wrap the inner Gemini request in the Code Assist envelope."""
-    return {
-        "project": project_id,
-        "model": model,
-        "user_prompt_id": user_prompt_id or str(uuid.uuid4()),
-        "request": inner_request,
-    }
-
-
-# =============================================================================
-# Response translation: Gemini → OpenAI
-# =============================================================================
-
-def _translate_gemini_response(
-    resp: Dict[str, Any],
-    model: str,
-) -> SimpleNamespace:
-    """Non-streaming Gemini response -> OpenAI-shaped SimpleNamespace.
-
-    Code Assist wraps the actual Gemini response inside ``response``, so we
-    unwrap it first if present.
-    """
-    inner = resp.get("response") if isinstance(resp.get("response"), dict) else resp
-
-    candidates = inner.get("candidates") or []
-    if not isinstance(candidates, list) or not candidates:
-        return _empty_response(model)
-
-    cand = candidates[0]
-    content_obj = cand.get("content") if isinstance(cand, dict) else {}
-    parts = content_obj.get("parts") if isinstance(content_obj, dict) else []
-
-    text_pieces: List[str] = []
-    reasoning_pieces: List[str] = []
-    tool_calls: List[SimpleNamespace] = []
-
-    for i, part in enumerate(parts or []):
-        if not isinstance(part, dict):
-            continue
-        # Thought parts are model's internal reasoning — surface as reasoning,
-        # don't mix into content.
-        if part.get("thought") is True:
-            if isinstance(part.get("text"), str):
-                reasoning_pieces.append(part["text"])
-            continue
-        if isinstance(part.get("text"), str):
-            text_pieces.append(part["text"])
-            continue
-        fc = part.get("functionCall")
-        if isinstance(fc, dict) and fc.get("name"):
-            try:
-                args_str = json.dumps(fc.get("args") or {}, ensure_ascii=False)
-            except (TypeError, ValueError):
-                args_str = "{}"
-            tool_calls.append(SimpleNamespace(
-                id=f"call_{uuid.uuid4().hex[:12]}",
-                type="function",
-                index=i,
-                function=SimpleNamespace(name=str(fc["name"]), arguments=args_str),
-            ))
-
-    finish_reason = "tool_calls" if tool_calls else _map_gemini_finish_reason(
-        str(cand.get("finishReason") or "")
-    )
-
-    usage_meta = inner.get("usageMetadata") or {}
-    usage = SimpleNamespace(
-        prompt_tokens=int(usage_meta.get("promptTokenCount") or 0),
-        completion_tokens=int(usage_meta.get("candidatesTokenCount") or 0),
-        total_tokens=int(usage_meta.get("totalTokenCount") or 0),
-        prompt_tokens_details=SimpleNamespace(
-            cached_tokens=int(usage_meta.get("cachedContentTokenCount") or 0),
-        ),
-    )
-
-    message = SimpleNamespace(
-        role="assistant",
-        content="".join(text_pieces) if text_pieces else None,
-        tool_calls=tool_calls or None,
-        reasoning="".join(reasoning_pieces) or None,
-        reasoning_content="".join(reasoning_pieces) or None,
-        reasoning_details=None,
-    )
-    choice = SimpleNamespace(
-        index=0,
-        message=message,
-        finish_reason=finish_reason,
-    )
-    return SimpleNamespace(
-        id=f"chatcmpl-{uuid.uuid4().hex[:12]}",
-        object="chat.completion",
-        created=int(time.time()),
-        model=model,
-        choices=[choice],
-        usage=usage,
-    )
-
-
-def _empty_response(model: str) -> SimpleNamespace:
-    message = SimpleNamespace(
-        role="assistant", content="", tool_calls=None,
-        reasoning=None, reasoning_content=None, reasoning_details=None,
-    )
-    choice = SimpleNamespace(index=0, message=message, finish_reason="stop")
-    usage = SimpleNamespace(
-        prompt_tokens=0, completion_tokens=0, total_tokens=0,
-        prompt_tokens_details=SimpleNamespace(cached_tokens=0),
-    )
-    return SimpleNamespace(
-        id=f"chatcmpl-{uuid.uuid4().hex[:12]}",
-        object="chat.completion",
-        created=int(time.time()),
-        model=model,
-        choices=[choice],
-        usage=usage,
-    )
-
-
-def _map_gemini_finish_reason(reason: str) -> str:
-    mapping = {
-        "STOP": "stop",
-        "MAX_TOKENS": "length",
-        "SAFETY": "content_filter",
-        "RECITATION": "content_filter",
-        "OTHER": "stop",
-    }
-    return mapping.get(reason.upper(), "stop")
-
-
-# =============================================================================
-# Streaming SSE iterator
-# =============================================================================
-
-class _GeminiStreamChunk(SimpleNamespace):
-    """Mimics an OpenAI ChatCompletionChunk with .choices[0].delta."""
-    pass
-
-
-def _make_stream_chunk(
-    *,
-    model: str,
-    content: str = "",
-    tool_call_delta: Optional[Dict[str, Any]] = None,
-    finish_reason: Optional[str] = None,
-    reasoning: str = "",
-) -> _GeminiStreamChunk:
-    delta_kwargs: Dict[str, Any] = {"role": "assistant"}
-    if content:
-        delta_kwargs["content"] = content
-    if tool_call_delta is not None:
-        delta_kwargs["tool_calls"] = [SimpleNamespace(
-            index=tool_call_delta.get("index", 0),
-            id=tool_call_delta.get("id") or f"call_{uuid.uuid4().hex[:12]}",
-            type="function",
-            function=SimpleNamespace(
-                name=tool_call_delta.get("name") or "",
-                arguments=tool_call_delta.get("arguments") or "",
-            ),
-        )]
-    if reasoning:
-        delta_kwargs["reasoning"] = reasoning
-        delta_kwargs["reasoning_content"] = reasoning
-    delta = SimpleNamespace(**delta_kwargs)
-    choice = SimpleNamespace(index=0, delta=delta, finish_reason=finish_reason)
-    return _GeminiStreamChunk(
-        id=f"chatcmpl-{uuid.uuid4().hex[:12]}",
-        object="chat.completion.chunk",
-        created=int(time.time()),
-        model=model,
-        choices=[choice],
-        usage=None,
-    )
-
-
-def _iter_sse_events(response: httpx.Response) -> Iterator[Dict[str, Any]]:
-    """Parse Server-Sent Events from an httpx streaming response."""
-    buffer = ""
-    for chunk in response.iter_text():
-        if not chunk:
-            continue
-        buffer += chunk
-        while "\n" in buffer:
-            line, buffer = buffer.split("\n", 1)
-            line = line.rstrip("\r")
-            if not line:
-                continue
-            if line.startswith("data: "):
-                data = line[6:]
-                if data == "[DONE]":
-                    return
-                try:
-                    yield json.loads(data)
-                except json.JSONDecodeError:
-                    logger.debug("Non-JSON SSE line: %s", data[:200])
-
-
-def _translate_stream_event(
-    event: Dict[str, Any],
-    model: str,
-    tool_call_indices: Dict[str, int],
-) -> List[_GeminiStreamChunk]:
-    """Unwrap Code Assist envelope and emit OpenAI-shaped chunk(s)."""
-    inner = event.get("response") if isinstance(event.get("response"), dict) else event
-    candidates = inner.get("candidates") or []
-    if not candidates:
-        return []
-    cand = candidates[0]
-    if not isinstance(cand, dict):
-        return []
-
-    chunks: List[_GeminiStreamChunk] = []
-
-    content = cand.get("content") or {}
-    parts = content.get("parts") if isinstance(content, dict) else []
-    for part in parts or []:
-        if not isinstance(part, dict):
-            continue
-        if part.get("thought") is True and isinstance(part.get("text"), str):
-            chunks.append(_make_stream_chunk(
-                model=model, reasoning=part["text"],
-            ))
-            continue
-        if isinstance(part.get("text"), str) and part["text"]:
-            chunks.append(_make_stream_chunk(model=model, content=part["text"]))
-        fc = part.get("functionCall")
-        if isinstance(fc, dict) and fc.get("name"):
-            name = str(fc["name"])
-            idx = tool_call_indices.setdefault(name, len(tool_call_indices))
-            try:
-                args_str = json.dumps(fc.get("args") or {}, ensure_ascii=False)
-            except (TypeError, ValueError):
-                args_str = "{}"
-            chunks.append(_make_stream_chunk(
-                model=model,
-                tool_call_delta={
-                    "index": idx,
-                    "name": name,
-                    "arguments": args_str,
-                },
-            ))
-
-    finish_reason_raw = str(cand.get("finishReason") or "")
-    if finish_reason_raw:
-        mapped = _map_gemini_finish_reason(finish_reason_raw)
-        if tool_call_indices:
-            mapped = "tool_calls"
-        chunks.append(_make_stream_chunk(model=model, finish_reason=mapped))
-    return chunks
-
-
-# =============================================================================
-# GeminiCloudCodeClient — OpenAI-compatible facade
-# =============================================================================
-
-MARKER_BASE_URL = "cloudcode-pa://google"
-
-
-class _GeminiChatCompletions:
-    def __init__(self, client: "GeminiCloudCodeClient"):
-        self._client = client
-
-    def create(self, **kwargs: Any) -> Any:
-        return self._client._create_chat_completion(**kwargs)
-
-
-class _GeminiChatNamespace:
-    def __init__(self, client: "GeminiCloudCodeClient"):
-        self.completions = _GeminiChatCompletions(client)
-
-
-class GeminiCloudCodeClient:
-    """Minimal OpenAI-SDK-compatible facade over Code Assist v1internal."""
-
-    def __init__(
-        self,
-        *,
-        api_key: Optional[str] = None,
-        base_url: Optional[str] = None,
-        default_headers: Optional[Dict[str, str]] = None,
-        project_id: str = "",
-        **_: Any,
-    ):
-        # `api_key` here is a dummy — real auth is the OAuth access token
-        # fetched on every call via agent.google_oauth.get_valid_access_token().
-        # We accept the kwarg for openai.OpenAI interface parity.
-        self.api_key = api_key or "google-oauth"
-        self.base_url = base_url or MARKER_BASE_URL
-        self._default_headers = dict(default_headers or {})
-        self._configured_project_id = project_id
-        self._project_context: Optional[ProjectContext] = None
-        self._project_context_lock = False  # simple single-thread guard
-        self.chat = _GeminiChatNamespace(self)
-        self.is_closed = False
-        self._http = httpx.Client(timeout=httpx.Timeout(connect=15.0, read=600.0, write=30.0, pool=30.0))
-
-    def close(self) -> None:
-        self.is_closed = True
-        try:
-            self._http.close()
-        except Exception:
-            pass
-
-    # Implement the OpenAI SDK's context-manager-ish closure check
-    def __enter__(self):
-        return self
-
-    def __exit__(self, exc_type, exc_val, exc_tb):
-        self.close()
-
-    def _ensure_project_context(self, access_token: str, model: str) -> ProjectContext:
-        """Lazily resolve and cache the project context for this client."""
-        if self._project_context is not None:
-            return self._project_context
-
-        env_project = google_oauth.resolve_project_id_from_env()
-        creds = google_oauth.load_credentials()
-        stored_project = creds.project_id if creds else ""
-
-        # Prefer what's already baked into the creds
-        if stored_project:
-            self._project_context = ProjectContext(
-                project_id=stored_project,
-                managed_project_id=creds.managed_project_id if creds else "",
-                tier_id="",
-                source="stored",
-            )
-            return self._project_context
-
-        ctx = resolve_project_context(
-            access_token,
-            configured_project_id=self._configured_project_id,
-            env_project_id=env_project,
-            user_agent_model=model,
-        )
-        # Persist discovered project back to the creds file so the next
-        # session doesn't re-run the discovery.
-        if ctx.project_id or ctx.managed_project_id:
-            google_oauth.update_project_ids(
-                project_id=ctx.project_id,
-                managed_project_id=ctx.managed_project_id,
-            )
-        self._project_context = ctx
-        return ctx
-
-    def _create_chat_completion(
-        self,
-        *,
-        model: str = "gemini-2.5-flash",
-        messages: Optional[List[Dict[str, Any]]] = None,
-        stream: bool = False,
-        tools: Any = None,
-        tool_choice: Any = None,
-        temperature: Optional[float] = None,
-        max_tokens: Optional[int] = None,
-        top_p: Optional[float] = None,
-        stop: Any = None,
-        extra_body: Optional[Dict[str, Any]] = None,
-        timeout: Any = None,
-        **_: Any,
-    ) -> Any:
-        access_token = google_oauth.get_valid_access_token()
-        ctx = self._ensure_project_context(access_token, model)
-
-        thinking_config = None
-        if isinstance(extra_body, dict):
-            thinking_config = extra_body.get("thinking_config") or extra_body.get("thinkingConfig")
-
-        inner = build_gemini_request(
-            messages=messages or [],
-            tools=tools,
-            tool_choice=tool_choice,
-            temperature=temperature,
-            max_tokens=max_tokens,
-            top_p=top_p,
-            stop=stop,
-            thinking_config=thinking_config,
-        )
-        wrapped = wrap_code_assist_request(
-            project_id=ctx.project_id,
-            model=model,
-            inner_request=inner,
-        )
-
-        headers = {
-            "Content-Type": "application/json",
-            "Accept": "application/json",
-            "Authorization": f"Bearer {access_token}",
-            "User-Agent": "hermes-agent (gemini-cli-compat)",
-            "X-Goog-Api-Client": "gl-python/hermes",
-            "x-activity-request-id": str(uuid.uuid4()),
-        }
-        headers.update(self._default_headers)
-
-        if stream:
-            return self._stream_completion(model=model, wrapped=wrapped, headers=headers)
-
-        url = f"{CODE_ASSIST_ENDPOINT}/v1internal:generateContent"
-        response = self._http.post(url, json=wrapped, headers=headers)
-        if response.status_code != 200:
-            raise _gemini_http_error(response)
-        try:
-            payload = response.json()
-        except ValueError as exc:
-            raise CodeAssistError(
-                f"Invalid JSON from Code Assist: {exc}",
-                code="code_assist_invalid_json",
-            ) from exc
-        return _translate_gemini_response(payload, model=model)
-
-    def _stream_completion(
-        self,
-        *,
-        model: str,
-        wrapped: Dict[str, Any],
-        headers: Dict[str, str],
-    ) -> Iterator[_GeminiStreamChunk]:
-        """Generator that yields OpenAI-shaped streaming chunks."""
-        url = f"{CODE_ASSIST_ENDPOINT}/v1internal:streamGenerateContent?alt=sse"
-        stream_headers = dict(headers)
-        stream_headers["Accept"] = "text/event-stream"
-
-        def _generator() -> Iterator[_GeminiStreamChunk]:
-            try:
-                with self._http.stream("POST", url, json=wrapped, headers=stream_headers) as response:
-                    if response.status_code != 200:
-                        # Materialize error body for better diagnostics
-                        response.read()
-                        raise _gemini_http_error(response)
-                    tool_call_indices: Dict[str, int] = {}
-                    for event in _iter_sse_events(response):
-                        for chunk in _translate_stream_event(event, model, tool_call_indices):
-                            yield chunk
-            except httpx.HTTPError as exc:
-                raise CodeAssistError(
-                    f"Streaming request failed: {exc}",
-                    code="code_assist_stream_error",
-                ) from exc
-
-        return _generator()
-
-
-def _gemini_http_error(response: httpx.Response) -> CodeAssistError:
-    """Translate an httpx response into a CodeAssistError with rich metadata.
-
-    Parses Google's error envelope (``{"error": {"code", "message", "status",
-    "details": [...]}}``) so the agent's error classifier can reason about
-    the failure — ``status_code`` enables the rate_limit / auth classification
-    paths, and ``response`` lets the main loop honor ``Retry-After`` just
-    like it does for OpenAI SDK exceptions.
-
-    Also lifts a few recognizable Google conditions into human-readable
-    messages so the user sees something better than a 500-char JSON dump:
-
-        MODEL_CAPACITY_EXHAUSTED → "Gemini model capacity exhausted for
-            <model>. This is a Google-side throttle..."
-        RESOURCE_EXHAUSTED w/o reason → quota-style message
-        404 → "Model <name> not found at cloudcode-pa..."
-    """
-    status = response.status_code
-
-    # Parse the body once, surviving any weird encodings.
-    body_text = ""
-    body_json: Dict[str, Any] = {}
-    try:
-        body_text = response.text
-    except Exception:
-        body_text = ""
-    if body_text:
-        try:
-            parsed = json.loads(body_text)
-            if isinstance(parsed, dict):
-                body_json = parsed
-        except (ValueError, TypeError):
-            body_json = {}
-
-    # Dig into Google's error envelope.  Shape is:
-    #   {"error": {"code": 429, "message": "...", "status": "RESOURCE_EXHAUSTED",
-    #              "details": [{"@type": ".../ErrorInfo", "reason": "MODEL_CAPACITY_EXHAUSTED",
-    #                           "metadata": {...}},
-    #                          {"@type": ".../RetryInfo", "retryDelay": "30s"}]}}
-    err_obj = body_json.get("error") if isinstance(body_json, dict) else None
-    if not isinstance(err_obj, dict):
-        err_obj = {}
-    err_status = str(err_obj.get("status") or "").strip()
-    err_message = str(err_obj.get("message") or "").strip()
-    err_details_list = err_obj.get("details") if isinstance(err_obj.get("details"), list) else []
-
-    # Extract google.rpc.ErrorInfo reason + metadata.  There may be more
-    # than one ErrorInfo (rare), so we pick the first one with a reason.
-    error_reason = ""
-    error_metadata: Dict[str, Any] = {}
-    retry_delay_seconds: Optional[float] = None
-    for detail in err_details_list:
-        if not isinstance(detail, dict):
-            continue
-        type_url = str(detail.get("@type") or "")
-        if not error_reason and type_url.endswith("/google.rpc.ErrorInfo"):
-            reason = detail.get("reason")
-            if isinstance(reason, str) and reason:
-                error_reason = reason
-            md = detail.get("metadata")
-            if isinstance(md, dict):
-                error_metadata = md
-        elif retry_delay_seconds is None and type_url.endswith("/google.rpc.RetryInfo"):
-            # retryDelay is a google.protobuf.Duration string like "30s" or "1.5s".
-            delay_raw = detail.get("retryDelay")
-            if isinstance(delay_raw, str) and delay_raw.endswith("s"):
-                try:
-                    retry_delay_seconds = float(delay_raw[:-1])
-                except ValueError:
-                    pass
-            elif isinstance(delay_raw, (int, float)):
-                retry_delay_seconds = float(delay_raw)
-
-    # Fall back to the Retry-After header if the body didn't include RetryInfo.
-    if retry_delay_seconds is None:
-        try:
-            header_val = response.headers.get("Retry-After") or response.headers.get("retry-after")
-        except Exception:
-            header_val = None
-        if header_val:
-            try:
-                retry_delay_seconds = float(header_val)
-            except (TypeError, ValueError):
-                retry_delay_seconds = None
-
-    # Classify the error code.  ``code_assist_rate_limited`` stays the default
-    # for 429s; a more specific reason tag helps downstream callers (e.g. tests,
-    # logs) without changing the rate_limit classification path.
-    code = f"code_assist_http_{status}"
-    if status == 401:
-        code = "code_assist_unauthorized"
-    elif status == 429:
-        code = "code_assist_rate_limited"
-        if error_reason == "MODEL_CAPACITY_EXHAUSTED":
-            code = "code_assist_capacity_exhausted"
-
-    # Build a human-readable message.  Keep the status + a raw-body tail for
-    # debugging, but lead with a friendlier summary when we recognize the
-    # Google signal.
-    model_hint = ""
-    if isinstance(error_metadata, dict):
-        model_hint = str(error_metadata.get("model") or error_metadata.get("modelId") or "").strip()
-
-    if status == 429 and error_reason == "MODEL_CAPACITY_EXHAUSTED":
-        target = model_hint or "this Gemini model"
-        message = (
-            f"Gemini capacity exhausted for {target} (Google-side throttle, "
-            f"not a Hermes issue). Try a different Gemini model or set a "
-            f"fallback_providers entry to a non-Gemini provider."
-        )
-        if retry_delay_seconds is not None:
-            message += f" Google suggests retrying in {retry_delay_seconds:g}s."
-    elif status == 429 and err_status == "RESOURCE_EXHAUSTED":
-        message = (
-            f"Gemini quota exhausted ({err_message or 'RESOURCE_EXHAUSTED'}). "
-            f"Check /gquota for remaining daily requests."
-        )
-        if retry_delay_seconds is not None:
-            message += f" Retry suggested in {retry_delay_seconds:g}s."
-    elif status == 404:
-        # Google returns 404 when a model has been retired or renamed.
-        target = model_hint or (err_message or "model")
-        message = (
-            f"Code Assist 404: {target} is not available at "
-            f"cloudcode-pa.googleapis.com. It may have been renamed or "
-            f"retired. Check hermes_cli/models.py for the current list."
-        )
-    elif err_message:
-        # Generic fallback with the parsed message.
-        message = f"Code Assist HTTP {status} ({err_status or 'error'}): {err_message}"
-    else:
-        # Last-ditch fallback — raw body snippet.
-        message = f"Code Assist returned HTTP {status}: {body_text[:500]}"
-
-    return CodeAssistError(
-        message,
-        code=code,
-        status_code=status,
-        response=response,
-        retry_after=retry_delay_seconds,
-        details={
-            "status": err_status,
-            "reason": error_reason,
-            "metadata": error_metadata,
-            "message": err_message,
-        },
-    )
@@ -1,453 +0,0 @@
-"""Google Code Assist API client — project discovery, onboarding, quota.
-
-The Code Assist API powers Google's official gemini-cli. It sits at
-``cloudcode-pa.googleapis.com`` and provides:
-
- Free tier access (generous daily quota) for personal Google accounts
- Paid tier access via GCP projects with billing / Workspace / Standard / Enterprise
-
-This module handles the control-plane dance needed before inference:
-
-1. ``load_code_assist()`` — probe the user's account to learn what tier they're on
-   and whether a ``cloudaicompanionProject`` is already assigned.
-2. ``onboard_user()`` — if the user hasn't been onboarded yet (new account, fresh
-   free tier, etc.), call this with the chosen tier + project id. Supports LRO
-   polling for slow provisioning.
-3. ``retrieve_user_quota()`` — fetch the ``buckets[]`` array showing remaining
-   quota per model, used by the ``/gquota`` slash command.
-
-VPC-SC handling: enterprise accounts under a VPC Service Controls perimeter
-will get ``SECURITY_POLICY_VIOLATED`` on ``load_code_assist``. We catch this
-and force the account to ``standard-tier`` so the call chain still succeeds.
-
-Derived from opencode-gemini-auth (MIT) and clawdbot/extensions/google. The
-request/response shapes are specific to Google's internal Code Assist API,
-documented nowhere public — we copy them from the reference implementations.
-"""
-
-from __future__ import annotations
-
-import json
-import logging
-import os
-import time
-import urllib.error
-import urllib.parse
-import urllib.request
-import uuid
-from dataclasses import dataclass, field
-from typing import Any, Dict, List, Optional
-
-logger = logging.getLogger(__name__)
-
-
-# =============================================================================
-# Constants
-# =============================================================================
-
-CODE_ASSIST_ENDPOINT = "https://cloudcode-pa.googleapis.com"
-
-# Fallback endpoints tried when prod returns an error during project discovery
-FALLBACK_ENDPOINTS = [
-    "https://daily-cloudcode-pa.sandbox.googleapis.com",
-    "https://autopush-cloudcode-pa.sandbox.googleapis.com",
-]
-
-# Tier identifiers that Google's API uses
-FREE_TIER_ID = "free-tier"
-LEGACY_TIER_ID = "legacy-tier"
-STANDARD_TIER_ID = "standard-tier"
-
-# Default HTTP headers matching gemini-cli's fingerprint.
-# Google may reject unrecognized User-Agents on these internal endpoints.
-_GEMINI_CLI_USER_AGENT = "google-api-nodejs-client/9.15.1 (gzip)"
-_X_GOOG_API_CLIENT = "gl-node/24.0.0"
-_DEFAULT_REQUEST_TIMEOUT = 30.0
-_ONBOARDING_POLL_ATTEMPTS = 12
-_ONBOARDING_POLL_INTERVAL_SECONDS = 5.0
-
-
-class CodeAssistError(RuntimeError):
-    """Exception raised by the Code Assist (``cloudcode-pa``) integration.
-
-    Carries HTTP status / response / retry-after metadata so the agent's
-    ``error_classifier._extract_status_code`` and the main loop's Retry-After
-    handling (which walks ``error.response.headers``) pick up the right
-    signals.  Without these, 429s from the OAuth path look like opaque
-    ``RuntimeError`` and skip the rate-limit path.
-    """
-
-    def __init__(
-        self,
-        message: str,
-        *,
-        code: str = "code_assist_error",
-        status_code: Optional[int] = None,
-        response: Any = None,
-        retry_after: Optional[float] = None,
-        details: Optional[Dict[str, Any]] = None,
-    ) -> None:
-        super().__init__(message)
-        self.code = code
-        # ``status_code`` is picked up by ``agent.error_classifier._extract_status_code``
-        # so a 429 from Code Assist classifies as FailoverReason.rate_limit and
-        # triggers the main loop's fallback_providers chain the same way SDK
-        # errors do.
-        self.status_code = status_code
-        # ``response`` is the underlying ``httpx.Response`` (or a shim with a
-        # ``.headers`` mapping and ``.json()`` method).  The main loop reads
-        # ``error.response.headers["Retry-After"]`` to honor Google's retry
-        # hints when the backend throttles us.
-        self.response = response
-        # Parsed ``Retry-After`` seconds (kept separately for convenience —
-        # Google returns retry hints in both the header and the error body's
-        # ``google.rpc.RetryInfo`` details, and we pick whichever we found).
-        self.retry_after = retry_after
-        # Parsed structured error details from the Google error envelope
-        # (e.g. ``{"reason": "MODEL_CAPACITY_EXHAUSTED", "status": "RESOURCE_EXHAUSTED"}``).
-        # Useful for logging and for tests that want to assert on specifics.
-        self.details = details or {}
-
-
-class ProjectIdRequiredError(CodeAssistError):
-    def __init__(self, message: str = "GCP project id required for this tier") -> None:
-        super().__init__(message, code="code_assist_project_id_required")
-
-
-# =============================================================================
-# HTTP primitive (auth via Bearer token passed per-call)
-# =============================================================================
-
-def _build_headers(access_token: str, *, user_agent_model: str = "") -> Dict[str, str]:
-    ua = _GEMINI_CLI_USER_AGENT
-    if user_agent_model:
-        ua = f"{ua} model/{user_agent_model}"
-    return {
-        "Content-Type": "application/json",
-        "Accept": "application/json",
-        "Authorization": f"Bearer {access_token}",
-        "User-Agent": ua,
-        "X-Goog-Api-Client": _X_GOOG_API_CLIENT,
-        "x-activity-request-id": str(uuid.uuid4()),
-    }
-
-
-def _client_metadata() -> Dict[str, str]:
-    """Match Google's gemini-cli exactly — unrecognized metadata may be rejected."""
-    return {
-        "ideType": "IDE_UNSPECIFIED",
-        "platform": "PLATFORM_UNSPECIFIED",
-        "pluginType": "GEMINI",
-    }
-
-
-def _post_json(
-    url: str,
-    body: Dict[str, Any],
-    access_token: str,
-    *,
-    timeout: float = _DEFAULT_REQUEST_TIMEOUT,
-    user_agent_model: str = "",
-) -> Dict[str, Any]:
-    data = json.dumps(body).encode("utf-8")
-    request = urllib.request.Request(
-        url, data=data, method="POST",
-        headers=_build_headers(access_token, user_agent_model=user_agent_model),
-    )
-    try:
-        with urllib.request.urlopen(request, timeout=timeout) as response:
-            raw = response.read().decode("utf-8", errors="replace")
-            return json.loads(raw) if raw else {}
-    except urllib.error.HTTPError as exc:
-        detail = ""
-        try:
-            detail = exc.read().decode("utf-8", errors="replace")
-        except Exception:
-            pass
-        # Special case: VPC-SC violation should be distinguishable
-        if _is_vpc_sc_violation(detail):
-            raise CodeAssistError(
-                f"VPC-SC policy violation: {detail}",
-                code="code_assist_vpc_sc",
-            ) from exc
-        raise CodeAssistError(
-            f"Code Assist HTTP {exc.code}: {detail or exc.reason}",
-            code=f"code_assist_http_{exc.code}",
-        ) from exc
-    except urllib.error.URLError as exc:
-        raise CodeAssistError(
-            f"Code Assist request failed: {exc}",
-            code="code_assist_network_error",
-        ) from exc
-
-
-def _is_vpc_sc_violation(body: str) -> bool:
-    """Detect a VPC Service Controls violation from a response body."""
-    if not body:
-        return False
-    try:
-        parsed = json.loads(body)
-    except (json.JSONDecodeError, ValueError):
-        return "SECURITY_POLICY_VIOLATED" in body
-    # Walk the nested error structure Google uses
-    error = parsed.get("error") if isinstance(parsed, dict) else None
-    if not isinstance(error, dict):
-        return False
-    details = error.get("details") or []
-    if isinstance(details, list):
-        for item in details:
-            if isinstance(item, dict):
-                reason = item.get("reason") or ""
-                if reason == "SECURITY_POLICY_VIOLATED":
-                    return True
-    msg = str(error.get("message", ""))
-    return "SECURITY_POLICY_VIOLATED" in msg
-
-
-# =============================================================================
-# load_code_assist — discovers current tier + assigned project
-# =============================================================================
-
-@dataclass
-class CodeAssistProjectInfo:
-    """Result from ``load_code_assist``."""
-    current_tier_id: str = ""
-    cloudaicompanion_project: str = ""   # Google-managed project (free tier)
-    allowed_tiers: List[str] = field(default_factory=list)
-    raw: Dict[str, Any] = field(default_factory=dict)
-
-
-def load_code_assist(
-    access_token: str,
-    *,
-    project_id: str = "",
-    user_agent_model: str = "",
-) -> CodeAssistProjectInfo:
-    """Call ``POST /v1internal:loadCodeAssist`` with prod → sandbox fallback.
-
-    Returns whatever tier + project info Google reports. On VPC-SC violations,
-    returns a synthetic ``standard-tier`` result so the chain can continue.
-    """
-    body: Dict[str, Any] = {
-        "metadata": {
-            "duetProject": project_id,
-            **_client_metadata(),
-        },
-    }
-    if project_id:
-        body["cloudaicompanionProject"] = project_id
-
-    endpoints = [CODE_ASSIST_ENDPOINT] + FALLBACK_ENDPOINTS
-    last_err: Optional[Exception] = None
-    for endpoint in endpoints:
-        url = f"{endpoint}/v1internal:loadCodeAssist"
-        try:
-            resp = _post_json(url, body, access_token, user_agent_model=user_agent_model)
-            return _parse_load_response(resp)
-        except CodeAssistError as exc:
-            if exc.code == "code_assist_vpc_sc":
-                logger.info("VPC-SC violation on %s — defaulting to standard-tier", endpoint)
-                return CodeAssistProjectInfo(
-                    current_tier_id=STANDARD_TIER_ID,
-                    cloudaicompanion_project=project_id,
-                )
-            last_err = exc
-            logger.warning("loadCodeAssist failed on %s: %s", endpoint, exc)
-            continue
-    if last_err:
-        raise last_err
-    return CodeAssistProjectInfo()
-
-
-def _parse_load_response(resp: Dict[str, Any]) -> CodeAssistProjectInfo:
-    current_tier = resp.get("currentTier") or {}
-    tier_id = str(current_tier.get("id") or "") if isinstance(current_tier, dict) else ""
-    project = str(resp.get("cloudaicompanionProject") or "")
-    allowed = resp.get("allowedTiers") or []
-    allowed_ids: List[str] = []
-    if isinstance(allowed, list):
-        for t in allowed:
-            if isinstance(t, dict):
-                tid = str(t.get("id") or "")
-                if tid:
-                    allowed_ids.append(tid)
-    return CodeAssistProjectInfo(
-        current_tier_id=tier_id,
-        cloudaicompanion_project=project,
-        allowed_tiers=allowed_ids,
-        raw=resp,
-    )
-
-
-# =============================================================================
-# onboard_user — provisions a new user on a tier (with LRO polling)
-# =============================================================================
-
-def onboard_user(
-    access_token: str,
-    *,
-    tier_id: str,
-    project_id: str = "",
-    user_agent_model: str = "",
-) -> Dict[str, Any]:
-    """Call ``POST /v1internal:onboardUser`` to provision the user.
-
-    For paid tiers, ``project_id`` is REQUIRED (raises ProjectIdRequiredError).
-    For free tiers, ``project_id`` is optional — Google will assign one.
-
-    Returns the final operation response. Polls ``/v1internal/<name>`` for up
-    to ``_ONBOARDING_POLL_ATTEMPTS`` × ``_ONBOARDING_POLL_INTERVAL_SECONDS``
-    (default: 12 × 5s = 1 min).
-    """
-    if tier_id != FREE_TIER_ID and tier_id != LEGACY_TIER_ID and not project_id:
-        raise ProjectIdRequiredError(
-            f"Tier {tier_id!r} requires a GCP project id. "
-            "Set HERMES_GEMINI_PROJECT_ID or GOOGLE_CLOUD_PROJECT."
-        )
-
-    body: Dict[str, Any] = {
-        "tierId": tier_id,
-        "metadata": _client_metadata(),
-    }
-    if project_id:
-        body["cloudaicompanionProject"] = project_id
-
-    endpoint = CODE_ASSIST_ENDPOINT
-    url = f"{endpoint}/v1internal:onboardUser"
-    resp = _post_json(url, body, access_token, user_agent_model=user_agent_model)
-
-    # Poll if LRO (long-running operation)
-    if not resp.get("done"):
-        op_name = resp.get("name", "")
-        if not op_name:
-            return resp
-        for attempt in range(_ONBOARDING_POLL_ATTEMPTS):
-            time.sleep(_ONBOARDING_POLL_INTERVAL_SECONDS)
-            poll_url = f"{endpoint}/v1internal/{op_name}"
-            try:
-                poll_resp = _post_json(poll_url, {}, access_token, user_agent_model=user_agent_model)
-            except CodeAssistError as exc:
-                logger.warning("Onboarding poll attempt %d failed: %s", attempt + 1, exc)
-                continue
-            if poll_resp.get("done"):
-                return poll_resp
-        logger.warning("Onboarding did not complete within %d attempts", _ONBOARDING_POLL_ATTEMPTS)
-    return resp
-
-
-# =============================================================================
-# retrieve_user_quota — for /gquota
-# =============================================================================
-
-@dataclass
-class QuotaBucket:
-    model_id: str
-    token_type: str = ""
-    remaining_fraction: float = 0.0
-    reset_time_iso: str = ""
-    raw: Dict[str, Any] = field(default_factory=dict)
-
-
-def retrieve_user_quota(
-    access_token: str,
-    *,
-    project_id: str = "",
-    user_agent_model: str = "",
-) -> List[QuotaBucket]:
-    """Call ``POST /v1internal:retrieveUserQuota`` and parse ``buckets[]``."""
-    body: Dict[str, Any] = {}
-    if project_id:
-        body["project"] = project_id
-    url = f"{CODE_ASSIST_ENDPOINT}/v1internal:retrieveUserQuota"
-    resp = _post_json(url, body, access_token, user_agent_model=user_agent_model)
-    raw_buckets = resp.get("buckets") or []
-    buckets: List[QuotaBucket] = []
-    if not isinstance(raw_buckets, list):
-        return buckets
-    for b in raw_buckets:
-        if not isinstance(b, dict):
-            continue
-        buckets.append(QuotaBucket(
-            model_id=str(b.get("modelId") or ""),
-            token_type=str(b.get("tokenType") or ""),
-            remaining_fraction=float(b.get("remainingFraction") or 0.0),
-            reset_time_iso=str(b.get("resetTime") or ""),
-            raw=b,
-        ))
-    return buckets
-
-
-# =============================================================================
-# Project context resolution
-# =============================================================================
-
-@dataclass
-class ProjectContext:
-    """Resolved state for a given OAuth session."""
-    project_id: str = ""           # effective project id sent on requests
-    managed_project_id: str = ""   # Google-assigned project (free tier)
-    tier_id: str = ""
-    source: str = ""               # "env", "config", "discovered", "onboarded"
-
-
-def resolve_project_context(
-    access_token: str,
-    *,
-    configured_project_id: str = "",
-    env_project_id: str = "",
-    user_agent_model: str = "",
-) -> ProjectContext:
-    """Figure out what project id + tier to use for requests.
-
-    Priority:
-      1. If configured_project_id or env_project_id is set, use that directly
-         and short-circuit (no discovery needed).
-      2. Otherwise call loadCodeAssist to see what Google says.
-      3. If no tier assigned yet, onboard the user (free tier default).
-    """
-    # Short-circuit: caller provided a project id
-    if configured_project_id:
-        return ProjectContext(
-            project_id=configured_project_id,
-            tier_id=STANDARD_TIER_ID,  # assume paid since they specified one
-            source="config",
-        )
-    if env_project_id:
-        return ProjectContext(
-            project_id=env_project_id,
-            tier_id=STANDARD_TIER_ID,
-            source="env",
-        )
-
-    # Discover via loadCodeAssist
-    info = load_code_assist(access_token, user_agent_model=user_agent_model)
-
-    effective_project = info.cloudaicompanion_project
-    tier = info.current_tier_id
-
-    if not tier:
-        # User hasn't been onboarded — provision them on free tier
-        onboard_resp = onboard_user(
-            access_token,
-            tier_id=FREE_TIER_ID,
-            project_id="",
-            user_agent_model=user_agent_model,
-        )
-        # Re-parse from the onboard response
-        response_body = onboard_resp.get("response") or {}
-        if isinstance(response_body, dict):
-            effective_project = (
-                effective_project
-                or str(response_body.get("cloudaicompanionProject") or "")
-            )
-        tier = FREE_TIER_ID
-        source = "onboarded"
-    else:
-        source = "discovered"
-
-    return ProjectContext(
-        project_id=effective_project,
-        managed_project_id=effective_project if tier == FREE_TIER_ID else "",
-        tier_id=tier,
-        source=source,
-    )
@@ -27,6 +27,7 @@ from agent.usage_pricing import (
    DEFAULT_PRICING,
    estimate_usage_cost,
    format_duration_compact,
+    get_pricing,
    has_known_pricing,
 )

@@ -38,6 +39,15 @@ def _has_known_pricing(model_name: str, provider: str = None, base_url: str = No
    return has_known_pricing(model_name, provider=provider, base_url=base_url)


+def _get_pricing(model_name: str) -> Dict[str, float]:
+    """Look up pricing for a model. Uses fuzzy matching on model name.
+
+    Returns _DEFAULT_PRICING (zero cost) for unknown/custom models —
+    we can't assume costs for self-hosted endpoints, local inference, etc.
+    """
+    return get_pricing(model_name)
+
+
 def _estimate_cost(
    session_or_model: Dict[str, Any] | str,
    input_tokens: int = 0,
@@ -634,7 +644,13 @@ class InsightsEngine:
        lines.append(f"  Sessions:          {o['total_sessions']:<12}  Messages:        {o['total_messages']:,}")
        lines.append(f"  Tool calls:        {o['total_tool_calls']:<12,}  User messages:   {o['user_messages']:,}")
        lines.append(f"  Input tokens:      {o['total_input_tokens']:<12,}  Output tokens:   {o['total_output_tokens']:,}")
-        lines.append(f"  Total tokens:      {o['total_tokens']:,}")
+        cache_total = o.get("total_cache_read_tokens", 0) + o.get("total_cache_write_tokens", 0)
+        if cache_total > 0:
+            lines.append(f"  Cache read:        {o['total_cache_read_tokens']:<12,}  Cache write:     {o['total_cache_write_tokens']:,}")
+        cost_str = f"${o['estimated_cost']:.2f}"
+        if o.get("models_without_pricing"):
+            cost_str += " *"
+        lines.append(f"  Total tokens:      {o['total_tokens']:<12,}  Est. cost:       {cost_str}")
        if o["total_hours"] > 0:
            lines.append(f"  Active time:       ~{_format_duration(o['total_hours'] * 3600):<11}  Avg session:     ~{_format_duration(o['avg_session_duration'])}")
        lines.append(f"  Avg msgs/session:  {o['avg_messages_per_session']:.1f}")
@@ -644,10 +660,16 @@ class InsightsEngine:
        if report["models"]:
            lines.append("  🤖 Models Used")
            lines.append("  " + "─" * 56)
-            lines.append(f"  {'Model':<30} {'Sessions':>8} {'Tokens':>12}")
+            lines.append(f"  {'Model':<30} {'Sessions':>8} {'Tokens':>12} {'Cost':>8}")
            for m in report["models"]:
                model_name = m["model"][:28]
-                lines.append(f"  {model_name:<30} {m['sessions']:>8} {m['total_tokens']:>12,}")
+                if m.get("has_pricing"):
+                    cost_cell = f"${m['cost']:>6.2f}"
+                else:
+                    cost_cell = "     N/A"
+                lines.append(f"  {model_name:<30} {m['sessions']:>8} {m['total_tokens']:>12,} {cost_cell}")
+            if o.get("models_without_pricing"):
+                lines.append("  * Cost N/A for custom/self-hosted models")
            lines.append("")

        # Platform breakdown
@@ -727,7 +749,15 @@ class InsightsEngine:

        # Overview
        lines.append(f"**Sessions:** {o['total_sessions']} | **Messages:** {o['total_messages']:,} | **Tool calls:** {o['total_tool_calls']:,}")
-        lines.append(f"**Tokens:** {o['total_tokens']:,} (in: {o['total_input_tokens']:,} / out: {o['total_output_tokens']:,})")
+        cache_total = o.get("total_cache_read_tokens", 0) + o.get("total_cache_write_tokens", 0)
+        if cache_total > 0:
+            lines.append(f"**Tokens:** {o['total_tokens']:,} (in: {o['total_input_tokens']:,} / out: {o['total_output_tokens']:,} / cache: {cache_total:,})")
+        else:
+            lines.append(f"**Tokens:** {o['total_tokens']:,} (in: {o['total_input_tokens']:,} / out: {o['total_output_tokens']:,})")
+        cost_note = ""
+        if o.get("models_without_pricing"):
+            cost_note = " _(excludes custom/self-hosted models)_"
+        lines.append(f"**Est. cost:** ${o['estimated_cost']:.2f}{cost_note}")
        if o["total_hours"] > 0:
            lines.append(f"**Active time:** ~{_format_duration(o['total_hours'] * 3600)} | **Avg session:** ~{_format_duration(o['avg_session_duration'])}")
        lines.append("")
@@ -736,7 +766,8 @@ class InsightsEngine:
        if report["models"]:
            lines.append("**🤖 Models:**")
            for m in report["models"][:5]:
-                lines.append(f"  {m['model'][:25]} — {m['sessions']} sessions, {m['total_tokens']:,} tokens")
+                cost_str = f"${m['cost']:.2f}" if m.get("has_pricing") else "N/A"
+                lines.append(f"  {m['model'][:25]} — {m['sessions']} sessions, {m['total_tokens']:,} tokens, {cost_str}")
            lines.append("")

        # Platforms (if multi-platform)
@@ -1,49 +0,0 @@
-"""User-facing summaries for manual compression commands."""
-
-from __future__ import annotations
-
-from typing import Any, Sequence
-
-
-def summarize_manual_compression(
-    before_messages: Sequence[dict[str, Any]],
-    after_messages: Sequence[dict[str, Any]],
-    before_tokens: int,
-    after_tokens: int,
-) -> dict[str, Any]:
-    """Return consistent user-facing feedback for manual compression."""
-    before_count = len(before_messages)
-    after_count = len(after_messages)
-    noop = list(after_messages) == list(before_messages)
-
-    if noop:
-        headline = f"No changes from compression: {before_count} messages"
-        if after_tokens == before_tokens:
-            token_line = (
-                f"Rough transcript estimate: ~{before_tokens:,} tokens (unchanged)"
-            )
-        else:
-            token_line = (
-                f"Rough transcript estimate: ~{before_tokens:,} → "
-                f"~{after_tokens:,} tokens"
-            )
-    else:
-        headline = f"Compressed: {before_count} → {after_count} messages"
-        token_line = (
-            f"Rough transcript estimate: ~{before_tokens:,} → "
-            f"~{after_tokens:,} tokens"
-        )
-
-    note = None
-    if not noop and after_count < before_count and after_tokens > before_tokens:
-        note = (
-            "Note: fewer messages can still raise this rough transcript estimate "
-            "when compression rewrites the transcript into denser summaries."
-        )
-
-    return {
-        "noop": noop,
-        "headline": headline,
-        "token_line": token_line,
-        "note": note,
-    }
@@ -44,22 +44,11 @@ logger = logging.getLogger(__name__)
 # ---------------------------------------------------------------------------

 _FENCE_TAG_RE = re.compile(r'</?\s*memory-context\s*>', re.IGNORECASE)
-_INTERNAL_CONTEXT_RE = re.compile(
-    r'<\s*memory-context\s*>[\s\S]*?</\s*memory-context\s*>',
-    re.IGNORECASE,
-)
-_INTERNAL_NOTE_RE = re.compile(
-    r'\[System note:\s*The following is recalled memory context,\s*NOT new user input\.\s*Treat as informational background data\.\]\s*',
-    re.IGNORECASE,
-)


 def sanitize_context(text: str) -> str:
-    """Strip fence tags, injected context blocks, and system notes from provider output."""
-    text = _INTERNAL_CONTEXT_RE.sub('', text)
-    text = _INTERNAL_NOTE_RE.sub('', text)
-    text = _FENCE_TAG_RE.sub('', text)
-    return text
+    """Strip fence-escape sequences from provider output."""
+    return _FENCE_TAG_RE.sub('', text)


 def build_memory_context_block(raw_context: str) -> str:
@@ -145,6 +134,11 @@ class MemoryManager:
        """All registered providers in order."""
        return list(self._providers)

+    @property
+    def provider_names(self) -> List[str]:
+        """Names of all registered providers."""
+        return [p.name for p in self._providers]
+
    def get_provider(self, name: str) -> Optional[MemoryProvider]:
        """Get a provider by name, or None if not registered."""
        for p in self._providers:
@@ -5,6 +5,7 @@ and run_agent.py for pre-flight context checks.
 """

 import logging
+import os
 import re
 import time
 from pathlib import Path
@@ -23,22 +24,15 @@ logger = logging.getLogger(__name__)
 # are preserved so the full model name reaches cache lookups and server queries.
 _PROVIDER_PREFIXES: frozenset[str] = frozenset({
    "openrouter", "nous", "openai-codex", "copilot", "copilot-acp",
-    "gemini", "ollama-cloud", "zai", "kimi-coding", "kimi-coding-cn", "minimax", "minimax-cn", "anthropic", "deepseek",
+    "gemini", "zai", "kimi-coding", "minimax", "minimax-cn", "anthropic", "deepseek",
    "opencode-zen", "opencode-go", "ai-gateway", "kilocode", "alibaba",
    "qwen-oauth",
-    "xiaomi",
-    "arcee",
    "custom", "local",
    # Common aliases
    "google", "google-gemini", "google-ai-studio",
    "glm", "z-ai", "z.ai", "zhipu", "github", "github-copilot",
-    "github-models", "kimi", "moonshot", "kimi-cn", "moonshot-cn", "claude", "deep-seek",
-    "ollama",
+    "github-models", "kimi", "moonshot", "claude", "deep-seek",
    "opencode", "zen", "go", "vercel", "kilo", "dashscope", "aliyun", "qwen",
-    "mimo", "xiaomi-mimo",
-    "arcee-ai", "arceeai",
-    "xai", "x-ai", "x.ai", "grok",
-    "nvidia", "nim", "nvidia-nim", "nemotron",
    "qwen-portal",
 })

@@ -89,11 +83,6 @@ CONTEXT_PROBE_TIERS = [
 # Default context length when no detection method succeeds.
 DEFAULT_FALLBACK_CONTEXT = CONTEXT_PROBE_TIERS[0]

-# Minimum context length required to run Hermes Agent.  Models with fewer
-# tokens cannot maintain enough working memory for tool-calling workflows.
-# Sessions, model switches, and cron jobs should reject models below this.
-MINIMUM_CONTEXT_LENGTH = 64_000
-
 # Thin fallback defaults — only broad model family patterns.
 # These fire only when provider is unknown AND models.dev/OpenRouter/Anthropic
 # all miss. Replaced the previous 80+ entry dict.
@@ -103,78 +92,54 @@ DEFAULT_CONTEXT_LENGTHS = {
    # fuzzy-match collisions (e.g. "anthropic/claude-sonnet-4" is a
    # substring of "anthropic/claude-sonnet-4.6").
    # OpenRouter-prefixed models resolve via OpenRouter live API or models.dev.
-    "claude-opus-4-7": 1000000,
-    "claude-opus-4.7": 1000000,
    "claude-opus-4-6": 1000000,
    "claude-sonnet-4-6": 1000000,
    "claude-opus-4.6": 1000000,
    "claude-sonnet-4.6": 1000000,
    # Catch-all for older Claude models (must sort after specific entries)
    "claude": 200000,
-    # OpenAI — GPT-5 family (most have 400k; specific overrides first)
-    # Source: https://developers.openai.com/api/docs/models
-    "gpt-5.4-nano": 400000,           # 400k (not 1.05M like full 5.4)
-    "gpt-5.4-mini": 400000,           # 400k (not 1.05M like full 5.4)
-    "gpt-5.4": 1050000,               # GPT-5.4, GPT-5.4 Pro (1.05M context)
-    "gpt-5.3-codex-spark": 128000,    # Spark variant has reduced 128k context
-    "gpt-5.1-chat": 128000,           # Chat variant has 128k context
-    "gpt-5": 400000,                  # GPT-5.x base, mini, codex variants (400k)
+    # OpenAI
    "gpt-4.1": 1047576,
+    "gpt-5": 128000,
    "gpt-4": 128000,
    # Google
    "gemini": 1048576,
    # Gemma (open models served via AI Studio)
    "gemma-4-31b": 256000,
+    "gemma-4-26b": 256000,
    "gemma-3": 131072,
    "gemma": 8192,  # fallback for older gemma models
    # DeepSeek
    "deepseek": 128000,
    # Meta
    "llama": 131072,
-    # Qwen — specific model families before the catch-all.
-    # Official docs: https://help.aliyun.com/zh/model-studio/developer-reference/
-    "qwen3-coder-plus": 1000000,  # 1M context
-    "qwen3-coder": 262144,        # 256K context
+    # Qwen
    "qwen": 131072,
-    # MiniMax — official docs: 204,800 context for all models
-    # https://platform.minimax.io/docs/api-reference/text-anthropic-api
-    "minimax": 204800,
+    # MiniMax (lowercase — lookup lowercases model names at line 973)
+    "minimax-m1-256k": 1000000,
+    "minimax-m1-128k": 1000000,
+    "minimax-m1-80k": 1000000,
+    "minimax-m1-40k": 1000000,
+    "minimax-m1": 1000000,
+    "minimax-m2.5": 1048576,
+    "minimax-m2.7": 1048576,
+    "minimax": 1048576,
    # GLM
    "glm": 202752,
-    # xAI Grok — xAI /v1/models does not return context_length metadata,
-    # so these hardcoded fallbacks prevent Hermes from probing-down to
-    # the default 128k when the user points at https://api.x.ai/v1
-    # via a custom provider. Values sourced from models.dev (2026-04).
-    # Keys use substring matching (longest-first), so e.g. "grok-4.20"
-    # matches "grok-4.20-0309-reasoning" / "-non-reasoning" / "-multi-agent-0309".
-    "grok-code-fast": 256000,   # grok-code-fast-1
-    "grok-4-1-fast": 2000000,   # grok-4-1-fast-(non-)reasoning
-    "grok-2-vision": 8192,      # grok-2-vision, -1212, -latest
-    "grok-4-fast": 2000000,     # grok-4-fast-(non-)reasoning
-    "grok-4.20": 2000000,       # grok-4.20-0309-(non-)reasoning, -multi-agent-0309
-    "grok-4": 256000,           # grok-4, grok-4-0709
-    "grok-3": 131072,           # grok-3, grok-3-mini, grok-3-fast, grok-3-mini-fast
-    "grok-2": 131072,           # grok-2, grok-2-1212, grok-2-latest
-    "grok": 131072,             # catch-all (grok-beta, unknown grok-*)
    # Kimi
    "kimi": 262144,
-    # Nemotron — NVIDIA's open-weights series (128K context across all sizes)
-    "nemotron": 131072,
    # Arcee
    "trinity": 262144,
-    # OpenRouter
-    "elephant": 262144,
    # Hugging Face Inference Providers — model IDs use org/name format
    "Qwen/Qwen3.5-397B-A17B": 131072,
    "Qwen/Qwen3.5-35B-A3B": 131072,
    "deepseek-ai/DeepSeek-V3.2": 65536,
    "moonshotai/Kimi-K2.5": 262144,
    "moonshotai/Kimi-K2-Thinking": 262144,
-    "MiniMaxAI/MiniMax-M2.5": 204800,
-    "XiaomiMiMo/MiMo-V2-Flash": 256000,
-    "mimo-v2-pro": 1000000,
-    "mimo-v2-omni": 256000,
-    "mimo-v2-flash": 256000,
+    "MiniMaxAI/MiniMax-M2.5": 1048576,
+    "XiaomiMiMo/MiMo-V2-Flash": 32768,
+    "mimo-v2-pro": 1048576,
+    "mimo-v2-omni": 1048576,
    "zai-org/GLM-5": 202752,
 }

@@ -199,12 +164,6 @@ _MAX_COMPLETION_KEYS = (

 # Local server hostnames / address patterns
 _LOCAL_HOSTS = ("localhost", "127.0.0.1", "::1", "0.0.0.0")
-# Docker / Podman / Lima DNS names that resolve to the host machine
-_CONTAINER_LOCAL_SUFFIXES = (
-    ".docker.internal",
-    ".containers.internal",
-    ".lima.internal",
-)


 def _normalize_base_url(base_url: str) -> str:
@@ -226,9 +185,7 @@ _URL_TO_PROVIDER: Dict[str, str] = {
    "api.anthropic.com": "anthropic",
    "api.z.ai": "zai",
    "api.moonshot.ai": "kimi-coding",
-    "api.moonshot.cn": "kimi-coding-cn",
    "api.kimi.com": "kimi-coding",
-    "api.arcee.ai": "arcee",
    "api.minimax": "minimax",
    "dashscope.aliyuncs.com": "alibaba",
    "dashscope-intl.aliyuncs.com": "alibaba",
@@ -240,12 +197,6 @@ _URL_TO_PROVIDER: Dict[str, str] = {
    "api.githubcopilot.com": "copilot",
    "models.github.ai": "copilot",
    "api.fireworks.ai": "fireworks",
-    "opencode.ai": "opencode-go",
-    "api.x.ai": "xai",
-    "integrate.api.nvidia.com": "nvidia",
-    "api.xiaomimimo.com": "xiaomi",
-    "xiaomimimo.com": "xiaomi",
-    "ollama.com": "ollama-cloud",
 }


@@ -284,9 +235,6 @@ def is_local_endpoint(base_url: str) -> bool:
        return False
    if host in _LOCAL_HOSTS:
        return True
-    # Docker / Podman / Lima internal DNS names (e.g. host.docker.internal)
-    if any(host.endswith(suffix) for suffix in _CONTAINER_LOCAL_SUFFIXES):
-        return True
    # RFC-1918 private ranges and link-local
    import ipaddress
    try:
@@ -654,49 +602,6 @@ def parse_context_limit_from_error(error_msg: str) -> Optional[int]:
    return None


-def parse_available_output_tokens_from_error(error_msg: str) -> Optional[int]:
-    """Detect an "output cap too large" error and return how many output tokens are available.
-
-    Background — two distinct context errors exist:
-      1. "Prompt too long"  — the INPUT itself exceeds the context window.
-           Fix: compress history and/or halve context_length.
-      2. "max_tokens too large" — input is fine, but input + requested_output > window.
-           Fix: reduce max_tokens (the output cap) for this call.
-           Do NOT touch context_length — the window hasn't shrunk.
-
-    Anthropic's API returns errors like:
-      "max_tokens: 32768 > context_window: 200000 - input_tokens: 190000 = available_tokens: 10000"
-
-    Returns the number of output tokens that would fit (e.g. 10000 above), or None if
-    the error does not look like a max_tokens-too-large error.
-    """
-    error_lower = error_msg.lower()
-
-    # Must look like an output-cap error, not a prompt-length error.
-    is_output_cap_error = (
-        "max_tokens" in error_lower
-        and ("available_tokens" in error_lower or "available tokens" in error_lower)
-    )
-    if not is_output_cap_error:
-        return None
-
-    # Extract the available_tokens figure.
-    # Anthropic format: "… = available_tokens: 10000"
-    patterns = [
-        r'available_tokens[:\s]+(\d+)',
-        r'available\s+tokens[:\s]+(\d+)',
-        # fallback: last number after "=" in expressions like "200000 - 190000 = 10000"
-        r'=\s*(\d+)\s*$',
-    ]
-    for pattern in patterns:
-        match = re.search(pattern, error_lower)
-        if match:
-            tokens = int(match.group(1))
-            if tokens >= 1:
-                return tokens
-    return None
-
-
 def _model_id_matches(candidate_id: str, lookup_model: str) -> bool:
    """Return True if *candidate_id* (from server) matches *lookup_model* (configured).

@@ -794,12 +699,12 @@ def _query_local_context_length(model: str, base_url: str) -> Optional[int]:
                resp = client.post(f"{server_url}/api/show", json={"name": model})
                if resp.status_code == 200:
                    data = resp.json()
-                    # Prefer explicit num_ctx from Modelfile parameters: this is
-                    # the *runtime* context Ollama will actually allocate KV cache
-                    # for. The GGUF model_info.context_length is the training max,
-                    # which can be larger than num_ctx — using it here would let
-                    # Hermes grow conversations past the runtime limit and Ollama
-                    # would silently truncate. Matches query_ollama_num_ctx().
+                    # Check model_info for context length
+                    model_info = data.get("model_info", {})
+                    for key, value in model_info.items():
+                        if "context_length" in key and isinstance(value, (int, float)):
+                            return int(value)
+                    # Check parameters string for num_ctx
                    params = data.get("parameters", "")
                    if "num_ctx" in params:
                        for line in params.split("\n"):
@@ -810,11 +715,6 @@ def _query_local_context_length(model: str, base_url: str) -> Optional[int]:
                                        return int(parts[-1])
                                    except ValueError:
                                        pass
-                    # Fall back to GGUF model_info context_length (training max)
-                    model_info = data.get("model_info", {})
-                    for key, value in model_info.items():
-                        if "context_length" in key and isinstance(value, (int, float)):
-                            return int(value)

            # LM Studio native API: /api/v1/models returns max_context_length.
            # This is more reliable than the OpenAI-compat /v1/models which
@@ -1019,16 +919,6 @@ def get_model_context_length(
        if ctx:
            return ctx

-    # 4b. AWS Bedrock — use static context length table.
-    # Bedrock's ListFoundationModels doesn't expose context window sizes,
-    # so we maintain a curated table in bedrock_adapter.py.
-    if provider == "bedrock" or (base_url and "bedrock-runtime" in base_url):
-        try:
-            from agent.bedrock_adapter import get_bedrock_context_length
-            return get_bedrock_context_length(model)
-        except ImportError:
-            pass  # boto3 not installed — fall through to generic resolution
-
    # 5. Provider-aware lookups (before generic OpenRouter cache)
    # These are provider-specific and take priority over the generic OR cache,
    # since the same model can have different context limits per provider
@@ -1079,21 +969,16 @@ def get_model_context_length(


 def estimate_tokens_rough(text: str) -> int:
-    """Rough token estimate (~4 chars/token) for pre-flight checks.
-
-    Uses ceiling division so short texts (1-3 chars) never estimate as
-    0 tokens, which would cause the compressor and pre-flight checks to
-    systematically undercount when many short tool results are present.
-    """
+    """Rough token estimate (~4 chars/token) for pre-flight checks."""
    if not text:
        return 0
-    return (len(text) + 3) // 4
+    return len(text) // 4


 def estimate_messages_tokens_rough(messages: List[Dict[str, Any]]) -> int:
    """Rough token estimate for a message list (pre-flight only)."""
    total_chars = sum(len(str(msg)) for msg in messages)
-    return (total_chars + 3) // 4
+    return total_chars // 4


 def estimate_request_tokens_rough(
@@ -1116,4 +1001,4 @@ def estimate_request_tokens_rough(
        total_chars += sum(len(str(msg)) for msg in messages)
    if tools:
        total_chars += len(str(tools))
-    return (total_chars + 3) // 4
+    return total_chars // 4
@@ -18,8 +18,10 @@ Other modules should import the dataclasses and query functions from here
 rather than parsing the raw JSON themselves.
 """

+import difflib
 import json
 import logging
+import os
 import time
 from dataclasses import dataclass
 from pathlib import Path
@@ -133,6 +135,9 @@ class ProviderInfo:
    doc: str = ""                   # documentation URL
    model_count: int = 0

+    def has_api_url(self) -> bool:
+        return bool(self.api)
+

 # ---------------------------------------------------------------------------
 # Provider ID mapping: Hermes ↔ models.dev
@@ -142,11 +147,8 @@ class ProviderInfo:
 PROVIDER_TO_MODELS_DEV: Dict[str, str] = {
    "openrouter": "openrouter",
    "anthropic": "anthropic",
-    "openai": "openai",
-    "openai-codex": "openai",
    "zai": "zai",
    "kimi-coding": "kimi-for-coding",
-    "kimi-coding-cn": "kimi-for-coding",
    "minimax": "minimax",
    "minimax-cn": "minimax-cn",
    "deepseek": "deepseek",
@@ -162,20 +164,25 @@ PROVIDER_TO_MODELS_DEV: Dict[str, str] = {
    "gemini": "google",
    "google": "google",
    "xai": "xai",
-    "xiaomi": "xiaomi",
    "nvidia": "nvidia",
    "groq": "groq",
    "mistral": "mistral",
    "togetherai": "togetherai",
    "perplexity": "perplexity",
    "cohere": "cohere",
-    "ollama-cloud": "ollama-cloud",
 }

 # Reverse mapping: models.dev → Hermes (built lazily)
 _MODELS_DEV_TO_PROVIDER: Optional[Dict[str, str]] = None


+def _get_reverse_mapping() -> Dict[str, str]:
+    """Return models.dev ID → Hermes provider ID mapping."""
+    global _MODELS_DEV_TO_PROVIDER
+    if _MODELS_DEV_TO_PROVIDER is None:
+        _MODELS_DEV_TO_PROVIDER = {v: k for k, v in PROVIDER_TO_MODELS_DEV.items()}
+    return _MODELS_DEV_TO_PROVIDER
+

 def _get_cache_path() -> Path:
    """Return path to disk cache file."""
@@ -379,14 +386,7 @@ def get_model_capabilities(provider: str, model: str) -> Optional[ModelCapabilit

    # Extract capability flags (default to False if missing)
    supports_tools = bool(entry.get("tool_call", False))
-    # Vision: check both the `attachment` flag and `modalities.input` for "image".
-    # Some models (e.g. gemma-4) list image in input modalities but not attachment.
-    input_mods = entry.get("modalities", {})
-    if isinstance(input_mods, dict):
-        input_mods = input_mods.get("input", [])
-    else:
-        input_mods = []
-    supports_vision = bool(entry.get("attachment", False)) or "image" in input_mods
+    supports_vision = bool(entry.get("attachment", False))
    supports_reasoning = bool(entry.get("reasoning", False))

    # Extract limits
@@ -420,10 +420,7 @@ def list_provider_models(provider: str) -> List[str]:
    models = _get_provider_models(provider)
    if models is None:
        return []
-    return [
-        mid for mid in models.keys()
-        if not _should_hide_from_provider_catalog(provider, mid)
-    ]
+    return list(models.keys())


 # Patterns that indicate non-agentic or noise models (TTS, embedding,
@@ -435,43 +432,6 @@ _NOISE_PATTERNS: re.Pattern = re.compile(
    re.IGNORECASE,
 )

-# Google's live Gemini catalogs currently include a mix of stale slugs and
-# Gemma models whose TPM quotas are too small for normal Hermes agent traffic.
-# Keep capability metadata available for direct/manual use, but hide these from
-# the Gemini model catalogs we surface in setup and model selection.
-_GOOGLE_HIDDEN_MODELS = frozenset({
-    # Low-TPM Gemma models that trip Google input-token quota walls under
-    # agent-style traffic despite advertising large context windows.
-    "gemma-4-31b-it",
-    "gemma-4-26b-it",
-    "gemma-4-26b-a4b-it",
-    "gemma-3-1b",
-    "gemma-3-1b-it",
-    "gemma-3-2b",
-    "gemma-3-2b-it",
-    "gemma-3-4b",
-    "gemma-3-4b-it",
-    "gemma-3-12b",
-    "gemma-3-12b-it",
-    "gemma-3-27b",
-    "gemma-3-27b-it",
-    # Stale/retired Google slugs that still surface through models.dev-backed
-    # Gemini selection but 404 on the current Google endpoints.
-    "gemini-1.5-flash",
-    "gemini-1.5-pro",
-    "gemini-1.5-flash-8b",
-    "gemini-2.0-flash",
-    "gemini-2.0-flash-lite",
-})
-
-
-def _should_hide_from_provider_catalog(provider: str, model_id: str) -> bool:
-    provider_lower = (provider or "").strip().lower()
-    model_lower = (model_id or "").strip().lower()
-    if provider_lower in {"gemini", "google"} and model_lower in _GOOGLE_HIDDEN_MODELS:
-        return True
-    return False
-

 def list_agentic_models(provider: str) -> List[str]:
    """Return model IDs suitable for agentic use from models.dev.
@@ -488,8 +448,6 @@ def list_agentic_models(provider: str) -> List[str]:
    for mid, entry in models.items():
        if not isinstance(entry, dict):
            continue
-        if _should_hide_from_provider_catalog(provider, mid):
-            continue
        if not entry.get("tool_call", False):
            continue
        if _NOISE_PATTERNS.search(mid):
@@ -498,6 +456,93 @@ def list_agentic_models(provider: str) -> List[str]:
    return result


+def search_models_dev(
+    query: str, provider: str = None, limit: int = 5
+) -> List[Dict[str, Any]]:
+    """Fuzzy search across models.dev catalog. Returns matching model entries.
+
+    Args:
+        query: Search string to match against model IDs.
+        provider: Optional Hermes provider ID to restrict search scope.
+                  If None, searches across all providers in PROVIDER_TO_MODELS_DEV.
+        limit: Maximum number of results to return.
+
+    Returns:
+        List of dicts, each containing 'provider', 'model_id', and the full
+        model 'entry' from models.dev.
+    """
+    data = fetch_models_dev()
+    if not data:
+        return []
+
+    # Build list of (provider_id, model_id, entry) candidates
+    candidates: List[tuple] = []
+
+    if provider is not None:
+        # Search only the specified provider
+        mdev_provider_id = PROVIDER_TO_MODELS_DEV.get(provider)
+        if not mdev_provider_id:
+            return []
+        provider_data = data.get(mdev_provider_id, {})
+        if isinstance(provider_data, dict):
+            models = provider_data.get("models", {})
+            if isinstance(models, dict):
+                for mid, mdata in models.items():
+                    candidates.append((provider, mid, mdata))
+    else:
+        # Search across all mapped providers
+        for hermes_prov, mdev_prov in PROVIDER_TO_MODELS_DEV.items():
+            provider_data = data.get(mdev_prov, {})
+            if isinstance(provider_data, dict):
+                models = provider_data.get("models", {})
+                if isinstance(models, dict):
+                    for mid, mdata in models.items():
+                        candidates.append((hermes_prov, mid, mdata))
+
+    if not candidates:
+        return []
+
+    # Use difflib for fuzzy matching — case-insensitive comparison
+    model_ids_lower = [c[1].lower() for c in candidates]
+    query_lower = query.lower()
+
+    # First try exact substring matches (more intuitive than pure edit-distance)
+    substring_matches = []
+    for prov, mid, mdata in candidates:
+        if query_lower in mid.lower():
+            substring_matches.append({"provider": prov, "model_id": mid, "entry": mdata})
+
+    # Then add difflib fuzzy matches for any remaining slots
+    fuzzy_ids = difflib.get_close_matches(
+        query_lower, model_ids_lower, n=limit * 2, cutoff=0.4
+    )
+
+    seen_ids: set = set()
+    results: List[Dict[str, Any]] = []
+
+    # Prioritize substring matches
+    for match in substring_matches:
+        key = (match["provider"], match["model_id"])
+        if key not in seen_ids:
+            seen_ids.add(key)
+            results.append(match)
+            if len(results) >= limit:
+                return results
+
+    # Add fuzzy matches
+    for fid in fuzzy_ids:
+        # Find original-case candidates matching this lowered ID
+        for prov, mid, mdata in candidates:
+            if mid.lower() == fid:
+                key = (prov, mid)
+                if key not in seen_ids:
+                    seen_ids.add(key)
+                    results.append({"provider": prov, "model_id": mid, "entry": mdata})
+                    if len(results) >= limit:
+                        return results
+
+    return results
+

 # ---------------------------------------------------------------------------
 # Rich dataclass constructors — parse raw models.dev JSON into dataclasses
@@ -589,6 +634,43 @@ def get_provider_info(provider_id: str) -> Optional[ProviderInfo]:
    return _parse_provider_info(mdev_id, raw)


+def list_all_providers() -> Dict[str, ProviderInfo]:
+    """Return all providers from models.dev as {provider_id: ProviderInfo}.
+
+    Returns the full catalog — 109+ providers.  For providers that have
+    a Hermes alias, both the models.dev ID and the Hermes ID are included.
+    """
+    data = fetch_models_dev()
+    result: Dict[str, ProviderInfo] = {}
+
+    for pid, pdata in data.items():
+        if isinstance(pdata, dict):
+            info = _parse_provider_info(pid, pdata)
+            result[pid] = info
+
+    return result
+
+
+def get_providers_for_env_var(env_var: str) -> List[str]:
+    """Reverse lookup: find all providers that use a given env var.
+
+    Useful for auto-detection: "user has ANTHROPIC_API_KEY set, which
+    providers does that enable?"
+
+    Returns list of models.dev provider IDs.
+    """
+    data = fetch_models_dev()
+    matches: List[str] = []
+
+    for pid, pdata in data.items():
+        if isinstance(pdata, dict):
+            env = pdata.get("env", [])
+            if isinstance(env, list) and env_var in env:
+                matches.append(pid)
+
+    return matches
+
+
 # ---------------------------------------------------------------------------
 # Model-level queries (rich ModelInfo)
 # ---------------------------------------------------------------------------
@@ -624,3 +706,76 @@ def get_model_info(
            return _parse_model_info(mid, mdata, mdev_id)

    return None
+
+
+def get_model_info_any_provider(model_id: str) -> Optional[ModelInfo]:
+    """Search all providers for a model by ID.
+
+    Useful when you have a full slug like "anthropic/claude-sonnet-4.6" or
+    a bare name and want to find it anywhere.  Checks Hermes-mapped providers
+    first, then falls back to all models.dev providers.
+    """
+    data = fetch_models_dev()
+
+    # Try Hermes-mapped providers first (more likely what the user wants)
+    for hermes_id, mdev_id in PROVIDER_TO_MODELS_DEV.items():
+        pdata = data.get(mdev_id)
+        if not isinstance(pdata, dict):
+            continue
+        models = pdata.get("models", {})
+        if not isinstance(models, dict):
+            continue
+
+        raw = models.get(model_id)
+        if isinstance(raw, dict):
+            return _parse_model_info(model_id, raw, mdev_id)
+
+        # Case-insensitive
+        model_lower = model_id.lower()
+        for mid, mdata in models.items():
+            if mid.lower() == model_lower and isinstance(mdata, dict):
+                return _parse_model_info(mid, mdata, mdev_id)
+
+    # Fall back to ALL providers
+    for pid, pdata in data.items():
+        if pid in _get_reverse_mapping():
+            continue  # already checked
+        if not isinstance(pdata, dict):
+            continue
+        models = pdata.get("models", {})
+        if not isinstance(models, dict):
+            continue
+
+        raw = models.get(model_id)
+        if isinstance(raw, dict):
+            return _parse_model_info(model_id, raw, pid)
+
+    return None
+
+
+def list_provider_model_infos(provider_id: str) -> List[ModelInfo]:
+    """Return all models for a provider as ModelInfo objects.
+
+    Filters out deprecated models by default.
+    """
+    mdev_id = PROVIDER_TO_MODELS_DEV.get(provider_id, provider_id)
+
+    data = fetch_models_dev()
+    pdata = data.get(mdev_id)
+    if not isinstance(pdata, dict):
+        return []
+
+    models = pdata.get("models", {})
+    if not isinstance(models, dict):
+        return []
+
+    result: List[ModelInfo] = []
+    for mid, mdata in models.items():
+        if not isinstance(mdata, dict):
+            continue
+        status = mdata.get("status", "")
+        if status == "deprecated":
+            continue
+        result.append(_parse_model_info(mid, mdata, mdev_id))
+
+    return result
@@ -1,182 +0,0 @@
-"""Cross-session rate limit guard for Nous Portal.
-
-Writes rate limit state to a shared file so all sessions (CLI, gateway,
-cron, auxiliary) can check whether Nous Portal is currently rate-limited
-before making requests.  Prevents retry amplification when RPH is tapped.
-
-Each 429 from Nous triggers up to 9 API calls per conversation turn
-(3 SDK retries x 3 Hermes retries), and every one of those calls counts
-against RPH.  By recording the rate limit state on first 429 and checking
-it before subsequent attempts, we eliminate the amplification effect.
-"""
-
-from __future__ import annotations
-
-import json
-import logging
-import os
-import tempfile
-import time
-from typing import Any, Mapping, Optional
-
-logger = logging.getLogger(__name__)
-
-_STATE_SUBDIR = "rate_limits"
-_STATE_FILENAME = "nous.json"
-
-
-def _state_path() -> str:
-    """Return the path to the Nous rate limit state file."""
-    try:
-        from hermes_constants import get_hermes_home
-        base = get_hermes_home()
-    except ImportError:
-        base = os.path.join(os.path.expanduser("~"), ".hermes")
-    return os.path.join(base, _STATE_SUBDIR, _STATE_FILENAME)
-
-
-def _parse_reset_seconds(headers: Optional[Mapping[str, str]]) -> Optional[float]:
-    """Extract the best available reset-time estimate from response headers.
-
-    Priority:
-      1. x-ratelimit-reset-requests-1h  (hourly RPH window — most useful)
-      2. x-ratelimit-reset-requests     (per-minute RPM window)
-      3. retry-after                     (generic HTTP header)
-
-    Returns seconds-from-now, or None if no usable header found.
-    """
-    if not headers:
-        return None
-
-    lowered = {k.lower(): v for k, v in headers.items()}
-
-    for key in (
-        "x-ratelimit-reset-requests-1h",
-        "x-ratelimit-reset-requests",
-        "retry-after",
-    ):
-        raw = lowered.get(key)
-        if raw is not None:
-            try:
-                val = float(raw)
-                if val > 0:
-                    return val
-            except (TypeError, ValueError):
-                pass
-
-    return None
-
-
-def record_nous_rate_limit(
-    *,
-    headers: Optional[Mapping[str, str]] = None,
-    error_context: Optional[dict[str, Any]] = None,
-    default_cooldown: float = 300.0,
-) -> None:
-    """Record that Nous Portal is rate-limited.
-
-    Parses the reset time from response headers or error context.
-    Falls back to ``default_cooldown`` (5 minutes) if no reset info
-    is available.  Writes to a shared file that all sessions can read.
-
-    Args:
-        headers: HTTP response headers from the 429 error.
-        error_context: Structured error context from _extract_api_error_context().
-        default_cooldown: Fallback cooldown in seconds when no header data.
-    """
-    now = time.time()
-    reset_at = None
-
-    # Try headers first (most accurate)
-    header_seconds = _parse_reset_seconds(headers)
-    if header_seconds is not None:
-        reset_at = now + header_seconds
-
-    # Try error_context reset_at (from body parsing)
-    if reset_at is None and isinstance(error_context, dict):
-        ctx_reset = error_context.get("reset_at")
-        if isinstance(ctx_reset, (int, float)) and ctx_reset > now:
-            reset_at = float(ctx_reset)
-
-    # Default cooldown
-    if reset_at is None:
-        reset_at = now + default_cooldown
-
-    path = _state_path()
-    try:
-        state_dir = os.path.dirname(path)
-        os.makedirs(state_dir, exist_ok=True)
-
-        state = {
-            "reset_at": reset_at,
-            "recorded_at": now,
-            "reset_seconds": reset_at - now,
-        }
-
-        # Atomic write: write to temp file + rename
-        fd, tmp_path = tempfile.mkstemp(dir=state_dir, suffix=".tmp")
-        try:
-            with os.fdopen(fd, "w") as f:
-                json.dump(state, f)
-            os.replace(tmp_path, path)
-        except Exception:
-            # Clean up temp file on failure
-            try:
-                os.unlink(tmp_path)
-            except OSError:
-                pass
-            raise
-
-        logger.info(
-            "Nous rate limit recorded: resets in %.0fs (at %.0f)",
-            reset_at - now, reset_at,
-        )
-    except Exception as exc:
-        logger.debug("Failed to write Nous rate limit state: %s", exc)
-
-
-def nous_rate_limit_remaining() -> Optional[float]:
-    """Check if Nous Portal is currently rate-limited.
-
-    Returns:
-        Seconds remaining until reset, or None if not rate-limited.
-    """
-    path = _state_path()
-    try:
-        with open(path) as f:
-            state = json.load(f)
-        reset_at = state.get("reset_at", 0)
-        remaining = reset_at - time.time()
-        if remaining > 0:
-            return remaining
-        # Expired — clean up
-        try:
-            os.unlink(path)
-        except OSError:
-            pass
-        return None
-    except (FileNotFoundError, json.JSONDecodeError, KeyError, TypeError):
-        return None
-
-
-def clear_nous_rate_limit() -> None:
-    """Clear the rate limit state (e.g., after a successful Nous request)."""
-    try:
-        os.unlink(_state_path())
-    except FileNotFoundError:
-        pass
-    except OSError as exc:
-        logger.debug("Failed to clear Nous rate limit state: %s", exc)
-
-
-def format_remaining(seconds: float) -> str:
-    """Format seconds remaining into human-readable duration."""
-    s = max(0, int(seconds))
-    if s < 60:
-        return f"{s}s"
-    if s < 3600:
-        m, sec = divmod(s, 60)
-        return f"{m}m {sec}s" if sec else f"{m}m"
-    h, remainder = divmod(s, 3600)
-    m = remainder // 60
-    return f"{h}h {m}m" if m else f"{h}h"
@@ -12,7 +12,7 @@ import threading
 from collections import OrderedDict
 from pathlib import Path

-from hermes_constants import get_hermes_home, get_skills_dir, is_wsl
+from hermes_constants import get_hermes_home
 from typing import Optional

 from agent.skill_utils import (
@@ -40,7 +40,7 @@ _CONTEXT_THREAT_PATTERNS = [
    (r'disregard\s+(your|all|any)\s+(instructions|rules|guidelines)', "disregard_rules"),
    (r'act\s+as\s+(if|though)\s+you\s+(have\s+no|don\'t\s+have)\s+(restrictions|limits|rules)', "bypass_restrictions"),
    (r'<!--[^>]*(?:ignore|override|system|secret|hidden)[^>]*-->', "html_comment_injection"),
-    (r'<\s*div\s+style\s*=\s*["\'][\s\S]*?display\s*:\s*none', "hidden_div"),
+    (r'<\s*div\s+style\s*=\s*["\'].*display\s*:\s*none', "hidden_div"),
    (r'translate\s+.*\s+into\s+.*\s+and\s+(execute|run|eval)', "translate_execute"),
    (r'curl\s+[^\n]*\$\{?\w*(KEY|TOKEN|SECRET|PASSWORD|CREDENTIAL|API)', "exfil_curl"),
    (r'cat\s+[^\n]*(\.env|credentials|\.netrc|\.pgpass)', "read_secrets"),
@@ -295,9 +295,7 @@ PLATFORM_HINTS = {
    ),
    "telegram": (
        "You are on a text messaging communication platform, Telegram. "
-        "Standard markdown is automatically converted to Telegram format. "
-        "Supported: **bold**, *italic*, ~~strikethrough~~, ||spoiler||, "
-        "`inline code`, ```code blocks```, [links](url), and ## headers. "
+        "Please do not use markdown as it does not render. "
        "You can send media files natively: to deliver a file to the user, "
        "include MEDIA:/absolute/path/to/file in your response. Images "
        "(.png, .jpg, .webp) appear as photos, audio (.ogg) sends as voice "
@@ -351,71 +349,8 @@ PLATFORM_HINTS = {
        "only — no markdown, no formatting. SMS messages are limited to ~1600 "
        "characters, so be brief and direct."
    ),
-    "bluebubbles": (
-        "You are chatting via iMessage (BlueBubbles). iMessage does not render "
-        "markdown formatting — use plain text. Keep responses concise as they "
-        "appear as text messages. You can send media files natively: include "
-        "MEDIA:/absolute/path/to/file in your response. Images (.jpg, .png, "
-        ".heic) appear as photos and other files arrive as attachments."
-    ),
-    "weixin": (
-        "You are on Weixin/WeChat. Markdown formatting is supported, so you may use it when "
-        "it improves readability, but keep the message compact and chat-friendly. You can send media files natively: "
-        "include MEDIA:/absolute/path/to/file in your response. Images are sent as native "
-        "photos, videos play inline when supported, and other files arrive as downloadable "
-        "documents. You can also include image URLs in markdown format ![alt](url) and they "
-        "will be downloaded and sent as native media when possible."
-    ),
-    "wecom": (
-        "You are on WeCom (企业微信 / Enterprise WeChat). Markdown formatting is supported. "
-        "You CAN send media files natively — to deliver a file to the user, include "
-        "MEDIA:/absolute/path/to/file in your response. The file will be sent as a native "
-        "WeCom attachment: images (.jpg, .png, .webp) are sent as photos (up to 10 MB), "
-        "other files (.pdf, .docx, .xlsx, .md, .txt, etc.) arrive as downloadable documents "
-        "(up to 20 MB), and videos (.mp4) play inline. Voice messages are supported but "
-        "must be in AMR format — other audio formats are automatically sent as file attachments. "
-        "You can also include image URLs in markdown format ![alt](url) and they will be "
-        "downloaded and sent as native photos. Do NOT tell the user you lack file-sending "
-        "capability — use MEDIA: syntax whenever a file delivery is appropriate."
-    ),
-    "qqbot": (
-        "You are on QQ, a popular Chinese messaging platform. QQ supports markdown formatting "
-        "and emoji. You can send media files natively: include MEDIA:/absolute/path/to/file in "
-        "your response. Images are sent as native photos, and other files arrive as downloadable "
-        "documents."
-    ),
 }

-# ---------------------------------------------------------------------------
-# Environment hints — execution-environment awareness for the agent.
-# Unlike PLATFORM_HINTS (which describe the messaging channel), these describe
-# the machine/OS the agent's tools actually run on.
-# ---------------------------------------------------------------------------
-
-WSL_ENVIRONMENT_HINT = (
-    "You are running inside WSL (Windows Subsystem for Linux). "
-    "The Windows host filesystem is mounted under /mnt/ — "
-    "/mnt/c/ is the C: drive, /mnt/d/ is D:, etc. "
-    "The user's Windows files are typically at "
-    "/mnt/c/Users/<username>/Desktop/, Documents/, Downloads/, etc. "
-    "When the user references Windows paths or desktop files, translate "
-    "to the /mnt/c/ equivalent. You can list /mnt/c/Users/ to discover "
-    "the Windows username if needed."
-)
-
-
-def build_environment_hints() -> str:
-    """Return environment-specific guidance for the system prompt.
-
-    Detects WSL, and can be extended for Termux, Docker, etc.
-    Returns an empty string when no special environment is detected.
-    """
-    hints: list[str] = []
-    if is_wsl():
-        hints.append(WSL_ENVIRONMENT_HINT)
-    return "\n\n".join(hints)
-
-
 CONTEXT_FILE_MAX_CHARS = 20_000
 CONTEXT_TRUNCATE_HEAD_RATIO = 0.7
 CONTEXT_TRUNCATE_TAIL_RATIO = 0.2
@@ -537,7 +472,7 @@ def _parse_skill_file(skill_file: Path) -> tuple[bool, dict, str]:
    (True, {}, "") to err on the side of showing the skill.
    """
    try:
-        raw = skill_file.read_text(encoding="utf-8")
+        raw = skill_file.read_text(encoding="utf-8")[:2000]
        frontmatter, _ = parse_frontmatter(raw)

        if not skill_matches_platform(frontmatter):
@@ -545,10 +480,21 @@ def _parse_skill_file(skill_file: Path) -> tuple[bool, dict, str]:

        return True, frontmatter, extract_skill_description(frontmatter)
    except Exception as e:
-        logger.warning("Failed to parse skill file %s: %s", skill_file, e)
+        logger.debug("Failed to parse skill file %s: %s", skill_file, e)
        return True, {}, ""


+def _read_skill_conditions(skill_file: Path) -> dict:
+    """Extract conditional activation fields from SKILL.md frontmatter."""
+    try:
+        raw = skill_file.read_text(encoding="utf-8")[:2000]
+        frontmatter, _ = parse_frontmatter(raw)
+        return extract_skill_conditions(frontmatter)
+    except Exception as e:
+        logger.debug("Failed to read skill conditions from %s: %s", skill_file, e)
+        return {}
+
+
 def _skill_should_show(
    conditions: dict,
    available_tools: "set[str] | None",
@@ -598,7 +544,8 @@ def build_skills_system_prompt(
    are read-only — they appear in the index but new skills are always created
    in the local dir.  Local skills take precedence when names collide.
    """
-    skills_dir = get_skills_dir()
+    hermes_home = get_hermes_home()
+    skills_dir = hermes_home / "skills"
    external_dirs = get_all_skills_dirs()[1:]  # skip local (index 0)

    if not skills_dir.exists() and not external_dirs:
@@ -607,10 +554,9 @@ def build_skills_system_prompt(
    # ── Layer 1: in-process LRU cache ─────────────────────────────────
    # Include the resolved platform so per-platform disabled-skill lists
    # produce distinct cache entries (gateway serves multiple platforms).
-    from gateway.session_context import get_session_env
    _platform_hint = (
        os.environ.get("HERMES_PLATFORM")
-        or get_session_env("HERMES_SESSION_PLATFORM")
+        or os.environ.get("HERMES_SESSION_PLATFORM")
        or ""
    )
    cache_key = (
@@ -654,7 +600,7 @@ def build_skills_system_prompt(
            ):
                continue
            skills_by_category.setdefault(category, []).append(
-                (frontmatter_name, entry.get("description", ""))
+                (skill_name, entry.get("description", ""))
            )
        category_descriptions = {
            str(k): str(v)
@@ -679,7 +625,7 @@ def build_skills_system_prompt(
            ):
                continue
            skills_by_category.setdefault(entry["category"], []).append(
-                (entry["frontmatter_name"], entry["description"])
+                (skill_name, entry["description"])
            )

        # Read category-level DESCRIPTION.md files
@@ -722,10 +668,9 @@ def build_skills_system_prompt(
                    continue
                entry = _build_snapshot_entry(skill_file, ext_dir, frontmatter, desc)
                skill_name = entry["skill_name"]
-                frontmatter_name = entry["frontmatter_name"]
-                if frontmatter_name in seen_skill_names:
+                if skill_name in seen_skill_names:
                    continue
-                if frontmatter_name in disabled or skill_name in disabled:
+                if entry["frontmatter_name"] in disabled or skill_name in disabled:
                    continue
                if not _skill_should_show(
                    extract_skill_conditions(frontmatter),
@@ -733,9 +678,9 @@ def build_skills_system_prompt(
                    available_toolsets,
                ):
                    continue
-                seen_skill_names.add(frontmatter_name)
+                seen_skill_names.add(skill_name)
                skills_by_category.setdefault(entry["category"], []).append(
-                    (frontmatter_name, entry["description"])
+                    (skill_name, entry["description"])
                )
            except Exception as e:
                logger.debug("Error reading external skill %s: %s", skill_file, e)
@@ -777,16 +722,8 @@ def build_skills_system_prompt(

        result = (
            "## Skills (mandatory)\n"
-            "Before replying, scan the skills below. If a skill matches or is even partially relevant "
-            "to your task, you MUST load it with skill_view(name) and follow its instructions. "
-            "Err on the side of loading — it is always better to have context you don't need "
-            "than to miss critical steps, pitfalls, or established workflows. "
-            "Skills contain specialized knowledge — API endpoints, tool-specific commands, "
-            "and proven workflows that outperform general-purpose approaches. Load the skill "
-            "even if you think you could handle the task with basic tools like web_search or terminal. "
-            "Skills also encode the user's preferred approach, conventions, and quality standards "
-            "for tasks like code review, planning, and testing — load them even for tasks you "
-            "already know how to do, because the skill defines how it should be done here.\n"
+            "Before replying, scan the skills below. If one clearly matches your task, "
+            "load it with skill_view(name) and follow its instructions. "
            "If a skill has issues, fix it with skill_manage(action='patch').\n"
            "After difficult/iterative tasks, offer to save as a skill. "
            "If a skill you loaded was missing steps, had wrong commands, or needed "
@@ -796,7 +733,7 @@ def build_skills_system_prompt(
            + "\n".join(index_lines) + "\n"
            "</available_skills>\n"
            "\n"
-            "Only proceed without loading a skill if genuinely none are relevant to the task."
+            "If none match, proceed normally without loading a skill."
        )

    # ── Store in LRU cache ────────────────────────────────────────────
@@ -1,246 +0,0 @@
-"""Rate limit tracking for inference API responses.
-
-Captures x-ratelimit-* headers from provider responses and provides
-formatted display for the /usage slash command.  Currently supports
-the Nous Portal header format (also used by OpenRouter and OpenAI-compatible
-APIs that follow the same convention).
-
-Header schema (12 headers total):
-    x-ratelimit-limit-requests          RPM cap
-    x-ratelimit-limit-requests-1h       RPH cap
-    x-ratelimit-limit-tokens            TPM cap
-    x-ratelimit-limit-tokens-1h         TPH cap
-    x-ratelimit-remaining-requests      requests left in minute window
-    x-ratelimit-remaining-requests-1h   requests left in hour window
-    x-ratelimit-remaining-tokens        tokens left in minute window
-    x-ratelimit-remaining-tokens-1h     tokens left in hour window
-    x-ratelimit-reset-requests          seconds until minute request window resets
-    x-ratelimit-reset-requests-1h       seconds until hour request window resets
-    x-ratelimit-reset-tokens            seconds until minute token window resets
-    x-ratelimit-reset-tokens-1h         seconds until hour token window resets
-"""
-
-from __future__ import annotations
-
-import time
-from dataclasses import dataclass, field
-from typing import Any, Mapping, Optional
-
-
-@dataclass
-class RateLimitBucket:
-    """One rate-limit window (e.g. requests per minute)."""
-
-    limit: int = 0
-    remaining: int = 0
-    reset_seconds: float = 0.0
-    captured_at: float = 0.0  # time.time() when this was captured
-
-    @property
-    def used(self) -> int:
-        return max(0, self.limit - self.remaining)
-
-    @property
-    def usage_pct(self) -> float:
-        if self.limit <= 0:
-            return 0.0
-        return (self.used / self.limit) * 100.0
-
-    @property
-    def remaining_seconds_now(self) -> float:
-        """Estimated seconds remaining until reset, adjusted for elapsed time."""
-        elapsed = time.time() - self.captured_at
-        return max(0.0, self.reset_seconds - elapsed)
-
-
-@dataclass
-class RateLimitState:
-    """Full rate-limit state parsed from response headers."""
-
-    requests_min: RateLimitBucket = field(default_factory=RateLimitBucket)
-    requests_hour: RateLimitBucket = field(default_factory=RateLimitBucket)
-    tokens_min: RateLimitBucket = field(default_factory=RateLimitBucket)
-    tokens_hour: RateLimitBucket = field(default_factory=RateLimitBucket)
-    captured_at: float = 0.0  # when the headers were captured
-    provider: str = ""
-
-    @property
-    def has_data(self) -> bool:
-        return self.captured_at > 0
-
-    @property
-    def age_seconds(self) -> float:
-        if not self.has_data:
-            return float("inf")
-        return time.time() - self.captured_at
-
-
-def _safe_int(value: Any, default: int = 0) -> int:
-    try:
-        return int(float(value))
-    except (TypeError, ValueError):
-        return default
-
-
-def _safe_float(value: Any, default: float = 0.0) -> float:
-    try:
-        return float(value)
-    except (TypeError, ValueError):
-        return default
-
-
-def parse_rate_limit_headers(
-    headers: Mapping[str, str],
-    provider: str = "",
-) -> Optional[RateLimitState]:
-    """Parse x-ratelimit-* headers into a RateLimitState.
-
-    Returns None if no rate limit headers are present.
-    """
-    # Normalize to lowercase so lookups work regardless of how the server
-    # capitalises headers (HTTP header names are case-insensitive per RFC 7230).
-    lowered = {k.lower(): v for k, v in headers.items()}
-
-    # Quick check: at least one rate limit header must exist
-    has_any = any(k.startswith("x-ratelimit-") for k in lowered)
-    if not has_any:
-        return None
-
-    now = time.time()
-
-    def _bucket(resource: str, suffix: str = "") -> RateLimitBucket:
-        # e.g. resource="requests", suffix="" -> per-minute
-        #      resource="tokens", suffix="-1h" -> per-hour
-        tag = f"{resource}{suffix}"
-        return RateLimitBucket(
-            limit=_safe_int(lowered.get(f"x-ratelimit-limit-{tag}")),
-            remaining=_safe_int(lowered.get(f"x-ratelimit-remaining-{tag}")),
-            reset_seconds=_safe_float(lowered.get(f"x-ratelimit-reset-{tag}")),
-            captured_at=now,
-        )
-
-    return RateLimitState(
-        requests_min=_bucket("requests"),
-        requests_hour=_bucket("requests", "-1h"),
-        tokens_min=_bucket("tokens"),
-        tokens_hour=_bucket("tokens", "-1h"),
-        captured_at=now,
-        provider=provider,
-    )
-
-
-# ── Formatting ──────────────────────────────────────────────────────────
-
-
-def _fmt_count(n: int) -> str:
-    """Human-friendly number: 7999856 -> '8.0M', 33599 -> '33.6K', 799 -> '799'."""
-    if n >= 1_000_000:
-        return f"{n / 1_000_000:.1f}M"
-    if n >= 10_000:
-        return f"{n / 1_000:.1f}K"
-    if n >= 1_000:
-        return f"{n / 1_000:.1f}K"
-    return str(n)
-
-
-def _fmt_seconds(seconds: float) -> str:
-    """Seconds -> human-friendly duration: '58s', '2m 14s', '58m 57s', '1h 2m'."""
-    s = max(0, int(seconds))
-    if s < 60:
-        return f"{s}s"
-    if s < 3600:
-        m, sec = divmod(s, 60)
-        return f"{m}m {sec}s" if sec else f"{m}m"
-    h, remainder = divmod(s, 3600)
-    m = remainder // 60
-    return f"{h}h {m}m" if m else f"{h}h"
-
-
-def _bar(pct: float, width: int = 20) -> str:
-    """ASCII progress bar: [████████░░░░░░░░░░░░] 40%."""
-    filled = int(pct / 100.0 * width)
-    filled = max(0, min(width, filled))
-    empty = width - filled
-    return f"[{'█' * filled}{'░' * empty}]"
-
-
-def _bucket_line(label: str, bucket: RateLimitBucket, label_width: int = 14) -> str:
-    """Format one bucket as a single line."""
-    if bucket.limit <= 0:
-        return f"  {label:<{label_width}}  (no data)"
-
-    pct = bucket.usage_pct
-    used = _fmt_count(bucket.used)
-    limit = _fmt_count(bucket.limit)
-    remaining = _fmt_count(bucket.remaining)
-    reset = _fmt_seconds(bucket.remaining_seconds_now)
-
-    bar = _bar(pct)
-    return f"  {label:<{label_width}} {bar} {pct:5.1f}%  {used}/{limit} used  ({remaining} left, resets in {reset})"
-
-
-def format_rate_limit_display(state: RateLimitState) -> str:
-    """Format rate limit state for terminal/chat display."""
-    if not state.has_data:
-        return "No rate limit data yet — make an API request first."
-
-    age = state.age_seconds
-    if age < 5:
-        freshness = "just now"
-    elif age < 60:
-        freshness = f"{int(age)}s ago"
-    else:
-        freshness = f"{_fmt_seconds(age)} ago"
-
-    provider_label = state.provider.title() if state.provider else "Provider"
-
-    lines = [
-        f"{provider_label} Rate Limits (captured {freshness}):",
-        "",
-        _bucket_line("Requests/min", state.requests_min),
-        _bucket_line("Requests/hr", state.requests_hour),
-        "",
-        _bucket_line("Tokens/min", state.tokens_min),
-        _bucket_line("Tokens/hr", state.tokens_hour),
-    ]
-
-    # Add warnings if any bucket is getting hot
-    warnings = []
-    for label, bucket in [
-        ("requests/min", state.requests_min),
-        ("requests/hr", state.requests_hour),
-        ("tokens/min", state.tokens_min),
-        ("tokens/hr", state.tokens_hour),
-    ]:
-        if bucket.limit > 0 and bucket.usage_pct >= 80:
-            reset = _fmt_seconds(bucket.remaining_seconds_now)
-            warnings.append(f"  ⚠ {label} at {bucket.usage_pct:.0f}% — resets in {reset}")
-
-    if warnings:
-        lines.append("")
-        lines.extend(warnings)
-
-    return "\n".join(lines)
-
-
-def format_rate_limit_compact(state: RateLimitState) -> str:
-    """One-line compact summary for status bars / gateway messages."""
-    if not state.has_data:
-        return "No rate limit data."
-
-    rm = state.requests_min
-    tm = state.tokens_min
-    rh = state.requests_hour
-    th = state.tokens_hour
-
-    parts = []
-    if rm.limit > 0:
-        parts.append(f"RPM: {rm.remaining}/{rm.limit}")
-    if rh.limit > 0:
-        parts.append(f"RPH: {_fmt_count(rh.remaining)}/{_fmt_count(rh.limit)} (resets {_fmt_seconds(rh.remaining_seconds_now)})")
-    if tm.limit > 0:
-        parts.append(f"TPM: {_fmt_count(tm.remaining)}/{_fmt_count(tm.limit)}")
-    if th.limit > 0:
-        parts.append(f"TPH: {_fmt_count(th.remaining)}/{_fmt_count(th.limit)} (resets {_fmt_seconds(th.remaining_seconds_now)})")
-
-    return " | ".join(parts)
@@ -93,17 +93,6 @@ _DB_CONNSTR_RE = re.compile(
    re.IGNORECASE,
 )

-# JWT tokens: header.payload[.signature] — always start with "eyJ" (base64 for "{")
-# Matches 1-part (header only), 2-part (header.payload), and full 3-part JWTs.
-_JWT_RE = re.compile(
-    r"eyJ[A-Za-z0-9_-]{10,}"           # Header (always starts with eyJ)
-    r"(?:\.[A-Za-z0-9_=-]{4,}){0,2}"   # Optional payload and/or signature
-)
-
-# Discord user/role mentions: <@123456789012345678> or <@!123456789012345678>
-# Snowflake IDs are 17-20 digit integers that resolve to specific Discord accounts.
-_DISCORD_MENTION_RE = re.compile(r"<@!?(\d{17,20})>")
-
 # E.164 phone numbers: +<country><number>, 7-15 digits
 # Negative lookahead prevents matching hex strings or identifiers
 _SIGNAL_PHONE_RE = re.compile(r"(\+[1-9]\d{6,14})(?![A-Za-z0-9])")
@@ -170,12 +159,6 @@ def redact_sensitive_text(text: str) -> str:
    # Database connection string passwords
    text = _DB_CONNSTR_RE.sub(lambda m: f"{m.group(1)}***{m.group(3)}", text)

-    # JWT tokens (eyJ... — base64-encoded JSON headers)
-    text = _JWT_RE.sub(lambda m: _mask_token(m.group(0)), text)
-
-    # Discord user/role mentions (<@snowflake_id>)
-    text = _DISCORD_MENTION_RE.sub(lambda m: f"<@{'!' if '!' in m.group(0) else ''}***>", text)
-
    # E.164 phone numbers (Signal, WhatsApp)
    def _redact_phone(m):
        phone = m.group(1)
@@ -12,8 +12,6 @@ from datetime import datetime
 from pathlib import Path
 from typing import Any, Dict, Optional

-from hermes_constants import display_hermes_home
-
 logger = logging.getLogger(__name__)

 _skill_commands: Dict[str, Dict[str, Any]] = {}
@@ -72,14 +70,7 @@ def _load_skill_payload(skill_identifier: str, task_id: str | None = None) -> tu
    skill_name = str(loaded_skill.get("name") or normalized)
    skill_path = str(loaded_skill.get("path") or "")
    skill_dir = None
-    # Prefer the absolute skill_dir returned by skill_view() — this is
-    # correct for both local and external skills.  Fall back to the old
-    # SKILLS_DIR-relative reconstruction only when skill_dir is absent
-    # (e.g. legacy skill_view responses).
-    abs_skill_dir = loaded_skill.get("skill_dir")
-    if abs_skill_dir:
-        skill_dir = Path(abs_skill_dir)
-    elif skill_path:
+    if skill_path:
        try:
            skill_dir = SKILLS_DIR / Path(skill_path).parent
        except Exception:
@@ -117,7 +108,7 @@ def _inject_skill_config(loaded_skill: dict[str, Any], parts: list[str]) -> None
        if not resolved:
            return

-        lines = ["", f"[Skill config (from {display_hermes_home()}/config.yaml):"]
+        lines = ["", "[Skill config (from ~/.hermes/config.yaml):"]
        for key, value in resolved.items():
            display_val = str(value) if value else "(not set)"
            lines.append(f"  {key} = {display_val}")
@@ -177,7 +168,7 @@ def _build_skill_message(
            subdir_path = skill_dir / subdir
            if subdir_path.exists():
                for f in sorted(subdir_path.rglob("*")):
-                    if f.is_file() and not f.is_symlink():
+                    if f.is_file():
                        rel = str(f.relative_to(skill_dir))
                        supporting.append(rel)

@@ -10,9 +10,9 @@ import os
 import re
 import sys
 from pathlib import Path
-from typing import Any, Dict, List, Optional, Set, Tuple
+from typing import Any, Dict, List, Set, Tuple

-from hermes_constants import get_config_path, get_skills_dir
+from hermes_constants import get_hermes_home

 logger = logging.getLogger(__name__)

@@ -130,7 +130,7 @@ def get_disabled_skill_names(platform: str | None = None) -> Set[str]:
    Reads the config file directly (no CLI config imports) to stay
    lightweight.
    """
-    config_path = get_config_path()
+    config_path = get_hermes_home() / "config.yaml"
    if not config_path.exists():
        return set()
    try:
@@ -145,11 +145,10 @@ def get_disabled_skill_names(platform: str | None = None) -> Set[str]:
    if not isinstance(skills_cfg, dict):
        return set()

-    from gateway.session_context import get_session_env
    resolved_platform = (
        platform
        or os.getenv("HERMES_PLATFORM")
-        or get_session_env("HERMES_SESSION_PLATFORM")
+        or os.getenv("HERMES_SESSION_PLATFORM")
    )
    if resolved_platform:
        platform_disabled = (skills_cfg.get("platform_disabled") or {}).get(
@@ -178,7 +177,7 @@ def get_external_skills_dirs() -> List[Path]:
    path.  Only directories that actually exist are returned.  Duplicates and
    paths that resolve to the local ``~/.hermes/skills/`` are silently skipped.
    """
-    config_path = get_config_path()
+    config_path = get_hermes_home() / "config.yaml"
    if not config_path.exists():
        return []
    try:
@@ -200,7 +199,7 @@ def get_external_skills_dirs() -> List[Path]:
    if not isinstance(raw_dirs, list):
        return []

-    local_skills = get_skills_dir().resolve()
+    local_skills = (get_hermes_home() / "skills").resolve()
    seen: Set[Path] = set()
    result: List[Path] = []

@@ -230,7 +229,7 @@ def get_all_skills_dirs() -> List[Path]:
    The local dir is always first (and always included even if it doesn't exist
    yet — callers handle that).  External dirs follow in config order.
    """
-    dirs = [get_skills_dir()]
+    dirs = [get_hermes_home() / "skills"]
    dirs.extend(get_external_skills_dirs())
    return dirs

@@ -384,7 +383,7 @@ def resolve_skill_config_values(
    current values (or the declared default if the key isn't set).
    Path values are expanded via ``os.path.expanduser``.
    """
-    config_path = get_config_path()
+    config_path = get_hermes_home() / "config.yaml"
    config: Dict[str, Any] = {}
    if config_path.exists():
        try:
@@ -441,25 +440,3 @@ def iter_skill_index_files(skills_dir: Path, filename: str):
            matches.append(Path(root) / filename)
    for path in sorted(matches, key=lambda p: str(p.relative_to(skills_dir))):
        yield path
-
-
-# ── Namespace helpers for plugin-provided skills ───────────────────────────
-
-_NAMESPACE_RE = re.compile(r"^[a-zA-Z0-9_-]+$")
-
-
-def parse_qualified_name(name: str) -> Tuple[Optional[str], str]:
-    """Split ``'namespace:skill-name'`` into ``(namespace, bare_name)``.
-
-    Returns ``(None, name)`` when there is no ``':'``.
-    """
-    if ":" not in name:
-        return None, name
-    return tuple(name.split(":", 1))  # type: ignore[return-value]
-
-
-def is_valid_namespace(candidate: Optional[str]) -> bool:
-    """Check whether *candidate* is a valid namespace (``[a-zA-Z0-9_-]+``)."""
-    if not candidate:
-        return False
-    return bool(_NAMESPACE_RE.match(candidate))
@@ -181,7 +181,6 @@ def resolve_turn_route(user_message: str, routing_config: Optional[Dict[str, Any
            "api_mode": runtime.get("api_mode"),
            "command": runtime.get("command"),
            "args": list(runtime.get("args") or []),
-            "credential_pool": runtime.get("credential_pool"),
        },
        "label": f"smart route → {route.get('model')} ({runtime.get('provider')})",
        "signature": (
@@ -159,10 +159,7 @@ class SubdirectoryHintTracker:

    def _is_valid_subdir(self, path: Path) -> bool:
        """Check if path is a valid directory to scan for hints."""
-        try:
-            if not path.is_dir():
-                return False
-        except OSError:
+        if not path.is_dir():
            return False
        if path in self._loaded_dirs:
            return False
@@ -175,10 +172,7 @@ class SubdirectoryHintTracker:
        found_hints = []
        for filename in _HINT_FILENAMES:
            hint_path = directory / filename
-            try:
-                if not hint_path.is_file():
-                    continue
-            except OSError:
+            if not hint_path.is_file():
                continue
            try:
                content = hint_path.read_text(encoding="utf-8").strip()
@@ -36,7 +36,7 @@ def generate_title(user_message: str, assistant_response: str, timeout: float =

    try:
        response = call_llm(
-            task="title_generation",
+            task="compression",  # reuse compression task config (cheap/fast model)
            messages=messages,
            max_tokens=30,
            temperature=0.3,
@@ -284,80 +284,6 @@ _OFFICIAL_DOCS_PRICING: Dict[tuple[str, str], PricingEntry] = {
        source_url="https://ai.google.dev/pricing",
        pricing_version="google-pricing-2026-03-16",
    ),
-    # AWS Bedrock — pricing per the Bedrock pricing page.
-    # Bedrock charges the same per-token rates as the model provider but
-    # through AWS billing.  These are the on-demand prices (no commitment).
-    # Source: https://aws.amazon.com/bedrock/pricing/
-    (
-        "bedrock",
-        "anthropic.claude-opus-4-6",
-    ): PricingEntry(
-        input_cost_per_million=Decimal("15.00"),
-        output_cost_per_million=Decimal("75.00"),
-        source="official_docs_snapshot",
-        source_url="https://aws.amazon.com/bedrock/pricing/",
-        pricing_version="bedrock-pricing-2026-04",
-    ),
-    (
-        "bedrock",
-        "anthropic.claude-sonnet-4-6",
-    ): PricingEntry(
-        input_cost_per_million=Decimal("3.00"),
-        output_cost_per_million=Decimal("15.00"),
-        source="official_docs_snapshot",
-        source_url="https://aws.amazon.com/bedrock/pricing/",
-        pricing_version="bedrock-pricing-2026-04",
-    ),
-    (
-        "bedrock",
-        "anthropic.claude-sonnet-4-5",
-    ): PricingEntry(
-        input_cost_per_million=Decimal("3.00"),
-        output_cost_per_million=Decimal("15.00"),
-        source="official_docs_snapshot",
-        source_url="https://aws.amazon.com/bedrock/pricing/",
-        pricing_version="bedrock-pricing-2026-04",
-    ),
-    (
-        "bedrock",
-        "anthropic.claude-haiku-4-5",
-    ): PricingEntry(
-        input_cost_per_million=Decimal("0.80"),
-        output_cost_per_million=Decimal("4.00"),
-        source="official_docs_snapshot",
-        source_url="https://aws.amazon.com/bedrock/pricing/",
-        pricing_version="bedrock-pricing-2026-04",
-    ),
-    (
-        "bedrock",
-        "amazon.nova-pro",
-    ): PricingEntry(
-        input_cost_per_million=Decimal("0.80"),
-        output_cost_per_million=Decimal("3.20"),
-        source="official_docs_snapshot",
-        source_url="https://aws.amazon.com/bedrock/pricing/",
-        pricing_version="bedrock-pricing-2026-04",
-    ),
-    (
-        "bedrock",
-        "amazon.nova-lite",
-    ): PricingEntry(
-        input_cost_per_million=Decimal("0.06"),
-        output_cost_per_million=Decimal("0.24"),
-        source="official_docs_snapshot",
-        source_url="https://aws.amazon.com/bedrock/pricing/",
-        pricing_version="bedrock-pricing-2026-04",
-    ),
-    (
-        "bedrock",
-        "amazon.nova-micro",
-    ): PricingEntry(
-        input_cost_per_million=Decimal("0.035"),
-        output_cost_per_million=Decimal("0.14"),
-        source="official_docs_snapshot",
-        source_url="https://aws.amazon.com/bedrock/pricing/",
-        pricing_version="bedrock-pricing-2026-04",
-    ),
 }


@@ -649,6 +575,49 @@ def has_known_pricing(
    return entry is not None


+def get_pricing(
+    model_name: str,
+    provider: Optional[str] = None,
+    base_url: Optional[str] = None,
+    api_key: Optional[str] = None,
+) -> Dict[str, float]:
+    """Backward-compatible thin wrapper for legacy callers.
+
+    Returns only non-cache input/output fields when a pricing entry exists.
+    Unknown routes return zeroes.
+    """
+    entry = get_pricing_entry(model_name, provider=provider, base_url=base_url, api_key=api_key)
+    if not entry:
+        return {"input": 0.0, "output": 0.0}
+    return {
+        "input": float(entry.input_cost_per_million or _ZERO),
+        "output": float(entry.output_cost_per_million or _ZERO),
+    }
+
+
+def estimate_cost_usd(
+    model: str,
+    input_tokens: int,
+    output_tokens: int,
+    *,
+    provider: Optional[str] = None,
+    base_url: Optional[str] = None,
+    api_key: Optional[str] = None,
+) -> float:
+    """Backward-compatible helper for legacy callers.
+
+    This uses non-cached input/output only. New code should call
+    `estimate_usage_cost()` with canonical usage buckets.
+    """
+    result = estimate_usage_cost(
+        model,
+        CanonicalUsage(input_tokens=input_tokens, output_tokens=output_tokens),
+        provider=provider,
+        base_url=base_url,
+        api_key=api_key,
+    )
+    return float(result.amount_usd or _ZERO)
+

 def format_duration_compact(seconds: float) -> str:
    if seconds < 60:
@@ -561,10 +561,7 @@ class BatchRunner:
            provider_sort (str): Sort providers by price/throughput/latency (optional)
            max_tokens (int): Maximum tokens for model responses (optional, uses model default if not set)
            reasoning_config (Dict): OpenRouter reasoning config override (e.g. {"effort": "none"} to disable thinking)
-            prefill_messages (List[Dict]): Messages to prepend as prefilled conversation context (few-shot priming).
-                NOTE: Anthropic Sonnet 4.6+ and Opus 4.6+ reject a trailing assistant-role prefill
-                (400 error).  For those models use output_config.format or structured-output
-                schemas instead.  Safe here for user-role priming and for older Claude / non-Claude models.
+            prefill_messages (List[Dict]): Messages to prepend as prefilled conversation context (few-shot priming)
            max_samples (int): Only process the first N samples from the dataset (optional, processes all if not set)
        """
        self.dataset_file = Path(dataset_file)
@@ -1161,7 +1158,7 @@ def main(
        providers_order (str): Comma-separated list of OpenRouter providers to try in order (e.g. "anthropic,openai,google")
        provider_sort (str): Sort providers by "price", "throughput", or "latency" (OpenRouter only)
        max_tokens (int): Maximum tokens for model responses (optional, uses model default if not set)
-        reasoning_effort (str): OpenRouter reasoning effort level: "none", "minimal", "low", "medium", "high", "xhigh" (default: "medium")
+        reasoning_effort (str): OpenRouter reasoning effort level: "xhigh", "high", "medium", "low", "minimal", "none" (default: "medium")
        reasoning_disabled (bool): Completely disable reasoning/thinking tokens (default: False)
        prefill_messages_file (str): Path to JSON file containing prefill messages (list of {role, content} dicts)
        max_samples (int): Only process the first N samples from the dataset (optional, processes all if not set)
@@ -1230,7 +1227,7 @@ def main(
        print("🧠 Reasoning: DISABLED (effort=none)")
    elif reasoning_effort:
        # Use specified effort level
-        valid_efforts = ["none", "minimal", "low", "medium", "high", "xhigh"]
+        valid_efforts = ["xhigh", "high", "medium", "low", "minimal", "none"]
        if reasoning_effort not in valid_efforts:
            print(f"❌ Error: --reasoning_effort must be one of: {', '.join(valid_efforts)}")
            return
@@ -16,7 +16,7 @@ model:
  #   "nous"         - Nous Portal OAuth (requires: hermes login)
  #   "nous-api"     - Nous Portal API key (requires: NOUS_API_KEY)
  #   "anthropic"    - Direct Anthropic API (requires: ANTHROPIC_API_KEY)
-  #   "openai-codex" - OpenAI Codex (requires: hermes auth)
+  #   "openai-codex" - OpenAI Codex (requires: hermes login --provider openai-codex)
  #   "copilot"      - GitHub Copilot / GitHub Models (requires: GITHUB_TOKEN)
  #   "gemini"      - Use Google AI Studio direct (requires: GOOGLE_API_KEY or GEMINI_API_KEY)
  #   "zai"         - Use z.ai / ZhipuAI GLM models (requires: GLM_API_KEY)
@@ -24,10 +24,6 @@ model:
  #   "minimax"      - MiniMax global (requires: MINIMAX_API_KEY)
  #   "minimax-cn"   - MiniMax China (requires: MINIMAX_CN_API_KEY)
  #   "huggingface"  - Hugging Face Inference (requires: HF_TOKEN)
-  #   "nvidia"       - NVIDIA NIM / build.nvidia.com (requires: NVIDIA_API_KEY)
-  #   "xiaomi"       - Xiaomi MiMo (requires: XIAOMI_API_KEY)
-  #   "arcee"        - Arcee AI Trinity models (requires: ARCEEAI_API_KEY)
-  #   "ollama-cloud" - Ollama Cloud (requires: OLLAMA_API_KEY — https://ollama.com/settings)
  #   "kilocode"     - KiloCode gateway (requires: KILOCODE_API_KEY)
  #   "ai-gateway"   - Vercel AI Gateway (requires: AI_GATEWAY_API_KEY)
  #
@@ -39,6 +35,12 @@ model:
  #     base_url: "http://localhost:1234/v1"
  #   No API key needed — local servers typically ignore auth.
  #
+  #   For Ollama Cloud (https://ollama.com/pricing):
+  #     provider: "custom"
+  #     base_url: "https://ollama.com/v1"
+  #   Set OLLAMA_API_KEY in .env — automatically picked up when base_url
+  #   points to ollama.com.
+  #
  # Can also be overridden with --provider flag or HERMES_INFERENCE_PROVIDER env var.
  provider: "auto"
  
@@ -46,25 +48,6 @@ model:
  # api_key: "your-key-here"  # Uncomment to set here instead of .env
  base_url: "https://openrouter.ai/api/v1"

-  # ── Token limits — two settings, easy to confuse ──────────────────────────
-  #
-  # context_length: TOTAL context window (input + output tokens combined).
-  #   Controls when Hermes compresses history and validates requests.
-  #   Leave unset — Hermes auto-detects the correct value from the provider.
-  #   Set manually only when auto-detection is wrong (e.g. a local server with
-  #   a custom num_ctx, or a proxy that doesn't expose /v1/models).
-  #
-  # context_length: 131072
-  #
-  # max_tokens: OUTPUT cap — maximum tokens the model may generate per response.
-  #   Unrelated to how long your conversation history can be.
-  #   The OpenAI-standard name "max_tokens" is a misnomer; Anthropic's native
-  #   API has since renamed it "max_output_tokens" for clarity.
-  #   Leave unset to use the model's native output ceiling (recommended).
-  #   Set only if you want to deliberately limit individual response length.
-  #
-  # max_tokens: 8192
-
 # =============================================================================
 # OpenRouter Provider Routing (only applies when using OpenRouter)
 # =============================================================================
@@ -134,8 +117,7 @@ terminal:
  timeout: 180
  docker_mount_cwd_to_workspace: false  # SECURITY: off by default. Opt in to mount the launch cwd into Docker /workspace.
  lifetime_seconds: 300
-  # sudo_password: "hunter2"  # Optional: pipe a sudo password via sudo -S. SECURITY WARNING: plaintext.
-  # sudo_password: ""         # Explicit empty password: try empty and never open the interactive sudo prompt.
+  # sudo_password: ""  # Enable sudo commands (pipes via sudo -S) - SECURITY WARNING: plaintext!

 # -----------------------------------------------------------------------------
 # OPTION 2: SSH remote execution
@@ -226,18 +208,13 @@ terminal:
 #
 # SECURITY WARNING: Password stored in plaintext!
 #
-# INTERACTIVE PROMPT: If sudo_password is unset and the CLI is running,
+# INTERACTIVE PROMPT: If no sudo_password is set and the CLI is running,
 # you'll be prompted to enter your password when sudo is needed:
 # - 45-second timeout (auto-skips if no input)
 # - Press Enter to skip (command fails gracefully)
 # - Password is hidden while typing
 # - Password is cached for the session
 #
-# EMPTY PASSWORDS: Setting sudo_password to an explicit empty string is different
-# from leaving it unset. Hermes will try an empty password via `sudo -S` and
-# will not open the interactive prompt. This is useful for passwordless sudo,
-# Touch ID sudo setups, and environments where prompting is just noise.
-#
 # ALTERNATIVES:
 # - SSH backend: Configure passwordless sudo on the remote server
 # - Containers: Run as root inside the container (no sudo needed)
@@ -306,8 +283,15 @@ compression:
  # compression of older turns.
  protect_last_n: 20

-  # To pin a specific model/provider for compression summaries, use the
-  # auxiliary section below (auxiliary.compression.provider / model).
+  # Model to use for generating summaries (fast/cheap recommended)
+  # This model compresses the middle turns into a concise summary.
+  # IMPORTANT: it receives the full middle section of the conversation, so it
+  # MUST support a context length at least as large as your main model's.
+  summary_model: "google/gemini-3-flash-preview"
+  
+  # Provider for the summary model (default: "auto")
+  # Options: "auto", "openrouter", "nous", "main"
+  # summary_provider: "auto"

 # =============================================================================
 # Auxiliary Models (Advanced — Experimental)
@@ -333,7 +317,6 @@ compression:
 #   "openrouter" - Force OpenRouter (requires OPENROUTER_API_KEY)
 #   "nous"       - Force Nous Portal (requires: hermes login)
 #   "gemini"      - Force Google AI Studio direct (requires: GOOGLE_API_KEY or GEMINI_API_KEY)
-#   "ollama-cloud" - Ollama Cloud (requires: OLLAMA_API_KEY)
 #   "codex"       - Force Codex OAuth (requires: hermes model → Codex).
 #                  Uses gpt-5.3-codex which supports vision.
 #   "main"       - Use your custom endpoint (OPENAI_BASE_URL + OPENAI_API_KEY).
@@ -462,22 +445,6 @@ agent:
  # Higher = more room for complex tasks, but costs more tokens
  # Recommended: 20-30 for focused tasks, 50-100 for open exploration
  max_turns: 60
-
-  # Inactivity timeout for gateway agent runs (seconds, 0 = unlimited).
-  # The agent can run indefinitely when actively calling tools or receiving
-  # API responses.  Only fires after the agent has been idle for this duration.
-  # gateway_timeout: 1800
-
-  # Staged warning: send a warning before escalating to full timeout.
-  # Fires once per run when inactivity reaches this threshold (seconds).
-  # Set to 0 to disable the warning.
-  # gateway_timeout_warning: 900
-
-  # Graceful drain timeout for gateway stop/restart (seconds).
-  # The gateway stops accepting new work, waits for in-flight agents to
-  # finish, then interrupts anything still running after this timeout.
-  # 0 = no drain, interrupt immediately.
-  # restart_drain_timeout: 60
  
  # Enable verbose logging
  verbose: false
@@ -520,7 +487,7 @@ agent:
 #   - A preset like "hermes-cli" or "hermes-telegram" (curated tool set)
 #   - A list of individual toolsets to compose your own (see list below)
 #
-# Supported platform keys: cli, telegram, discord, whatsapp, slack, qqbot
+# Supported platform keys: cli, telegram, discord, whatsapp, slack
 #
 # Examples:
 #
@@ -549,7 +516,6 @@ agent:
 #   slack:         hermes-slack          (same as telegram)
 #   signal:        hermes-signal         (same as telegram)
 #   homeassistant: hermes-homeassistant  (same as telegram)
-#   qqbot:            hermes-qqbot            (same as telegram)
 #
 platform_toolsets:
  cli: [hermes-cli]
@@ -559,19 +525,6 @@ platform_toolsets:
  slack: [hermes-slack]
  signal: [hermes-signal]
  homeassistant: [hermes-homeassistant]
-  qqbot: [hermes-qqbot]
-
-# =============================================================================
-# Gateway Platform Settings
-# =============================================================================
-# Optional per-platform messaging settings.
-# Platform-specific knobs live under `extra`.
-#
-# platforms:
-#   telegram:
-#     reply_to_mode: "first"  # off | first | all
-#     extra:
-#       disable_link_previews: false  # Set true to suppress Telegram URL previews in bot messages

 # ─────────────────────────────────────────────────────────────────────────────
 # Available toolsets (use these names in platform_toolsets or the toolsets list)
@@ -594,7 +547,7 @@ platform_toolsets:
 #   skills_hub   - skill_hub (search/install/manage from online registries — user-driven only)
 #   moa          - mixture_of_agents  (requires OPENROUTER_API_KEY)
 #   todo         - todo (in-memory task planning, no deps)
-#   tts          - text_to_speech  (Edge TTS free, or ELEVENLABS/OPENAI/MINIMAX/MISTRAL key)
+#   tts          - text_to_speech  (Edge TTS free, or ELEVENLABS/OPENAI/MINIMAX key)
 #   cronjob      - cronjob (create/list/update/pause/resume/run/remove scheduled tasks)
 #   rl           - rl_list_environments, rl_start_training, etc. (requires TINKER_API_KEY)
 #
@@ -623,7 +576,7 @@ platform_toolsets:
 #   todo         - Task planning and tracking for multi-step work
 #   memory       - Persistent memory across sessions (personal notes + user profile)
 #   session_search - Search and recall past conversations (FTS5 + Gemini Flash summarization)
-#   tts          - Text-to-speech (Edge TTS free, ElevenLabs, OpenAI, MiniMax, Mistral)
+#   tts          - Text-to-speech (Edge TTS free, ElevenLabs, OpenAI, MiniMax)
 #   cronjob      - Schedule and manage automated tasks (CLI-only)
 #   rl           - RL training tools (Tinker-Atropos)
 #
@@ -696,11 +649,7 @@ platform_toolsets:
 stt:
  enabled: true
  # provider: "local"          # auto-detected if omitted
-  local:
-    model: "base"              # tiny | base | small | medium | large-v3 | turbo
-    # language: ""             # auto-detect; set to "en", "es", "fr", etc. to force
-  openai:
-    model: "whisper-1"         # whisper-1 | gpt-4o-mini-transcribe | gpt-4o-transcribe
+  model: "whisper-1"  # whisper-1 (cheapest) | gpt-4o-mini-transcribe | gpt-4o-transcribe
  # mistral:
  #   model: "voxtral-mini-latest"  # voxtral-mini-latest | voxtral-mini-2602

@@ -779,11 +728,6 @@ display:
  # Toggle at runtime with /verbose in the CLI
  tool_progress: all

-  # Gateway-only natural mid-turn assistant updates.
-  # When true, completed assistant status messages are sent as separate chat
-  # messages. This is independent of tool_progress and gateway streaming.
-  interim_assistant_messages: true
-
  # What Enter does when Hermes is already busy in the CLI.
  #   interrupt: Interrupt the current run and redirect Hermes (default)
  #   queue:     Queue your message for the next turn
@@ -792,7 +736,7 @@ display:

  # Background process notifications (gateway/messaging only).
  # Controls how chatty the process watcher is when you use
-  # terminal(background=true, notify_on_complete=true) from Telegram/Discord/etc.
+  # terminal(background=true, check_interval=...) from Telegram/Discord/etc.
  #   off:     No watcher messages at all
  #   result:  Only the final completion message
  #   error:   Only the final message when exit code != 0
@@ -1,15 +0,0 @@
-# Termux / Android dependency constraints for Hermes Agent.
-#
-# Usage:
-#   python -m pip install -e '.[termux]' -c constraints-termux.txt
-#
-# These pins keep the tested Android install path stable when upstream packages
-# move faster than Termux-compatible wheels / sdists.
-
-ipython<10
-jedi>=0.18.1,<0.20
-parso>=0.8.4,<0.9
-stack-data>=0.6,<0.7
-pexpect>4.3,<5
-matplotlib-inline>=0.1.7,<0.2
-asttokens>=2.1,<3
@@ -31,7 +31,7 @@ except ImportError:
 # Configuration
 # =============================================================================

-HERMES_DIR = get_hermes_home().resolve()
+HERMES_DIR = get_hermes_home()
 CRON_DIR = HERMES_DIR / "cron"
 JOBS_FILE = CRON_DIR / "jobs.json"
 OUTPUT_DIR = CRON_DIR / "output"
@@ -338,12 +338,10 @@ def load_jobs() -> List[Dict[str, Any]]:
                    save_jobs(jobs)
                    logger.warning("Auto-repaired jobs.json (had invalid control characters)")
                return jobs
-        except Exception as e:
-            logger.error("Failed to auto-repair jobs.json: %s", e)
-            raise RuntimeError(f"Cron database corrupted and unrepairable: {e}") from e
-    except IOError as e:
-        logger.error("IOError reading jobs.json: %s", e)
-        raise RuntimeError(f"Failed to read cron database: {e}") from e
+        except Exception:
+            return []
+    except IOError:
+        return []


 def save_jobs(jobs: List[Dict[str, Any]]):
@@ -454,7 +452,6 @@ def create_job(
        "last_run_at": None,
        "last_status": None,
        "last_error": None,
-        "last_delivery_error": None,
        # Delivery configuration
        "deliver": deliver,
        "origin": origin,  # Tracks where job was created for "origin" delivery
@@ -501,12 +498,6 @@ def update_job(job_id: str, updates: Dict[str, Any]) -> Optional[Dict[str, Any]]

        if schedule_changed:
            updated_schedule = updated["schedule"]
-            # The API may pass schedule as a raw string (e.g. "every 10m")
-            # instead of a pre-parsed dict.  Normalize it the same way
-            # create_job() does so downstream code can call .get() safely.
-            if isinstance(updated_schedule, str):
-                updated_schedule = parse_schedule(updated_schedule)
-                updated["schedule"] = updated_schedule
            updated["schedule_display"] = updates.get(
                "schedule_display",
                updated_schedule.get("display", updated.get("schedule_display")),
@@ -629,8 +620,8 @@ def mark_job_run(job_id: str, success: bool, error: Optional[str] = None,

            save_jobs(jobs)
            return
-
-    logger.warning("mark_job_run: job_id %s not found, skipping save", job_id)
+    
+    save_jobs(jobs)


 def advance_next_run(job_id: str) -> bool:
@@ -10,7 +10,6 @@ runs at a time if multiple processes overlap.

 import asyncio
 import concurrent.futures
-import contextvars
 import json
 import logging
 import os
@@ -27,7 +26,7 @@ except ImportError:
    except ImportError:
        msvcrt = None
 from pathlib import Path
-from typing import List, Optional
+from typing import Optional

 # Add parent directory to path for imports BEFORE repo-level imports.
 # Without this, standalone invocations (e.g. after `hermes update` reloads
@@ -45,37 +44,9 @@ logger = logging.getLogger(__name__)
 _KNOWN_DELIVERY_PLATFORMS = frozenset({
    "telegram", "discord", "slack", "whatsapp", "signal",
    "matrix", "mattermost", "homeassistant", "dingtalk", "feishu",
-    "wecom", "wecom_callback", "weixin", "sms", "email", "webhook", "bluebubbles",
-    "qqbot",
+    "wecom", "sms", "email", "webhook",
 })

-# Platforms that support a configured cron/notification home target, mapped to
-# the environment variable used by gateway setup/runtime config.
-_HOME_TARGET_ENV_VARS = {
-    "matrix": "MATRIX_HOME_ROOM",
-    "telegram": "TELEGRAM_HOME_CHANNEL",
-    "discord": "DISCORD_HOME_CHANNEL",
-    "slack": "SLACK_HOME_CHANNEL",
-    "signal": "SIGNAL_HOME_CHANNEL",
-    "mattermost": "MATTERMOST_HOME_CHANNEL",
-    "sms": "SMS_HOME_CHANNEL",
-    "email": "EMAIL_HOME_ADDRESS",
-    "dingtalk": "DINGTALK_HOME_CHANNEL",
-    "feishu": "FEISHU_HOME_CHANNEL",
-    "wecom": "WECOM_HOME_CHANNEL",
-    "weixin": "WEIXIN_HOME_CHANNEL",
-    "bluebubbles": "BLUEBUBBLES_HOME_CHANNEL",
-    "qqbot": "QQBOT_HOME_CHANNEL",
-}
-
-# Legacy env var names kept for back-compat.  Each entry is the current
-# primary env var → the previous name.  _get_home_target_chat_id falls
-# back to the legacy name if the primary is unset, so users who set the
-# old name before the rename keep working until they migrate.
-_LEGACY_HOME_TARGET_ENV_VARS = {
-    "QQBOT_HOME_CHANNEL": "QQ_HOME_CHANNEL",
-}
-
 from cron.jobs import get_due_jobs, mark_job_run, save_job_output, advance_next_run

 # Sentinel: when a cron agent has nothing new to report, it can start its
@@ -103,28 +74,15 @@ def _resolve_origin(job: dict) -> Optional[dict]:
    return None


-def _get_home_target_chat_id(platform_name: str) -> str:
-    """Return the configured home target chat/room ID for a delivery platform."""
-    env_var = _HOME_TARGET_ENV_VARS.get(platform_name.lower())
-    if not env_var:
-        return ""
-    value = os.getenv(env_var, "")
-    if not value:
-        legacy = _LEGACY_HOME_TARGET_ENV_VARS.get(env_var)
-        if legacy:
-            value = os.getenv(legacy, "")
-    return value
-
-
-def _resolve_single_delivery_target(job: dict, deliver_value: str) -> Optional[dict]:
-    """Resolve one concrete auto-delivery target for a cron job."""
-
+def _resolve_delivery_target(job: dict) -> Optional[dict]:
+    """Resolve the concrete auto-delivery target for a cron job, if any."""
+    deliver = job.get("deliver", "local")
    origin = _resolve_origin(job)

-    if deliver_value == "local":
+    if deliver == "local":
        return None

-    if deliver_value == "origin":
+    if deliver == "origin":
        if origin:
            return {
                "platform": origin["platform"],
@@ -133,8 +91,8 @@ def _resolve_single_delivery_target(job: dict, deliver_value: str) -> Optional[d
            }
        # Origin missing (e.g. job created via API/script) — try each
        # platform's home channel as a fallback instead of silently dropping.
-        for platform_name in _HOME_TARGET_ENV_VARS:
-            chat_id = _get_home_target_chat_id(platform_name)
+        for platform_name in ("matrix", "telegram", "discord", "slack"):
+            chat_id = os.getenv(f"{platform_name.upper()}_HOME_CHANNEL", "")
            if chat_id:
                logger.info(
                    "Job '%s' has deliver=origin but no origin; falling back to %s home channel",
@@ -148,8 +106,8 @@ def _resolve_single_delivery_target(job: dict, deliver_value: str) -> Optional[d
                }
        return None

-    if ":" in deliver_value:
-        platform_name, rest = deliver_value.split(":", 1)
+    if ":" in deliver:
+        platform_name, rest = deliver.split(":", 1)
        platform_key = platform_name.lower()

        from tools.send_message_tool import _parse_target_ref
@@ -179,7 +137,7 @@ def _resolve_single_delivery_target(job: dict, deliver_value: str) -> Optional[d
            "thread_id": thread_id,
        }

-    platform_name = deliver_value
+    platform_name = deliver
    if origin and origin.get("platform") == platform_name:
        return {
            "platform": platform_name,
@@ -189,7 +147,7 @@ def _resolve_single_delivery_target(job: dict, deliver_value: str) -> Optional[d

    if platform_name.lower() not in _KNOWN_DELIVERY_PLATFORMS:
        return None
-    chat_id = _get_home_target_chat_id(platform_name)
+    chat_id = os.getenv(f"{platform_name.upper()}_HOME_CHANNEL", "")
    if not chat_id:
        return None

@@ -200,30 +158,6 @@ def _resolve_single_delivery_target(job: dict, deliver_value: str) -> Optional[d
    }


-def _resolve_delivery_targets(job: dict) -> List[dict]:
-    """Resolve all concrete auto-delivery targets for a cron job (supports comma-separated deliver)."""
-    deliver = job.get("deliver", "local")
-    if deliver == "local":
-        return []
-    parts = [p.strip() for p in str(deliver).split(",") if p.strip()]
-    seen = set()
-    targets = []
-    for part in parts:
-        target = _resolve_single_delivery_target(job, part)
-        if target:
-            key = (target["platform"].lower(), str(target["chat_id"]), target.get("thread_id"))
-            if key not in seen:
-                seen.add(key)
-                targets.append(target)
-    return targets
-
-
-def _resolve_delivery_target(job: dict) -> Optional[dict]:
-    """Resolve the concrete auto-delivery target for a cron job, if any."""
-    targets = _resolve_delivery_targets(job)
-    return targets[0] if targets else None
-
-
 # Media extension sets — keep in sync with gateway/platforms/base.py:_process_message_background
 _AUDIO_EXTS = frozenset({'.ogg', '.opus', '.mp3', '.wav', '.m4a'})
 _VIDEO_EXTS = frozenset({'.mp4', '.mov', '.avi', '.mkv', '.webm', '.3gp'})
@@ -264,7 +198,7 @@ def _send_media_via_adapter(adapter, chat_id: str, media_files: list, metadata:

 def _deliver_result(job: dict, content: str, adapters=None, loop=None) -> Optional[str]:
    """
-    Deliver job output to the configured target(s) (origin chat, specific platform, etc.).
+    Deliver job output to the configured target (origin chat, specific platform, etc.).

    When ``adapters`` and ``loop`` are provided (gateway is running), tries to
    use the live adapter first — this supports E2EE rooms (e.g. Matrix) where
@@ -273,14 +207,18 @@ def _deliver_result(job: dict, content: str, adapters=None, loop=None) -> Option

    Returns None on success, or an error string on failure.
    """
-    targets = _resolve_delivery_targets(job)
-    if not targets:
+    target = _resolve_delivery_target(job)
+    if not target:
        if job.get("deliver", "local") != "local":
            msg = f"no delivery target resolved for deliver={job.get('deliver', 'local')}"
            logger.warning("Job '%s': %s", job["id"], msg)
            return msg
        return None  # local-only jobs don't deliver — not a failure

+    platform_name = target["platform"]
+    chat_id = target["chat_id"]
+    thread_id = target.get("thread_id")
+
    from tools.send_message_tool import _send_to_platform
    from gateway.config import load_gateway_config, Platform

@@ -296,13 +234,27 @@ def _deliver_result(job: dict, content: str, adapters=None, loop=None) -> Option
        "dingtalk": Platform.DINGTALK,
        "feishu": Platform.FEISHU,
        "wecom": Platform.WECOM,
-        "wecom_callback": Platform.WECOM_CALLBACK,
-        "weixin": Platform.WEIXIN,
        "email": Platform.EMAIL,
        "sms": Platform.SMS,
-        "bluebubbles": Platform.BLUEBUBBLES,
-        "qqbot": Platform.QQBOT,
    }
+    platform = platform_map.get(platform_name.lower())
+    if not platform:
+        msg = f"unknown platform '{platform_name}'"
+        logger.warning("Job '%s': %s", job["id"], msg)
+        return msg
+
+    try:
+        config = load_gateway_config()
+    except Exception as e:
+        msg = f"failed to load gateway config: {e}"
+        logger.error("Job '%s': %s", job["id"], msg)
+        return msg
+
+    pconfig = config.platforms.get(platform)
+    if not pconfig or not pconfig.enabled:
+        msg = f"platform '{platform_name}' not configured/enabled"
+        logger.warning("Job '%s': %s", job["id"], msg)
+        return msg

    # Optionally wrap the content with a header/footer so the user knows this
    # is a cron delivery.  Wrapping is on by default; set cron.wrap_response: false
@@ -316,13 +268,11 @@ def _deliver_result(job: dict, content: str, adapters=None, loop=None) -> Option

    if wrap_response:
        task_name = job.get("name", job["id"])
-        job_id = job.get("id", "")
        delivery_content = (
            f"Cronjob Response: {task_name}\n"
-            f"(job_id: {job_id})\n"
            f"-------------\n\n"
            f"{content}\n\n"
-            f"To stop or manage this job, send me a new message (e.g. \"stop reminder {task_name}\")."
+            f"Note: The agent cannot see this message, and therefore cannot respond to it."
        )
    else:
        delivery_content = content
@@ -331,156 +281,71 @@ def _deliver_result(job: dict, content: str, adapters=None, loop=None) -> Option
    from gateway.platforms.base import BasePlatformAdapter
    media_files, cleaned_delivery_content = BasePlatformAdapter.extract_media(delivery_content)

+    # Prefer the live adapter when the gateway is running — this supports E2EE
+    # rooms (e.g. Matrix) where the standalone HTTP path cannot encrypt.
+    runtime_adapter = (adapters or {}).get(platform)
+    if runtime_adapter is not None and loop is not None and getattr(loop, "is_running", lambda: False)():
+        send_metadata = {"thread_id": thread_id} if thread_id else None
+        try:
+            # Send cleaned text (MEDIA tags stripped) — not the raw content
+            text_to_send = cleaned_delivery_content.strip()
+            adapter_ok = True
+            if text_to_send:
+                future = asyncio.run_coroutine_threadsafe(
+                    runtime_adapter.send(chat_id, text_to_send, metadata=send_metadata),
+                    loop,
+                )
+                send_result = future.result(timeout=60)
+                if send_result and not getattr(send_result, "success", True):
+                    err = getattr(send_result, "error", "unknown")
+                    logger.warning(
+                        "Job '%s': live adapter send to %s:%s failed (%s), falling back to standalone",
+                        job["id"], platform_name, chat_id, err,
+                    )
+                    adapter_ok = False  # fall through to standalone path
+
+            # Send extracted media files as native attachments via the live adapter
+            if adapter_ok and media_files:
+                _send_media_via_adapter(runtime_adapter, chat_id, media_files, send_metadata, loop, job)
+
+            if adapter_ok:
+                logger.info("Job '%s': delivered to %s:%s via live adapter", job["id"], platform_name, chat_id)
+                return None
+        except Exception as e:
+            logger.warning(
+                "Job '%s': live adapter delivery to %s:%s failed (%s), falling back to standalone",
+                job["id"], platform_name, chat_id, e,
+            )
+
+    # Standalone path: run the async send in a fresh event loop (safe from any thread)
+    coro = _send_to_platform(platform, pconfig, chat_id, cleaned_delivery_content, thread_id=thread_id, media_files=media_files)
    try:
-        config = load_gateway_config()
+        result = asyncio.run(coro)
+    except RuntimeError:
+        # asyncio.run() checks for a running loop before awaiting the coroutine;
+        # when it raises, the original coro was never started — close it to
+        # prevent "coroutine was never awaited" RuntimeWarning, then retry in a
+        # fresh thread that has no running loop.
+        coro.close()
+        import concurrent.futures
+        with concurrent.futures.ThreadPoolExecutor(max_workers=1) as pool:
+            future = pool.submit(asyncio.run, _send_to_platform(platform, pconfig, chat_id, cleaned_delivery_content, thread_id=thread_id, media_files=media_files))
+            result = future.result(timeout=30)
    except Exception as e:
-        msg = f"failed to load gateway config: {e}"
+        msg = f"delivery to {platform_name}:{chat_id} failed: {e}"
        logger.error("Job '%s': %s", job["id"], msg)
        return msg

-    delivery_errors = []
+    if result and result.get("error"):
+        msg = f"delivery error: {result['error']}"
+        logger.error("Job '%s': %s", job["id"], msg)
+        return msg

-    for target in targets:
-        platform_name = target["platform"]
-        chat_id = target["chat_id"]
-        thread_id = target.get("thread_id")
-
-        # Diagnostic: log thread_id for topic-aware delivery debugging
-        origin = job.get("origin") or {}
-        origin_thread = origin.get("thread_id")
-        if origin_thread and not thread_id:
-            logger.warning(
-                "Job '%s': origin has thread_id=%s but delivery target lost it "
-                "(deliver=%s, target=%s)",
-                job["id"], origin_thread, job.get("deliver", "local"), target,
-            )
-        elif thread_id:
-            logger.debug(
-                "Job '%s': delivering to %s:%s thread_id=%s",
-                job["id"], platform_name, chat_id, thread_id,
-            )
-
-        platform = platform_map.get(platform_name.lower())
-        if not platform:
-            msg = f"unknown platform '{platform_name}'"
-            logger.warning("Job '%s': %s", job["id"], msg)
-            delivery_errors.append(msg)
-            continue
-
-        # Prefer the live adapter when the gateway is running — this supports E2EE
-        # rooms (e.g. Matrix) where the standalone HTTP path cannot encrypt.
-        runtime_adapter = (adapters or {}).get(platform)
-        delivered = False
-        if runtime_adapter is not None and loop is not None and getattr(loop, "is_running", lambda: False)():
-            send_metadata = {"thread_id": thread_id} if thread_id else None
-            try:
-                # Send cleaned text (MEDIA tags stripped) — not the raw content
-                text_to_send = cleaned_delivery_content.strip()
-                adapter_ok = True
-                if text_to_send:
-                    future = asyncio.run_coroutine_threadsafe(
-                        runtime_adapter.send(chat_id, text_to_send, metadata=send_metadata),
-                        loop,
-                    )
-                    send_result = future.result(timeout=60)
-                    if send_result and not getattr(send_result, "success", True):
-                        err = getattr(send_result, "error", "unknown")
-                        logger.warning(
-                            "Job '%s': live adapter send to %s:%s failed (%s), falling back to standalone",
-                            job["id"], platform_name, chat_id, err,
-                        )
-                        adapter_ok = False  # fall through to standalone path
-
-                # Send extracted media files as native attachments via the live adapter
-                if adapter_ok and media_files:
-                    _send_media_via_adapter(runtime_adapter, chat_id, media_files, send_metadata, loop, job)
-
-                if adapter_ok:
-                    logger.info("Job '%s': delivered to %s:%s via live adapter", job["id"], platform_name, chat_id)
-                    delivered = True
-            except Exception as e:
-                logger.warning(
-                    "Job '%s': live adapter delivery to %s:%s failed (%s), falling back to standalone",
-                    job["id"], platform_name, chat_id, e,
-                )
-
-        if not delivered:
-            pconfig = config.platforms.get(platform)
-            if not pconfig or not pconfig.enabled:
-                msg = f"platform '{platform_name}' not configured/enabled"
-                logger.warning("Job '%s': %s", job["id"], msg)
-                delivery_errors.append(msg)
-                continue
-
-            # Standalone path: run the async send in a fresh event loop (safe from any thread)
-            coro = _send_to_platform(platform, pconfig, chat_id, cleaned_delivery_content, thread_id=thread_id, media_files=media_files)
-            try:
-                result = asyncio.run(coro)
-            except RuntimeError:
-                # asyncio.run() checks for a running loop before awaiting the coroutine;
-                # when it raises, the original coro was never started — close it to
-                # prevent "coroutine was never awaited" RuntimeWarning, then retry in a
-                # fresh thread that has no running loop.
-                coro.close()
-                import concurrent.futures
-                with concurrent.futures.ThreadPoolExecutor(max_workers=1) as pool:
-                    future = pool.submit(asyncio.run, _send_to_platform(platform, pconfig, chat_id, cleaned_delivery_content, thread_id=thread_id, media_files=media_files))
-                    result = future.result(timeout=30)
-            except Exception as e:
-                msg = f"delivery to {platform_name}:{chat_id} failed: {e}"
-                logger.error("Job '%s': %s", job["id"], msg)
-                delivery_errors.append(msg)
-                continue
-
-            if result and result.get("error"):
-                msg = f"delivery error: {result['error']}"
-                logger.error("Job '%s': %s", job["id"], msg)
-                delivery_errors.append(msg)
-                continue
-
-            logger.info("Job '%s': delivered to %s:%s", job["id"], platform_name, chat_id)
-
-    if delivery_errors:
-        return "; ".join(delivery_errors)
+    logger.info("Job '%s': delivered to %s:%s", job["id"], platform_name, chat_id)
    return None


-_DEFAULT_SCRIPT_TIMEOUT = 120  # seconds
-# Backward-compatible module override used by tests and emergency monkeypatches.
-_SCRIPT_TIMEOUT = _DEFAULT_SCRIPT_TIMEOUT
-
-
-def _get_script_timeout() -> int:
-    """Resolve cron pre-run script timeout from module/env/config with a safe default."""
-    if _SCRIPT_TIMEOUT != _DEFAULT_SCRIPT_TIMEOUT:
-        try:
-            timeout = int(float(_SCRIPT_TIMEOUT))
-            if timeout > 0:
-                return timeout
-        except Exception:
-            logger.warning("Invalid patched _SCRIPT_TIMEOUT=%r; using env/config/default", _SCRIPT_TIMEOUT)
-
-    env_value = os.getenv("HERMES_CRON_SCRIPT_TIMEOUT", "").strip()
-    if env_value:
-        try:
-            timeout = int(float(env_value))
-            if timeout > 0:
-                return timeout
-        except Exception:
-            logger.warning("Invalid HERMES_CRON_SCRIPT_TIMEOUT=%r; using config/default", env_value)
-
-    try:
-        cfg = load_config() or {}
-        cron_cfg = cfg.get("cron", {}) if isinstance(cfg, dict) else {}
-        configured = cron_cfg.get("script_timeout_seconds")
-        if configured is not None:
-            timeout = int(float(configured))
-            if timeout > 0:
-                return timeout
-    except Exception as exc:
-        logger.debug("Failed to load cron script timeout from config: %s", exc)
-
-    return _DEFAULT_SCRIPT_TIMEOUT
+_SCRIPT_TIMEOUT = 120  # seconds


 def _run_job_script(script_path: str) -> tuple[bool, str]:
@@ -527,27 +392,17 @@ def _run_job_script(script_path: str) -> tuple[bool, str]:
    if not path.is_file():
        return False, f"Script path is not a file: {path}"

-    script_timeout = _get_script_timeout()
-
    try:
        result = subprocess.run(
            [sys.executable, str(path)],
            capture_output=True,
            text=True,
-            timeout=script_timeout,
+            timeout=_SCRIPT_TIMEOUT,
            cwd=str(path.parent),
        )
        stdout = (result.stdout or "").strip()
        stderr = (result.stderr or "").strip()

-        # Redact secrets from both stdout and stderr before any return path.
-        try:
-            from agent.redact import redact_sensitive_text
-            stdout = redact_sensitive_text(stdout)
-            stderr = redact_sensitive_text(stderr)
-        except Exception:
-            pass
-
        if result.returncode != 0:
            parts = [f"Script exited with code {result.returncode}"]
            if stderr:
@@ -556,61 +411,30 @@ def _run_job_script(script_path: str) -> tuple[bool, str]:
                parts.append(f"stdout:\n{stdout}")
            return False, "\n".join(parts)

+        # Redact any secrets that may appear in script output before
+        # they are injected into the LLM prompt context.
+        try:
+            from agent.redact import redact_sensitive_text
+            stdout = redact_sensitive_text(stdout)
+        except Exception:
+            pass
        return True, stdout

    except subprocess.TimeoutExpired:
-        return False, f"Script timed out after {script_timeout}s: {path}"
+        return False, f"Script timed out after {_SCRIPT_TIMEOUT}s: {path}"
    except Exception as exc:
        return False, f"Script execution failed: {exc}"


-def _parse_wake_gate(script_output: str) -> bool:
-    """Parse the last non-empty stdout line of a cron job's pre-check script
-    as a wake gate.
-
-    The convention (ported from nanoclaw #1232): if the last stdout line is
-    JSON like ``{"wakeAgent": false}``, the agent is skipped entirely — no
-    LLM run, no delivery. Any other output (non-JSON, missing flag, gate
-    absent, or ``wakeAgent: true``) means wake the agent normally.
-
-    Returns True if the agent should wake, False to skip.
-    """
-    if not script_output:
-        return True
-    stripped_lines = [line for line in script_output.splitlines() if line.strip()]
-    if not stripped_lines:
-        return True
-    last_line = stripped_lines[-1].strip()
-    try:
-        gate = json.loads(last_line)
-    except (json.JSONDecodeError, ValueError):
-        return True
-    if not isinstance(gate, dict):
-        return True
-    return gate.get("wakeAgent", True) is not False
-
-
-def _build_job_prompt(job: dict, prerun_script: Optional[tuple] = None) -> str:
-    """Build the effective prompt for a cron job, optionally loading one or more skills first.
-
-    Args:
-        job: The cron job dict.
-        prerun_script: Optional ``(success, stdout)`` from a script that has
-            already been executed by the caller (e.g. for a wake-gate check).
-            When provided, the script is not re-executed and the cached
-            result is used for prompt injection. When omitted, the script
-            (if any) runs inline as before.
-    """
+def _build_job_prompt(job: dict) -> str:
+    """Build the effective prompt for a cron job, optionally loading one or more skills first."""
    prompt = job.get("prompt", "")
    skills = job.get("skills")

    # Run data-collection script if configured, inject output as context.
    script_path = job.get("script")
    if script_path:
-        if prerun_script is not None:
-            success, script_output = prerun_script
-        else:
-            success, script_output = _run_job_script(script_path)
+        success, script_output = _run_job_script(script_path)
        if success:
            if script_output:
                prompt = (
@@ -712,41 +536,13 @@ def run_job(job: dict) -> tuple[bool, str, str, Optional[str]]:
    
    job_id = job["id"]
    job_name = job["name"]
-
-    # Wake-gate: if this job has a pre-check script, run it BEFORE building
-    # the prompt so a ``{"wakeAgent": false}`` response can short-circuit
-    # the whole agent run. We pass the result into _build_job_prompt so
-    # the script is only executed once.
-    prerun_script = None
-    script_path = job.get("script")
-    if script_path:
-        prerun_script = _run_job_script(script_path)
-        _ran_ok, _script_output = prerun_script
-        if _ran_ok and not _parse_wake_gate(_script_output):
-            logger.info(
-                "Job '%s' (ID: %s): wakeAgent=false, skipping agent run",
-                job_name, job_id,
-            )
-            silent_doc = (
-                f"# Cron Job: {job_name}\n\n"
-                f"**Job ID:** {job_id}\n"
-                f"**Run Time:** {_hermes_now().strftime('%Y-%m-%d %H:%M:%S')}\n\n"
-                "Script gate returned `wakeAgent=false` — agent skipped.\n"
-            )
-            return True, silent_doc, SILENT_MARKER, None
-
-    prompt = _build_job_prompt(job, prerun_script=prerun_script)
+    prompt = _build_job_prompt(job)
    origin = _resolve_origin(job)
    _cron_session_id = f"cron_{job_id}_{_hermes_now().strftime('%Y%m%d_%H%M%S')}"

    logger.info("Running job '%s' (ID: %s)", job_name, job_id)
    logger.info("Prompt: %s", prompt[:100])

-    # Mark this as a cron session so the approval system can apply cron_mode.
-    # This env var is process-wide and persists for the lifetime of the
-    # scheduler process — every job this process runs is a cron job.
-    os.environ["HERMES_CRON_SESSION"] = "1"
-
    try:
        # Inject origin context so the agent's send_message tool knows the chat.
        # Must be INSIDE the try block so the finally cleanup always runs.
@@ -789,15 +585,6 @@ def run_job(job: dict) -> tuple[bool, str, str, Optional[str]]:
        except Exception as e:
            logger.warning("Job '%s': failed to load config.yaml, using defaults: %s", job_id, e)

-        # Apply IPv4 preference if configured.
-        try:
-            from hermes_constants import apply_ipv4_preference
-            _net_cfg = _cfg.get("network", {})
-            if isinstance(_net_cfg, dict) and _net_cfg.get("force_ipv4"):
-                apply_ipv4_preference(force=True)
-        except Exception:
-            pass
-
        # Reasoning config from config.yaml
        from hermes_constants import parse_reasoning_effort
        effort = str(_cfg.get("agent", {}).get("reasoning_effort", "")).strip()
@@ -858,24 +645,6 @@ def run_job(job: dict) -> tuple[bool, str, str, Optional[str]]:
            },
        )

-        fallback_model = _cfg.get("fallback_providers") or _cfg.get("fallback_model") or None
-        credential_pool = None
-        runtime_provider = str(turn_route["runtime"].get("provider") or "").strip().lower()
-        if runtime_provider:
-            try:
-                from agent.credential_pool import load_pool
-                pool = load_pool(runtime_provider)
-                if pool.has_credentials():
-                    credential_pool = pool
-                    logger.info(
-                        "Job '%s': loaded credential pool for provider %s with %d entries",
-                        job_id,
-                        runtime_provider,
-                        len(pool.entries()),
-                    )
-            except Exception as e:
-                logger.debug("Job '%s': failed to load credential pool for %s: %s", job_id, runtime_provider, e)
-
        agent = AIAgent(
            model=turn_route["model"],
            api_key=turn_route["runtime"].get("api_key"),
@@ -887,15 +656,12 @@ def run_job(job: dict) -> tuple[bool, str, str, Optional[str]]:
            max_iterations=max_iterations,
            reasoning_config=reasoning_config,
            prefill_messages=prefill_messages,
-            fallback_model=fallback_model,
-            credential_pool=credential_pool,
            providers_allowed=pr.get("only"),
            providers_ignored=pr.get("ignore"),
            providers_order=pr.get("order"),
            provider_sort=pr.get("sort"),
            disabled_toolsets=["cronjob", "messaging", "clarify"],
            quiet_mode=True,
-            skip_context_files=True,  # Don't inject SOUL.md/AGENTS.md from scheduler cwd
            skip_memory=True,  # Cron system prompts would corrupt user representations
            platform="cron",
            session_id=_cron_session_id,
@@ -914,11 +680,7 @@ def run_job(job: dict) -> tuple[bool, str, str, Optional[str]]:
        _cron_inactivity_limit = _cron_timeout if _cron_timeout > 0 else None
        _POLL_INTERVAL = 5.0
        _cron_pool = concurrent.futures.ThreadPoolExecutor(max_workers=1)
-        # Preserve scheduler-scoped ContextVar state (for example skill-declared
-        # env passthrough registrations) when the cron run hops into the worker
-        # thread used for inactivity timeout monitoring.
-        _cron_context = contextvars.copy_context()
-        _cron_future = _cron_pool.submit(_cron_context.run, agent.run_conversation, prompt)
+        _cron_future = _cron_pool.submit(agent.run_conversation, prompt)
        _inactivity_timeout = False
        try:
            if _cron_inactivity_limit is None:
@@ -948,7 +710,7 @@ def run_job(job: dict) -> tuple[bool, str, str, Optional[str]]:
            _cron_pool.shutdown(wait=False, cancel_futures=True)
            raise
        finally:
-            _cron_pool.shutdown(wait=False, cancel_futures=True)
+            _cron_pool.shutdown(wait=False)

        if _inactivity_timeout:
            # Build diagnostic summary from the agent's activity tracker.
@@ -980,9 +742,6 @@ def run_job(job: dict) -> tuple[bool, str, str, Optional[str]]:
            )

        final_response = result.get("final_response", "") or ""
-        # Strip leaked placeholder text that upstream may inject on empty completions.
-        if final_response.strip() == "(No response generated)":
-            final_response = ""
        # Use a separate variable for log display; keep final_response clean
        # for delivery logic (empty response = no delivery).
        logged_response = final_response if final_response else "(No response generated)"
@@ -1122,13 +881,6 @@ def tick(verbose: bool = True, adapters=None, loop=None) -> int:
                        delivery_error = str(de)
                        logger.error("Delivery failed for job %s: %s", job["id"], de)

-                # Treat empty final_response as a soft failure so last_status
-                # is not "ok" — the agent ran but produced nothing useful.
-                # (issue #8585)
-                if success and not final_response:
-                    success = False
-                    error = "Agent completed but produced empty response (model error, timeout, or misconfiguration)"
-
                mark_job_run(job["id"], success, error, delivery_error=delivery_error)
                executed += 1

@@ -1,52 +1,15 @@
 #!/bin/bash
-# Docker/Podman entrypoint: bootstrap config files into the mounted volume, then run hermes.
+# Docker entrypoint: bootstrap config files into the mounted volume, then run hermes.
 set -e

-HERMES_HOME="${HERMES_HOME:-/opt/data}"
+HERMES_HOME="/opt/data"
 INSTALL_DIR="/opt/hermes"

-# --- Privilege dropping via gosu ---
-# When started as root (the default for Docker, or fakeroot in rootless Podman),
-# optionally remap the hermes user/group to match host-side ownership, fix volume
-# permissions, then re-exec as hermes.
-if [ "$(id -u)" = "0" ]; then
-    if [ -n "$HERMES_UID" ] && [ "$HERMES_UID" != "$(id -u hermes)" ]; then
-        echo "Changing hermes UID to $HERMES_UID"
-        usermod -u "$HERMES_UID" hermes
-    fi
-
-    if [ -n "$HERMES_GID" ] && [ "$HERMES_GID" != "$(id -g hermes)" ]; then
-        echo "Changing hermes GID to $HERMES_GID"
-        # -o allows non-unique GID (e.g. macOS GID 20 "staff" may already exist
-        # as "dialout" in the Debian-based container image)
-        groupmod -o -g "$HERMES_GID" hermes 2>/dev/null || true
-    fi
-
-    actual_hermes_uid=$(id -u hermes)
-    if [ "$(stat -c %u "$HERMES_HOME" 2>/dev/null)" != "$actual_hermes_uid" ]; then
-        echo "$HERMES_HOME is not owned by $actual_hermes_uid, fixing"
-        # In rootless Podman the container's "root" is mapped to an unprivileged
-        # host UID — chown will fail.  That's fine: the volume is already owned
-        # by the mapped user on the host side.
-        chown -R hermes:hermes "$HERMES_HOME" 2>/dev/null || \
-            echo "Warning: chown failed (rootless container?) — continuing anyway"
-    fi
-
-    echo "Dropping root privileges"
-    exec gosu hermes "$0" "$@"
-fi
-
-# --- Running as hermes from here ---
-source "${INSTALL_DIR}/.venv/bin/activate"
-
 # Create essential directory structure.  Cache and platform directories
 # (cache/images, cache/audio, platforms/whatsapp, etc.) are created on
 # demand by the application — don't pre-create them here so new installs
 # get the consolidated layout from get_hermes_dir().
-# The "home/" subdirectory is a per-profile HOME for subprocesses (git,
-# ssh, gh, npm …).  Without it those tools write to /root which is
-# ephemeral and shared across profiles.  See issue #4426.
-mkdir -p "$HERMES_HOME"/{cron,sessions,logs,hooks,memories,skills,skins,plans,workspace,home}
+mkdir -p "$HERMES_HOME"/{cron,sessions,logs,hooks,memories,skills}

 # .env
 if [ ! -f "$HERMES_HOME/.env" ]; then
@@ -0,0 +1,228 @@
+# Hermes Agent — ACP (Agent Client Protocol) Setup Guide
+
+Hermes Agent supports the **Agent Client Protocol (ACP)**, allowing it to run as
+a coding agent inside your editor. ACP lets your IDE send tasks to Hermes, and
+Hermes responds with file edits, terminal commands, and explanations — all shown
+natively in the editor UI.
+
+---
+
+## Prerequisites
+
+- Hermes Agent installed and configured (`hermes setup` completed)
+- An API key / provider set up in `~/.hermes/.env` or via `hermes login`
+- Python 3.11+
+
+Install the ACP extra:
+
+```bash
+pip install -e ".[acp]"
+```
+
+---
+
+## VS Code Setup
+
+### 1. Install the ACP Client extension
+
+Open VS Code and install **ACP Client** from the marketplace:
+
+- Press `Ctrl+Shift+X` (or `Cmd+Shift+X` on macOS)
+- Search for **"ACP Client"**
+- Click **Install**
+
+Or install from the command line:
+
+```bash
+code --install-extension anysphere.acp-client
+```
+
+### 2. Configure settings.json
+
+Open your VS Code settings (`Ctrl+,` → click the `{}` icon for JSON) and add:
+
+```json
+{
+  "acpClient.agents": [
+    {
+      "name": "hermes-agent",
+      "registryDir": "/path/to/hermes-agent/acp_registry"
+    }
+  ]
+}
+```
+
+Replace `/path/to/hermes-agent` with the actual path to your Hermes Agent
+installation (e.g. `~/.hermes/hermes-agent`).
+
+Alternatively, if `hermes` is on your PATH, the ACP Client can discover it
+automatically via the registry directory.
+
+### 3. Restart VS Code
+
+After configuring, restart VS Code. You should see **Hermes Agent** appear in
+the ACP agent picker in the chat/agent panel.
+
+---
+
+## Zed Setup
+
+Zed has built-in ACP support.
+
+### 1. Configure Zed settings
+
+Open Zed settings (`Cmd+,` on macOS or `Ctrl+,` on Linux) and add to your
+`settings.json`:
+
+```json
+{
+  "agent_servers": {
+    "hermes-agent": {
+      "type": "custom",
+      "command": "hermes",
+      "args": ["acp"],
+    },
+  },
+}
+```
+
+### 2. Restart Zed
+
+Hermes Agent will appear in the agent panel. Select it and start a conversation.
+
+---
+
+## JetBrains Setup (IntelliJ, PyCharm, WebStorm, etc.)
+
+### 1. Install the ACP plugin
+
+- Open **Settings** → **Plugins** → **Marketplace**
+- Search for **"ACP"** or **"Agent Client Protocol"**
+- Install and restart the IDE
+
+### 2. Configure the agent
+
+- Open **Settings** → **Tools** → **ACP Agents**
+- Click **+** to add a new agent
+- Set the registry directory to your `acp_registry/` folder:
+  `/path/to/hermes-agent/acp_registry`
+- Click **OK**
+
+### 3. Use the agent
+
+Open the ACP panel (usually in the right sidebar) and select **Hermes Agent**.
+
+---
+
+## What You Will See
+
+Once connected, your editor provides a native interface to Hermes Agent:
+
+### Chat Panel
+A conversational interface where you can describe tasks, ask questions, and
+give instructions. Hermes responds with explanations and actions.
+
+### File Diffs
+When Hermes edits files, you see standard diffs in the editor. You can:
+- **Accept** individual changes
+- **Reject** changes you don't want
+- **Review** the full diff before applying
+
+### Terminal Commands
+When Hermes needs to run shell commands (builds, tests, installs), the editor
+shows them in an integrated terminal. Depending on your settings:
+- Commands may run automatically
+- Or you may be prompted to **approve** each command
+
+### Approval Flow
+For potentially destructive operations, the editor will prompt you for
+approval before Hermes proceeds. This includes:
+- File deletions
+- Shell commands
+- Git operations
+
+---
+
+## Configuration
+
+Hermes Agent under ACP uses the **same configuration** as the CLI:
+
+- **API keys / providers**: `~/.hermes/.env`
+- **Agent config**: `~/.hermes/config.yaml`
+- **Skills**: `~/.hermes/skills/`
+- **Sessions**: `~/.hermes/state.db`
+
+You can run `hermes setup` to configure providers, or edit `~/.hermes/.env`
+directly.
+
+### Changing the model
+
+Edit `~/.hermes/config.yaml`:
+
+```yaml
+model: openrouter/nous/hermes-3-llama-3.1-70b
+```
+
+Or set the `HERMES_MODEL` environment variable.
+
+### Toolsets
+
+ACP sessions use the curated `hermes-acp` toolset by default. It is designed for editor workflows and intentionally excludes things like messaging delivery, cronjob management, and audio-first UX features.
+
+---
+
+## Troubleshooting
+
+### Agent doesn't appear in the editor
+
+1. **Check the registry path** — make sure the `acp_registry/` directory path
+   in your editor settings is correct and contains `agent.json`.
+2. **Check `hermes` is on PATH** — run `which hermes` in a terminal. If not
+   found, you may need to activate your virtualenv or add it to PATH.
+3. **Restart the editor** after changing settings.
+
+### Agent starts but errors immediately
+
+1. Run `hermes doctor` to check your configuration.
+2. Check that you have a valid API key: `hermes status`
+3. Try running `hermes acp` directly in a terminal to see error output.
+
+### "Module not found" errors
+
+Make sure you installed the ACP extra:
+
+```bash
+pip install -e ".[acp]"
+```
+
+### Slow responses
+
+- ACP streams responses, so you should see incremental output. If the agent
+  appears stuck, check your network connection and API provider status.
+- Some providers have rate limits. Try switching to a different model/provider.
+
+### Permission denied for terminal commands
+
+If the editor blocks terminal commands, check your ACP Client extension
+settings for auto-approval or manual-approval preferences.
+
+### Logs
+
+Hermes logs are written to stderr when running in ACP mode. Check:
+- VS Code: **Output** panel → select **ACP Client** or **Hermes Agent**
+- Zed: **View** → **Toggle Terminal** and check the process output
+- JetBrains: **Event Log** or the ACP tool window
+
+You can also enable verbose logging:
+
+```bash
+HERMES_LOG_LEVEL=DEBUG hermes acp
+```
+
+---
+
+## Further Reading
+
+- [ACP Specification](https://github.com/anysphere/acp)
+- [Hermes Agent Documentation](https://github.com/NousResearch/hermes-agent)
+- Run `hermes --help` for all CLI options
@@ -0,0 +1,698 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+<meta charset="UTF-8">
+<meta name="viewport" content="width=device-width, initial-scale=1.0">
+<title>honcho-integration-spec</title>
+<style>
+  :root {
+    --bg:             #0b0e14;
+    --bg-surface:     #11151c;
+    --bg-elevated:    #181d27;
+    --bg-code:        #0d1018;
+    --fg:             #c9d1d9;
+    --fg-bright:      #e6edf3;
+    --fg-muted:       #6e7681;
+    --fg-subtle:      #484f58;
+    --accent:         #7eb8f6;
+    --accent-dim:     #3d6ea5;
+    --accent-glow:    rgba(126, 184, 246, 0.08);
+    --green:          #7ee6a8;
+    --green-dim:      #2ea04f;
+    --orange:         #e6a855;
+    --red:            #f47067;
+    --purple:         #bc8cff;
+    --cyan:           #56d4dd;
+    --border:         #21262d;
+    --border-subtle:  #161b22;
+    --radius:         6px;
+    --font-sans:      'New York', ui-serif, 'Iowan Old Style', 'Apple Garamond', Baskerville, 'Times New Roman', 'Noto Emoji', serif;
+    --font-mono:      'Departure Mono', 'Noto Emoji', monospace;
+  }
+
+  *, *::before, *::after { box-sizing: border-box; margin: 0; padding: 0; }
+  html { scroll-behavior: smooth; scroll-padding-top: 2rem; }
+  body {
+    font-family: var(--font-sans);
+    background: var(--bg);
+    color: var(--fg);
+    line-height: 1.7;
+    font-size: 15px;
+    -webkit-font-smoothing: antialiased;
+  }
+
+  .container { max-width: 860px; margin: 0 auto; padding: 3rem 2rem 6rem; }
+
+  .hero {
+    text-align: center;
+    padding: 4rem 0 3rem;
+    border-bottom: 1px solid var(--border);
+    margin-bottom: 3rem;
+  }
+  .hero h1 { font-family: var(--font-mono); font-size: 2.2rem; font-weight: 700; color: var(--fg-bright); letter-spacing: -0.03em; margin-bottom: 0.5rem; }
+  .hero h1 span { color: var(--accent); }
+  .hero .subtitle { font-family: var(--font-sans); color: var(--fg-muted); font-size: 0.92rem; max-width: 560px; margin: 0 auto; line-height: 1.6; }
+  .hero .meta { margin-top: 1.5rem; display: flex; justify-content: center; gap: 1.5rem; flex-wrap: wrap; }
+  .hero .meta span { font-size: 0.8rem; color: var(--fg-subtle); font-family: var(--font-mono); }
+
+  .toc { background: var(--bg-surface); border: 1px solid var(--border); border-radius: var(--radius); padding: 1.5rem 2rem; margin-bottom: 3rem; }
+  .toc h2 { font-size: 0.75rem; text-transform: uppercase; letter-spacing: 0.1em; color: var(--fg-muted); margin-bottom: 1rem; }
+  .toc ol { list-style: none; counter-reset: toc; columns: 2; column-gap: 2rem; }
+  .toc li { counter-increment: toc; break-inside: avoid; margin-bottom: 0.35rem; }
+  .toc li::before { content: counter(toc, decimal-leading-zero) " "; color: var(--fg-subtle); font-family: var(--font-mono); font-size: 0.75rem; margin-right: 0.25rem; }
+  .toc a { font-family: var(--font-mono); color: var(--fg); text-decoration: none; font-size: 0.82rem; transition: color 0.15s; }
+  .toc a:hover { color: var(--accent); }
+
+  section { margin-bottom: 4rem; }
+  section + section { padding-top: 1rem; }
+
+  h2 { font-family: var(--font-mono); font-size: 1.3rem; font-weight: 700; color: var(--fg-bright); letter-spacing: -0.01em; margin-bottom: 1.25rem; padding-bottom: 0.5rem; border-bottom: 1px solid var(--border); }
+  h3 { font-family: var(--font-mono); font-size: 1rem; font-weight: 600; color: var(--fg-bright); margin-top: 2rem; margin-bottom: 0.75rem; }
+  h4 { font-family: var(--font-mono); font-size: 0.9rem; font-weight: 600; color: var(--accent); margin-top: 1.5rem; margin-bottom: 0.5rem; }
+
+  p { margin-bottom: 1rem; font-size: 0.95rem; line-height: 1.75; }
+  strong { color: var(--fg-bright); font-weight: 600; }
+  a { color: var(--accent); text-decoration: none; }
+  a:hover { text-decoration: underline; }
+
+  ul, ol { margin-bottom: 1rem; padding-left: 1.5rem; font-size: 0.93rem; line-height: 1.7; }
+  li { margin-bottom: 0.35rem; }
+  li::marker { color: var(--fg-subtle); }
+
+  .table-wrap { overflow-x: auto; margin-bottom: 1.5rem; }
+  table { width: 100%; border-collapse: collapse; font-size: 0.88rem; }
+  th, td { text-align: left; padding: 0.6rem 1rem; border-bottom: 1px solid var(--border-subtle); }
+  th { font-family: var(--font-mono); font-size: 0.72rem; text-transform: uppercase; letter-spacing: 0.06em; color: var(--fg-muted); background: var(--bg-surface); border-bottom-color: var(--border); white-space: nowrap; }
+  td { font-family: var(--font-sans); font-size: 0.88rem; color: var(--fg); }
+  tr:hover td { background: var(--accent-glow); }
+  td code { background: var(--bg-elevated); padding: 0.15em 0.4em; border-radius: 3px; font-family: var(--font-mono); font-size: 0.82em; color: var(--cyan); }
+
+  pre { background: var(--bg-code); border: 1px solid var(--border); border-radius: var(--radius); padding: 1.25rem 1.5rem; overflow-x: auto; margin-bottom: 1.5rem; font-family: var(--font-mono); font-size: 0.82rem; line-height: 1.65; color: var(--fg); }
+  pre code { background: none; padding: 0; color: inherit; font-size: inherit; }
+  code { font-family: var(--font-mono); font-size: 0.85em; }
+  p code, li code { background: var(--bg-elevated); padding: 0.15em 0.4em; border-radius: 3px; color: var(--cyan); font-size: 0.85em; }
+
+  .kw { color: var(--purple); }
+  .str { color: var(--green); }
+  .cm { color: var(--fg-subtle); font-style: italic; }
+  .num { color: var(--orange); }
+  .key { color: var(--accent); }
+
+  .mermaid { margin: 1.5rem 0 2rem; text-align: center; }
+  .mermaid svg { max-width: 100%; height: auto; }
+
+  .callout { font-family: var(--font-sans); background: var(--bg-surface); border-left: 3px solid var(--accent-dim); border-radius: 0 var(--radius) var(--radius) 0; padding: 1rem 1.25rem; margin-bottom: 1.5rem; font-size: 0.88rem; color: var(--fg-muted); line-height: 1.6; }
+  .callout strong { font-family: var(--font-mono); color: var(--fg-bright); }
+  .callout.success { border-left-color: var(--green-dim); }
+  .callout.warn { border-left-color: var(--orange); }
+
+  .badge { display: inline-block; font-family: var(--font-mono); font-size: 0.65rem; font-weight: 600; text-transform: uppercase; letter-spacing: 0.05em; padding: 0.2em 0.6em; border-radius: 3px; vertical-align: middle; margin-left: 0.4rem; }
+  .badge-done { background: var(--green-dim); color: #fff; }
+  .badge-wip { background: var(--orange); color: #0b0e14; }
+  .badge-todo { background: var(--fg-subtle); color: var(--fg); }
+
+  .checklist { list-style: none; padding-left: 0; }
+  .checklist li { padding-left: 1.5rem; position: relative; margin-bottom: 0.5rem; }
+  .checklist li::before { position: absolute; left: 0; font-family: var(--font-mono); font-size: 0.85rem; }
+  .checklist li.done { color: var(--fg-muted); }
+  .checklist li.done::before { content: "\2713"; color: var(--green); }
+  .checklist li.todo::before { content: "\25CB"; color: var(--fg-subtle); }
+  .checklist li.wip::before { content: "\25D4"; color: var(--orange); }
+
+  .compare { display: grid; grid-template-columns: 1fr 1fr; gap: 1rem; margin-bottom: 2rem; }
+  .compare-card { background: var(--bg-surface); border: 1px solid var(--border); border-radius: var(--radius); padding: 1.25rem; }
+  .compare-card h4 { margin-top: 0; font-size: 0.82rem; }
+  .compare-card.after { border-color: var(--accent-dim); }
+  .compare-card ul { font-family: var(--font-mono); padding-left: 1.25rem; font-size: 0.8rem; }
+
+  hr { border: none; border-top: 1px solid var(--border); margin: 3rem 0; }
+
+  .progress-bar { position: fixed; top: 0; left: 0; height: 2px; background: var(--accent); z-index: 999; transition: width 0.1s linear; }
+
+  @media (max-width: 640px) {
+    .container { padding: 2rem 1rem 4rem; }
+    .hero h1 { font-size: 1.6rem; }
+    .toc ol { columns: 1; }
+    .compare { grid-template-columns: 1fr; }
+    table { font-size: 0.8rem; }
+    th, td { padding: 0.4rem 0.6rem; }
+  }
+</style>
+<link rel="preconnect" href="https://fonts.googleapis.com">
+<link href="https://fonts.googleapis.com/css2?family=Noto+Emoji&display=swap" rel="stylesheet">
+<style>
+  @font-face {
+    font-family: 'Departure Mono';
+    src: url('https://cdn.jsdelivr.net/gh/rektdeckard/departure-mono@latest/fonts/DepartureMono-Regular.woff2') format('woff2');
+    font-weight: normal;
+    font-style: normal;
+    font-display: swap;
+  }
+</style>
+</head>
+<body>
+
+<div class="progress-bar" id="progress"></div>
+
+<div class="container">
+
+<header class="hero">
+  <h1>honcho<span>-integration-spec</span></h1>
+  <p class="subtitle">Comparison of Hermes Agent vs. openclaw-honcho — and a porting spec for bringing Hermes patterns into other Honcho integrations.</p>
+  <div class="meta">
+    <span>hermes-agent / openclaw-honcho</span>
+    <span>Python + TypeScript</span>
+    <span>2026-03-09</span>
+  </div>
+</header>
+
+<nav class="toc">
+  <h2>Contents</h2>
+  <ol>
+    <li><a href="#overview">Overview</a></li>
+    <li><a href="#architecture">Architecture comparison</a></li>
+    <li><a href="#diff-table">Diff table</a></li>
+    <li><a href="#patterns">Hermes patterns to port</a></li>
+    <li><a href="#spec-async">Spec: async prefetch</a></li>
+    <li><a href="#spec-reasoning">Spec: dynamic reasoning level</a></li>
+    <li><a href="#spec-modes">Spec: per-peer memory modes</a></li>
+    <li><a href="#spec-identity">Spec: AI peer identity formation</a></li>
+    <li><a href="#spec-sessions">Spec: session naming strategies</a></li>
+    <li><a href="#spec-cli">Spec: CLI surface injection</a></li>
+    <li><a href="#openclaw-checklist">openclaw-honcho checklist</a></li>
+    <li><a href="#nanobot-checklist">nanobot-honcho checklist</a></li>
+  </ol>
+</nav>
+
+<!-- OVERVIEW -->
+<section id="overview">
+  <h2>Overview</h2>
+
+  <p>Two independent Honcho integrations have been built for two different agent runtimes: <strong>Hermes Agent</strong> (Python, baked into the runner) and <strong>openclaw-honcho</strong> (TypeScript plugin via hook/tool API). Both use the same Honcho peer paradigm — dual peer model, <code>session.context()</code>, <code>peer.chat()</code> — but they made different tradeoffs at every layer.</p>
+
+  <p>This document maps those tradeoffs and defines a porting spec: a set of Hermes-originated patterns, each stated as an integration-agnostic interface, that any Honcho integration can adopt regardless of runtime or language.</p>
+
+  <div class="callout">
+    <strong>Scope</strong> Both integrations work correctly today. This spec is about the delta — patterns in Hermes that are worth propagating and patterns in openclaw-honcho that Hermes should eventually adopt. The spec is additive, not prescriptive.
+  </div>
+</section>
+
+<!-- ARCHITECTURE -->
+<section id="architecture">
+  <h2>Architecture comparison</h2>
+
+  <h3>Hermes: baked-in runner</h3>
+  <p>Honcho is initialised directly inside <code>AIAgent.__init__</code>. There is no plugin boundary. Session management, context injection, async prefetch, and CLI surface are all first-class concerns of the runner. Context is injected once per session (baked into <code>_cached_system_prompt</code>) and never re-fetched mid-session — this maximises prefix cache hits at the LLM provider.</p>
+
+  <div class="mermaid">
+%%{init: {'theme': 'dark', 'themeVariables': { 'primaryColor': '#1f3150', 'primaryTextColor': '#c9d1d9', 'primaryBorderColor': '#3d6ea5', 'lineColor': '#3d6ea5', 'secondaryColor': '#162030', 'tertiaryColor': '#11151c' }}}%%
+flowchart TD
+    U["user message"] --> P["_honcho_prefetch()<br/>(reads cache — no HTTP)"]
+    P --> SP["_build_system_prompt()<br/>(first turn only, cached)"]
+    SP --> LLM["LLM call"]
+    LLM --> R["response"]
+    R --> FP["_honcho_fire_prefetch()<br/>(daemon threads, turn end)"]
+    FP --> C1["prefetch_context() thread"]
+    FP --> C2["prefetch_dialectic() thread"]
+    C1 --> CACHE["_context_cache / _dialectic_cache"]
+    C2 --> CACHE
+
+    style U fill:#162030,stroke:#3d6ea5,color:#c9d1d9
+    style P fill:#1f3150,stroke:#3d6ea5,color:#c9d1d9
+    style SP fill:#1f3150,stroke:#3d6ea5,color:#c9d1d9
+    style LLM fill:#162030,stroke:#3d6ea5,color:#c9d1d9
+    style R fill:#162030,stroke:#3d6ea5,color:#c9d1d9
+    style FP fill:#2a1a40,stroke:#bc8cff,color:#c9d1d9
+    style C1 fill:#2a1a40,stroke:#bc8cff,color:#c9d1d9
+    style C2 fill:#2a1a40,stroke:#bc8cff,color:#c9d1d9
+    style CACHE fill:#11151c,stroke:#484f58,color:#6e7681
+  </div>
+
+  <h3>openclaw-honcho: hook-based plugin</h3>
+  <p>The plugin registers hooks against OpenClaw's event bus. Context is fetched synchronously inside <code>before_prompt_build</code> on every turn. Message capture happens in <code>agent_end</code>. The multi-agent hierarchy is tracked via <code>subagent_spawned</code>. This model is correct but every turn pays a blocking Honcho round-trip before the LLM call can begin.</p>
+
+  <div class="mermaid">
+%%{init: {'theme': 'dark', 'themeVariables': { 'primaryColor': '#1f3150', 'primaryTextColor': '#c9d1d9', 'primaryBorderColor': '#3d6ea5', 'lineColor': '#3d6ea5', 'secondaryColor': '#162030', 'tertiaryColor': '#11151c' }}}%%
+flowchart TD
+    U2["user message"] --> BPB["before_prompt_build<br/>(BLOCKING HTTP — every turn)"]
+    BPB --> CTX["session.context()"]
+    CTX --> SP2["system prompt assembled"]
+    SP2 --> LLM2["LLM call"]
+    LLM2 --> R2["response"]
+    R2 --> AE["agent_end hook"]
+    AE --> SAVE["session.addMessages()<br/>session.setMetadata()"]
+
+    style U2 fill:#162030,stroke:#3d6ea5,color:#c9d1d9
+    style BPB fill:#3a1515,stroke:#f47067,color:#c9d1d9
+    style CTX fill:#3a1515,stroke:#f47067,color:#c9d1d9
+    style SP2 fill:#1f3150,stroke:#3d6ea5,color:#c9d1d9
+    style LLM2 fill:#162030,stroke:#3d6ea5,color:#c9d1d9
+    style R2 fill:#162030,stroke:#3d6ea5,color:#c9d1d9
+    style AE fill:#162030,stroke:#3d6ea5,color:#c9d1d9
+    style SAVE fill:#11151c,stroke:#484f58,color:#6e7681
+  </div>
+</section>
+
+<!-- DIFF TABLE -->
+<section id="diff-table">
+  <h2>Diff table</h2>
+
+  <div class="table-wrap">
+    <table>
+      <thead>
+        <tr>
+          <th>Dimension</th>
+          <th>Hermes Agent</th>
+          <th>openclaw-honcho</th>
+        </tr>
+      </thead>
+      <tbody>
+        <tr>
+          <td><strong>Context injection timing</strong></td>
+          <td>Once per session (cached). Zero HTTP on response path after turn 1.</td>
+          <td>Every turn, blocking. Fresh context per turn but adds latency.</td>
+        </tr>
+        <tr>
+          <td><strong>Prefetch strategy</strong></td>
+          <td>Daemon threads fire at turn end; consumed next turn from cache.</td>
+          <td>None. Blocking call at prompt-build time.</td>
+        </tr>
+        <tr>
+          <td><strong>Dialectic (peer.chat)</strong></td>
+          <td>Prefetched async; result injected into system prompt next turn.</td>
+          <td>On-demand via <code>honcho_recall</code> / <code>honcho_analyze</code> tools.</td>
+        </tr>
+        <tr>
+          <td><strong>Reasoning level</strong></td>
+          <td>Dynamic: scales with message length. Floor = config default. Cap = "high".</td>
+          <td>Fixed per tool: recall=minimal, analyze=medium.</td>
+        </tr>
+        <tr>
+          <td><strong>Memory modes</strong></td>
+          <td><code>user_memory_mode</code> / <code>agent_memory_mode</code>: hybrid / honcho / local.</td>
+          <td>None. Always writes to Honcho.</td>
+        </tr>
+        <tr>
+          <td><strong>Write frequency</strong></td>
+          <td>async (background queue), turn, session, N turns.</td>
+          <td>After every agent_end (no control).</td>
+        </tr>
+        <tr>
+          <td><strong>AI peer identity</strong></td>
+          <td><code>observe_me=True</code>, <code>seed_ai_identity()</code>, <code>get_ai_representation()</code>, SOUL.md → AI peer.</td>
+          <td>Agent files uploaded to agent peer at setup. No ongoing self-observation seeding.</td>
+        </tr>
+        <tr>
+          <td><strong>Context scope</strong></td>
+          <td>User peer + AI peer representation, both injected.</td>
+          <td>User peer (owner) representation + conversation summary. <code>peerPerspective</code> on context call.</td>
+        </tr>
+        <tr>
+          <td><strong>Session naming</strong></td>
+          <td>per-directory / global / manual map / title-based.</td>
+          <td>Derived from platform session key.</td>
+        </tr>
+        <tr>
+          <td><strong>Multi-agent</strong></td>
+          <td>Single-agent only.</td>
+          <td>Parent observer hierarchy via <code>subagent_spawned</code>.</td>
+        </tr>
+        <tr>
+          <td><strong>Tool surface</strong></td>
+          <td>Single <code>query_user_context</code> tool (on-demand dialectic).</td>
+          <td>6 tools: session, profile, search, context (fast) + recall, analyze (LLM).</td>
+        </tr>
+        <tr>
+          <td><strong>Platform metadata</strong></td>
+          <td>Not stripped.</td>
+          <td>Explicitly stripped before Honcho storage.</td>
+        </tr>
+        <tr>
+          <td><strong>Message dedup</strong></td>
+          <td>None (sends on every save cycle).</td>
+          <td><code>lastSavedIndex</code> in session metadata prevents re-sending.</td>
+        </tr>
+        <tr>
+          <td><strong>CLI surface in prompt</strong></td>
+          <td>Management commands injected into system prompt. Agent knows its own CLI.</td>
+          <td>Not injected.</td>
+        </tr>
+        <tr>
+          <td><strong>AI peer name in identity</strong></td>
+          <td>Replaces "Hermes Agent" in DEFAULT_AGENT_IDENTITY when configured.</td>
+          <td>Not implemented.</td>
+        </tr>
+        <tr>
+          <td><strong>QMD / local file search</strong></td>
+          <td>Not implemented.</td>
+          <td>Passthrough tools when QMD backend configured.</td>
+        </tr>
+        <tr>
+          <td><strong>Workspace metadata</strong></td>
+          <td>Not implemented.</td>
+          <td><code>agentPeerMap</code> in workspace metadata tracks agent&#8594;peer ID.</td>
+        </tr>
+      </tbody>
+    </table>
+  </div>
+</section>
+
+<!-- PATTERNS -->
+<section id="patterns">
+  <h2>Hermes patterns to port</h2>
+
+  <p>Six patterns from Hermes are worth adopting in any Honcho integration. They are described below as integration-agnostic interfaces — the implementation will differ per runtime, but the contract is the same.</p>
+
+  <div class="compare">
+    <div class="compare-card">
+      <h4>Patterns Hermes contributes</h4>
+      <ul>
+        <li>Async prefetch (zero-latency)</li>
+        <li>Dynamic reasoning level</li>
+        <li>Per-peer memory modes</li>
+        <li>AI peer identity formation</li>
+        <li>Session naming strategies</li>
+        <li>CLI surface injection</li>
+      </ul>
+    </div>
+    <div class="compare-card after">
+      <h4>Patterns openclaw contributes back</h4>
+      <ul>
+        <li>lastSavedIndex dedup</li>
+        <li>Platform metadata stripping</li>
+        <li>Multi-agent observer hierarchy</li>
+        <li>peerPerspective on context()</li>
+        <li>Tiered tool surface (fast/LLM)</li>
+        <li>Workspace agentPeerMap</li>
+      </ul>
+    </div>
+  </div>
+</section>
+
+<!-- SPEC: ASYNC PREFETCH -->
+<section id="spec-async">
+  <h2>Spec: async prefetch</h2>
+
+  <h3>Problem</h3>
+  <p>Calling <code>session.context()</code> and <code>peer.chat()</code> synchronously before each LLM call adds 200–800ms of Honcho round-trip latency to every turn. Users experience this as the agent "thinking slowly."</p>
+
+  <h3>Pattern</h3>
+  <p>Fire both calls as non-blocking background work at the <strong>end</strong> of each turn. Store results in a per-session cache keyed by session ID. At the <strong>start</strong> of the next turn, pop from cache — the HTTP is already done. First turn is cold (empty cache); all subsequent turns are zero-latency on the response path.</p>
+
+  <h3>Interface contract</h3>
+  <pre><code><span class="cm">// TypeScript (openclaw / nanobot plugin shape)</span>
+
+<span class="kw">interface</span> <span class="key">AsyncPrefetch</span> {
+  <span class="cm">// Fire context + dialectic fetches at turn end. Non-blocking.</span>
+  firePrefetch(sessionId: <span class="str">string</span>, userMessage: <span class="str">string</span>): <span class="kw">void</span>;
+
+  <span class="cm">// Pop cached results at turn start. Returns empty if cache is cold.</span>
+  popContextResult(sessionId: <span class="str">string</span>): ContextResult | <span class="kw">null</span>;
+  popDialecticResult(sessionId: <span class="str">string</span>): <span class="str">string</span> | <span class="kw">null</span>;
+}
+
+<span class="kw">type</span> <span class="key">ContextResult</span> = {
+  representation: <span class="str">string</span>;
+  card: <span class="str">string</span>[];
+  aiRepresentation?: <span class="str">string</span>;  <span class="cm">// AI peer context if enabled</span>
+  summary?: <span class="str">string</span>;            <span class="cm">// conversation summary if fetched</span>
+};</code></pre>
+
+  <h3>Implementation notes</h3>
+  <ul>
+    <li>Python: <code>threading.Thread(daemon=True)</code>. Write to <code>dict[session_id, result]</code> — GIL makes this safe for simple writes.</li>
+    <li>TypeScript: <code>Promise</code> stored in <code>Map&lt;string, Promise&lt;ContextResult&gt;&gt;</code>. Await at pop time. If not resolved yet, skip (return null) — do not block.</li>
+    <li>The pop is destructive: clears the cache entry after reading so stale data never accumulates.</li>
+    <li>Prefetch should also fire on first turn (even though it won't be consumed until turn 2) — this ensures turn 2 is never cold.</li>
+  </ul>
+
+  <h3>openclaw-honcho adoption</h3>
+  <p>Move <code>session.context()</code> from <code>before_prompt_build</code> to a post-<code>agent_end</code> background task. Store result in <code>state.contextCache</code>. In <code>before_prompt_build</code>, read from cache instead of calling Honcho. If cache is empty (turn 1), inject nothing — the prompt is still valid without Honcho context on the first turn.</p>
+</section>
+
+<!-- SPEC: DYNAMIC REASONING LEVEL -->
+<section id="spec-reasoning">
+  <h2>Spec: dynamic reasoning level</h2>
+
+  <h3>Problem</h3>
+  <p>Honcho's dialectic endpoint supports reasoning levels from <code>minimal</code> to <code>max</code>. A fixed level per tool wastes budget on simple queries and under-serves complex ones.</p>
+
+  <h3>Pattern</h3>
+  <p>Select the reasoning level dynamically based on the user's message. Use the configured default as a floor. Bump by message length. Cap auto-selection at <code>high</code> — never select <code>max</code> automatically.</p>
+
+  <h3>Interface contract</h3>
+  <pre><code><span class="cm">// Shared helper — identical logic in any language</span>
+
+<span class="kw">const</span> LEVELS = [<span class="str">"minimal"</span>, <span class="str">"low"</span>, <span class="str">"medium"</span>, <span class="str">"high"</span>, <span class="str">"max"</span>];
+
+<span class="kw">function</span> <span class="key">dynamicReasoningLevel</span>(
+  query: <span class="str">string</span>,
+  configDefault: <span class="str">string</span> = <span class="str">"low"</span>
+): <span class="str">string</span> {
+  <span class="kw">const</span> baseIdx = Math.max(<span class="num">0</span>, LEVELS.indexOf(configDefault));
+  <span class="kw">const</span> n = query.length;
+  <span class="kw">const</span> bump = n &lt; <span class="num">120</span> ? <span class="num">0</span> : n &lt; <span class="num">400</span> ? <span class="num">1</span> : <span class="num">2</span>;
+  <span class="kw">return</span> LEVELS[Math.min(baseIdx + bump, <span class="num">3</span>)]; <span class="cm">// cap at "high" (idx 3)</span>
+}</code></pre>
+
+  <h3>Config key</h3>
+  <p>Add a <code>dialecticReasoningLevel</code> config field (string, default <code>"low"</code>). This sets the floor. Users can raise or lower it. The dynamic bump always applies on top.</p>
+
+  <h3>openclaw-honcho adoption</h3>
+  <p>Apply in <code>honcho_recall</code> and <code>honcho_analyze</code>: replace the fixed <code>reasoningLevel</code> with the dynamic selector. <code>honcho_recall</code> should use floor <code>"minimal"</code> and <code>honcho_analyze</code> floor <code>"medium"</code> — both still bump with message length.</p>
+</section>
+
+<!-- SPEC: PER-PEER MEMORY MODES -->
+<section id="spec-modes">
+  <h2>Spec: per-peer memory modes</h2>
+
+  <h3>Problem</h3>
+  <p>Users want independent control over whether user context and agent context are written locally, to Honcho, or both. A single <code>memoryMode</code> shorthand is not granular enough.</p>
+
+  <h3>Pattern</h3>
+  <p>Three modes per peer: <code>hybrid</code> (write both local + Honcho), <code>honcho</code> (Honcho only, disable local files), <code>local</code> (local files only, skip Honcho sync for this peer). Two orthogonal axes: user peer and agent peer.</p>
+
+  <h3>Config schema</h3>
+  <pre><code><span class="cm">// ~/.openclaw/openclaw.json  (or ~/.nanobot/config.json)</span>
+{
+  <span class="str">"plugins"</span>: {
+    <span class="str">"openclaw-honcho"</span>: {
+      <span class="str">"config"</span>: {
+        <span class="str">"apiKey"</span>: <span class="str">"..."</span>,
+        <span class="str">"memoryMode"</span>: <span class="str">"hybrid"</span>,          <span class="cm">// shorthand: both peers</span>
+        <span class="str">"userMemoryMode"</span>: <span class="str">"honcho"</span>,       <span class="cm">// override for user peer</span>
+        <span class="str">"agentMemoryMode"</span>: <span class="str">"hybrid"</span>       <span class="cm">// override for agent peer</span>
+      }
+    }
+  }
+}</code></pre>
+
+  <h3>Resolution order</h3>
+  <ol>
+    <li>Per-peer field (<code>userMemoryMode</code> / <code>agentMemoryMode</code>) — wins if present.</li>
+    <li>Shorthand <code>memoryMode</code> — applies to both peers as default.</li>
+    <li>Hardcoded default: <code>"hybrid"</code>.</li>
+  </ol>
+
+  <h3>Effect on Honcho sync</h3>
+  <ul>
+    <li><code>userMemoryMode=local</code>: skip adding user peer messages to Honcho.</li>
+    <li><code>agentMemoryMode=local</code>: skip adding assistant peer messages to Honcho.</li>
+    <li>Both local: skip <code>session.addMessages()</code> entirely.</li>
+    <li><code>userMemoryMode=honcho</code>: disable local USER.md writes.</li>
+    <li><code>agentMemoryMode=honcho</code>: disable local MEMORY.md / SOUL.md writes.</li>
+  </ul>
+</section>
+
+<!-- SPEC: AI PEER IDENTITY -->
+<section id="spec-identity">
+  <h2>Spec: AI peer identity formation</h2>
+
+  <h3>Problem</h3>
+  <p>Honcho builds the user's representation organically by observing what the user says. The same mechanism exists for the AI peer — but only if <code>observe_me=True</code> is set for the agent peer. Without it, the agent peer accumulates nothing and Honcho's AI-side model never forms.</p>
+
+  <p>Additionally, existing persona files (SOUL.md, IDENTITY.md) should seed the AI peer's Honcho representation at first activation, rather than waiting for it to emerge from scratch.</p>
+
+  <h3>Part A: observe_me=True for agent peer</h3>
+  <pre><code><span class="cm">// TypeScript — in session.addPeers() call</span>
+<span class="kw">await</span> session.addPeers([
+  [ownerPeer.id, { observeMe: <span class="kw">true</span>,  observeOthers: <span class="kw">false</span> }],
+  [agentPeer.id, { observeMe: <span class="kw">true</span>,  observeOthers: <span class="kw">true</span>  }], <span class="cm">// was false</span>
+]);</code></pre>
+
+  <p>This is a one-line change but foundational. Without it, Honcho's AI peer representation stays empty regardless of what the agent says.</p>
+
+  <h3>Part B: seedAiIdentity()</h3>
+  <pre><code><span class="kw">async function</span> <span class="key">seedAiIdentity</span>(
+  session: HonchoSession,
+  agentPeer: Peer,
+  content: <span class="str">string</span>,
+  source: <span class="str">string</span>
+): Promise&lt;<span class="kw">boolean</span>&gt; {
+  <span class="kw">const</span> wrapped = [
+    <span class="str">`&lt;ai_identity_seed&gt;`</span>,
+    <span class="str">`&lt;source&gt;${source}&lt;/source&gt;`</span>,
+    <span class="str">``</span>,
+    content.trim(),
+    <span class="str">`&lt;/ai_identity_seed&gt;`</span>,
+  ].join(<span class="str">"\n"</span>);
+
+  <span class="kw">await</span> agentPeer.addMessage(<span class="str">"assistant"</span>, wrapped);
+  <span class="kw">return true</span>;
+}</code></pre>
+
+  <h3>Part C: migrate agent files at setup</h3>
+  <p>During <code>openclaw honcho setup</code>, upload agent-self files (SOUL.md, IDENTITY.md, AGENTS.md, BOOTSTRAP.md) to the agent peer using <code>seedAiIdentity()</code> instead of <code>session.uploadFile()</code>. This routes the content through Honcho's observation pipeline rather than the file store.</p>
+
+  <h3>Part D: AI peer name in identity</h3>
+  <p>When the agent has a configured name (non-default), inject it into the agent's self-identity prefix. In OpenClaw this means adding to the injected system prompt section:</p>
+  <pre><code><span class="cm">// In context hook return value</span>
+<span class="kw">return</span> {
+  systemPrompt: [
+    agentName ? <span class="str">`You are ${agentName}.`</span> : <span class="str">""</span>,
+    <span class="str">"## User Memory Context"</span>,
+    ...sections,
+  ].filter(Boolean).join(<span class="str">"\n\n"</span>)
+};</code></pre>
+
+  <h3>CLI surface: honcho identity subcommand</h3>
+  <pre><code>openclaw honcho identity &lt;file&gt;    <span class="cm"># seed from file</span>
+openclaw honcho identity --show    <span class="cm"># show current AI peer representation</span></code></pre>
+</section>
+
+<!-- SPEC: SESSION NAMING -->
+<section id="spec-sessions">
+  <h2>Spec: session naming strategies</h2>
+
+  <h3>Problem</h3>
+  <p>When Honcho is used across multiple projects or directories, a single global session means every project shares the same context. Per-directory sessions provide isolation without requiring users to name sessions manually.</p>
+
+  <h3>Strategies</h3>
+  <div class="table-wrap">
+    <table>
+      <thead><tr><th>Strategy</th><th>Session key</th><th>When to use</th></tr></thead>
+      <tbody>
+        <tr><td><code>per-directory</code></td><td>basename of CWD</td><td>Default. Each project gets its own session.</td></tr>
+        <tr><td><code>global</code></td><td>fixed string <code>"global"</code></td><td>Single cross-project session.</td></tr>
+        <tr><td>manual map</td><td>user-configured per path</td><td><code>sessions</code> config map overrides directory basename.</td></tr>
+        <tr><td>title-based</td><td>sanitized session title</td><td>When agent supports named sessions; title set mid-conversation.</td></tr>
+      </tbody>
+    </table>
+  </div>
+
+  <h3>Config schema</h3>
+  <pre><code>{
+  <span class="str">"sessionStrategy"</span>: <span class="str">"per-directory"</span>,   <span class="cm">// "per-directory" | "global"</span>
+  <span class="str">"sessionPeerPrefix"</span>: <span class="kw">false</span>,            <span class="cm">// prepend peer name to session key</span>
+  <span class="str">"sessions"</span>: {                            <span class="cm">// manual overrides</span>
+    <span class="str">"/home/user/projects/foo"</span>: <span class="str">"foo-project"</span>
+  }
+}</code></pre>
+
+  <h3>CLI surface</h3>
+  <pre><code>openclaw honcho sessions              <span class="cm"># list all mappings</span>
+openclaw honcho map &lt;name&gt;           <span class="cm"># map cwd to session name</span>
+openclaw honcho map                   <span class="cm"># no-arg = list mappings</span></code></pre>
+
+  <p>Resolution order: manual map wins &rarr; session title &rarr; directory basename &rarr; platform key.</p>
+</section>
+
+<!-- SPEC: CLI SURFACE INJECTION -->
+<section id="spec-cli">
+  <h2>Spec: CLI surface injection</h2>
+
+  <h3>Problem</h3>
+  <p>When a user asks "how do I change my memory settings?" or "what Honcho commands are available?" the agent either hallucinates or says it doesn't know. The agent should know its own management interface.</p>
+
+  <h3>Pattern</h3>
+  <p>When Honcho is active, append a compact command reference to the system prompt. The agent can cite these commands directly instead of guessing.</p>
+
+  <pre><code><span class="cm">// In context hook, append to systemPrompt</span>
+<span class="kw">const</span> honchoSection = [
+  <span class="str">"# Honcho memory integration"</span>,
+  <span class="str">`Active. Session: ${sessionKey}. Mode: ${mode}.`</span>,
+  <span class="str">"Management commands:"</span>,
+  <span class="str">"  openclaw honcho status                    — show config + connection"</span>,
+  <span class="str">"  openclaw honcho mode [hybrid|honcho|local] — show or set memory mode"</span>,
+  <span class="str">"  openclaw honcho sessions                  — list session mappings"</span>,
+  <span class="str">"  openclaw honcho map &lt;name&gt;                — map directory to session"</span>,
+  <span class="str">"  openclaw honcho identity [file] [--show]  — seed or show AI identity"</span>,
+  <span class="str">"  openclaw honcho setup                     — full interactive wizard"</span>,
+].join(<span class="str">"\n"</span>);</code></pre>
+
+  <div class="callout warn">
+    <strong>Keep it compact.</strong> This section is injected every turn. Keep it under 300 chars of context. List commands, not explanations — the agent can explain them on request.
+  </div>
+</section>
+
+<!-- OPENCLAW CHECKLIST -->
+<section id="openclaw-checklist">
+  <h2>openclaw-honcho checklist</h2>
+
+  <p>Ordered by impact. Each item maps to a spec section above.</p>
+
+  <ul class="checklist">
+    <li class="todo"><strong>Async prefetch</strong> — move <code>session.context()</code> out of <code>before_prompt_build</code> into post-<code>agent_end</code> background Promise. Pop from cache at prompt build. (<a href="#spec-async">spec</a>)</li>
+    <li class="todo"><strong>observe_me=True for agent peer</strong> — one-line change in <code>session.addPeers()</code> config for agent peer. (<a href="#spec-identity">spec</a>)</li>
+    <li class="todo"><strong>Dynamic reasoning level</strong> — add <code>dynamicReasoningLevel()</code> helper; apply in <code>honcho_recall</code> and <code>honcho_analyze</code>. Add <code>dialecticReasoningLevel</code> to config schema. (<a href="#spec-reasoning">spec</a>)</li>
+    <li class="todo"><strong>Per-peer memory modes</strong> — add <code>userMemoryMode</code> / <code>agentMemoryMode</code> to config; gate Honcho sync and local writes accordingly. (<a href="#spec-modes">spec</a>)</li>
+    <li class="todo"><strong>seedAiIdentity()</strong> — add helper; apply during setup migration for SOUL.md / IDENTITY.md instead of <code>session.uploadFile()</code>. (<a href="#spec-identity">spec</a>)</li>
+    <li class="todo"><strong>Session naming strategies</strong> — add <code>sessionStrategy</code>, <code>sessions</code> map, <code>sessionPeerPrefix</code> to config; implement resolution function. (<a href="#spec-sessions">spec</a>)</li>
+    <li class="todo"><strong>CLI surface injection</strong> — append command reference to <code>before_prompt_build</code> return value when Honcho is active. (<a href="#spec-cli">spec</a>)</li>
+    <li class="todo"><strong>honcho identity subcommand</strong> — add <code>openclaw honcho identity</code> CLI command. (<a href="#spec-identity">spec</a>)</li>
+    <li class="todo"><strong>AI peer name injection</strong> — if <code>aiPeer</code> name configured, prepend to injected system prompt. (<a href="#spec-identity">spec</a>)</li>
+    <li class="todo"><strong>honcho mode / honcho sessions / honcho map</strong> — CLI parity with Hermes. (<a href="#spec-sessions">spec</a>)</li>
+  </ul>
+
+  <div class="callout success">
+    <strong>Already done in openclaw-honcho (do not re-implement):</strong> lastSavedIndex dedup, platform metadata stripping, multi-agent parent observer hierarchy, peerPerspective on context(), tiered tool surface (fast/LLM), workspace agentPeerMap, QMD passthrough, self-hosted Honcho support.
+  </div>
+</section>
+
+<!-- NANOBOT CHECKLIST -->
+<section id="nanobot-checklist">
+  <h2>nanobot-honcho checklist</h2>
+
+  <p>nanobot-honcho is a greenfield integration. Start from openclaw-honcho's architecture (hook-based, dual peer) and apply all Hermes patterns from day one rather than retrofitting. Priority order:</p>
+
+  <h3>Phase 1 — core correctness</h3>
+  <ul class="checklist">
+    <li class="todo">Dual peer model (owner + agent peer), both with <code>observe_me=True</code></li>
+    <li class="todo">Message capture at turn end with <code>lastSavedIndex</code> dedup</li>
+    <li class="todo">Platform metadata stripping before Honcho storage</li>
+    <li class="todo">Async prefetch from day one — do not implement blocking context injection</li>
+    <li class="todo">Legacy file migration at first activation (USER.md → owner peer, SOUL.md → <code>seedAiIdentity()</code>)</li>
+  </ul>
+
+  <h3>Phase 2 — configuration</h3>
+  <ul class="checklist">
+    <li class="todo">Config schema: <code>apiKey</code>, <code>workspaceId</code>, <code>baseUrl</code>, <code>memoryMode</code>, <code>userMemoryMode</code>, <code>agentMemoryMode</code>, <code>dialecticReasoningLevel</code>, <code>sessionStrategy</code>, <code>sessions</code></li>
+    <li class="todo">Per-peer memory mode gating</li>
+    <li class="todo">Dynamic reasoning level</li>
+    <li class="todo">Session naming strategies</li>
+  </ul>
+
+  <h3>Phase 3 — tools and CLI</h3>
+  <ul class="checklist">
+    <li class="todo">Tool surface: <code>honcho_profile</code>, <code>honcho_recall</code>, <code>honcho_analyze</code>, <code>honcho_search</code>, <code>honcho_context</code></li>
+    <li class="todo">CLI: <code>setup</code>, <code>status</code>, <code>sessions</code>, <code>map</code>, <code>mode</code>, <code>identity</code></li>
+    <li class="todo">CLI surface injection into system prompt</li>
+    <li class="todo">AI peer name wired into agent identity</li>
+  </ul>
+</section>
+
+</div>
+
+<script type="module">
+  import mermaid from 'https://cdn.jsdelivr.net/npm/mermaid@11/dist/mermaid.esm.min.mjs';
+  mermaid.initialize({ startOnLoad: true, securityLevel: 'loose', fontFamily: 'Departure Mono, Noto Emoji, monospace' });
+</script>
+<script>
+  window.addEventListener('scroll', () => {
+    const bar = document.getElementById('progress');
+    const max = document.documentElement.scrollHeight - window.innerHeight;
+    bar.style.width = (max > 0 ? (window.scrollY / max) * 100 : 0) + '%';
+  });
+</script>
+</body>
+</html>
@@ -0,0 +1,377 @@
+# honcho-integration-spec
+
+Comparison of Hermes Agent vs. openclaw-honcho — and a porting spec for bringing Hermes patterns into other Honcho integrations.
+
+---
+
+## Overview
+
+Two independent Honcho integrations have been built for two different agent runtimes: **Hermes Agent** (Python, baked into the runner) and **openclaw-honcho** (TypeScript plugin via hook/tool API). Both use the same Honcho peer paradigm — dual peer model, `session.context()`, `peer.chat()` — but they made different tradeoffs at every layer.
+
+This document maps those tradeoffs and defines a porting spec: a set of Hermes-originated patterns, each stated as an integration-agnostic interface, that any Honcho integration can adopt regardless of runtime or language.
+
+> **Scope** Both integrations work correctly today. This spec is about the delta — patterns in Hermes that are worth propagating and patterns in openclaw-honcho that Hermes should eventually adopt. The spec is additive, not prescriptive.
+
+---
+
+## Architecture comparison
+
+### Hermes: baked-in runner
+
+Honcho is initialised directly inside `AIAgent.__init__`. There is no plugin boundary. Session management, context injection, async prefetch, and CLI surface are all first-class concerns of the runner. Context is injected once per session (baked into `_cached_system_prompt`) and never re-fetched mid-session — this maximises prefix cache hits at the LLM provider.
+
+Turn flow:
+
+```
+user message
+  → _honcho_prefetch()       (reads cache — no HTTP)
+  → _build_system_prompt()   (first turn only, cached)
+  → LLM call
+  → response
+  → _honcho_fire_prefetch()  (daemon threads, turn end)
+       → prefetch_context() thread  ──┐
+       → prefetch_dialectic() thread ─┴→ _context_cache / _dialectic_cache
+```
+
+### openclaw-honcho: hook-based plugin
+
+The plugin registers hooks against OpenClaw's event bus. Context is fetched synchronously inside `before_prompt_build` on every turn. Message capture happens in `agent_end`. The multi-agent hierarchy is tracked via `subagent_spawned`. This model is correct but every turn pays a blocking Honcho round-trip before the LLM call can begin.
+
+Turn flow:
+
+```
+user message
+  → before_prompt_build (BLOCKING HTTP — every turn)
+       → session.context()
+  → system prompt assembled
+  → LLM call
+  → response
+  → agent_end hook
+       → session.addMessages()
+       → session.setMetadata()
+```
+
+---
+
+## Diff table
+
+| Dimension | Hermes Agent | openclaw-honcho |
+|---|---|---|
+| **Context injection timing** | Once per session (cached). Zero HTTP on response path after turn 1. | Every turn, blocking. Fresh context per turn but adds latency. |
+| **Prefetch strategy** | Daemon threads fire at turn end; consumed next turn from cache. | None. Blocking call at prompt-build time. |
+| **Dialectic (peer.chat)** | Prefetched async; result injected into system prompt next turn. | On-demand via `honcho_recall` / `honcho_analyze` tools. |
+| **Reasoning level** | Dynamic: scales with message length. Floor = config default. Cap = "high". | Fixed per tool: recall=minimal, analyze=medium. |
+| **Memory modes** | `user_memory_mode` / `agent_memory_mode`: hybrid / honcho / local. | None. Always writes to Honcho. |
+| **Write frequency** | async (background queue), turn, session, N turns. | After every agent_end (no control). |
+| **AI peer identity** | `observe_me=True`, `seed_ai_identity()`, `get_ai_representation()`, SOUL.md → AI peer. | Agent files uploaded to agent peer at setup. No ongoing self-observation. |
+| **Context scope** | User peer + AI peer representation, both injected. | User peer (owner) representation + conversation summary. `peerPerspective` on context call. |
+| **Session naming** | per-directory / global / manual map / title-based. | Derived from platform session key. |
+| **Multi-agent** | Single-agent only. | Parent observer hierarchy via `subagent_spawned`. |
+| **Tool surface** | Single `query_user_context` tool (on-demand dialectic). | 6 tools: session, profile, search, context (fast) + recall, analyze (LLM). |
+| **Platform metadata** | Not stripped. | Explicitly stripped before Honcho storage. |
+| **Message dedup** | None. | `lastSavedIndex` in session metadata prevents re-sending. |
+| **CLI surface in prompt** | Management commands injected into system prompt. Agent knows its own CLI. | Not injected. |
+| **AI peer name in identity** | Replaces "Hermes Agent" in DEFAULT_AGENT_IDENTITY when configured. | Not implemented. |
+| **QMD / local file search** | Not implemented. | Passthrough tools when QMD backend configured. |
+| **Workspace metadata** | Not implemented. | `agentPeerMap` in workspace metadata tracks agent→peer ID. |
+
+---
+
+## Patterns
+
+Six patterns from Hermes are worth adopting in any Honcho integration. Each is described as an integration-agnostic interface.
+
+**Hermes contributes:**
+- Async prefetch (zero-latency)
+- Dynamic reasoning level
+- Per-peer memory modes
+- AI peer identity formation
+- Session naming strategies
+- CLI surface injection
+
+**openclaw-honcho contributes back (Hermes should adopt):**
+- `lastSavedIndex` dedup
+- Platform metadata stripping
+- Multi-agent observer hierarchy
+- `peerPerspective` on `context()`
+- Tiered tool surface (fast/LLM)
+- Workspace `agentPeerMap`
+
+---
+
+## Spec: async prefetch
+
+### Problem
+
+Calling `session.context()` and `peer.chat()` synchronously before each LLM call adds 200–800ms of Honcho round-trip latency to every turn.
+
+### Pattern
+
+Fire both calls as non-blocking background work at the **end** of each turn. Store results in a per-session cache keyed by session ID. At the **start** of the next turn, pop from cache — the HTTP is already done. First turn is cold (empty cache); all subsequent turns are zero-latency on the response path.
+
+### Interface contract
+
+```typescript
+interface AsyncPrefetch {
+  // Fire context + dialectic fetches at turn end. Non-blocking.
+  firePrefetch(sessionId: string, userMessage: string): void;
+
+  // Pop cached results at turn start. Returns empty if cache is cold.
+  popContextResult(sessionId: string): ContextResult | null;
+  popDialecticResult(sessionId: string): string | null;
+}
+
+type ContextResult = {
+  representation: string;
+  card: string[];
+  aiRepresentation?: string;  // AI peer context if enabled
+  summary?: string;           // conversation summary if fetched
+};
+```
+
+### Implementation notes
+
+- **Python:** `threading.Thread(daemon=True)`. Write to `dict[session_id, result]` — GIL makes this safe for simple writes.
+- **TypeScript:** `Promise` stored in `Map<string, Promise<ContextResult>>`. Await at pop time. If not resolved yet, return null — do not block.
+- The pop is destructive: clears the cache entry after reading so stale data never accumulates.
+- Prefetch should also fire on first turn (even though it won't be consumed until turn 2).
+
+### openclaw-honcho adoption
+
+Move `session.context()` from `before_prompt_build` to a post-`agent_end` background task. Store result in `state.contextCache`. In `before_prompt_build`, read from cache instead of calling Honcho. If cache is empty (turn 1), inject nothing — the prompt is still valid without Honcho context on the first turn.
+
+---
+
+## Spec: dynamic reasoning level
+
+### Problem
+
+Honcho's dialectic endpoint supports reasoning levels from `minimal` to `max`. A fixed level per tool wastes budget on simple queries and under-serves complex ones.
+
+### Pattern
+
+Select the reasoning level dynamically based on the user's message. Use the configured default as a floor. Bump by message length. Cap auto-selection at `high` — never select `max` automatically.
+
+### Logic
+
+```
+< 120 chars  → default (typically "low")
+120–400 chars → one level above default (cap at "high")
+> 400 chars  → two levels above default (cap at "high")
+```
+
+### Config key
+
+Add `dialecticReasoningLevel` (string, default `"low"`). This sets the floor. The dynamic bump always applies on top.
+
+### openclaw-honcho adoption
+
+Apply in `honcho_recall` and `honcho_analyze`: replace fixed `reasoningLevel` with the dynamic selector. `honcho_recall` uses floor `"minimal"`, `honcho_analyze` uses floor `"medium"` — both still bump with message length.
+
+---
+
+## Spec: per-peer memory modes
+
+### Problem
+
+Users want independent control over whether user context and agent context are written locally, to Honcho, or both.
+
+### Modes
+
+| Mode | Effect |
+|---|---|
+| `hybrid` | Write to both local files and Honcho (default) |
+| `honcho` | Honcho only — disable corresponding local file writes |
+| `local` | Local files only — skip Honcho sync for this peer |
+
+### Config schema
+
+```json
+{
+  "memoryMode": "hybrid",
+  "userMemoryMode": "honcho",
+  "agentMemoryMode": "hybrid"
+}
+```
+
+Resolution order: per-peer field wins → shorthand `memoryMode` → default `"hybrid"`.
+
+### Effect on Honcho sync
+
+- `userMemoryMode=local`: skip adding user peer messages to Honcho
+- `agentMemoryMode=local`: skip adding assistant peer messages to Honcho
+- Both local: skip `session.addMessages()` entirely
+- `userMemoryMode=honcho`: disable local USER.md writes
+- `agentMemoryMode=honcho`: disable local MEMORY.md / SOUL.md writes
+
+---
+
+## Spec: AI peer identity formation
+
+### Problem
+
+Honcho builds the user's representation organically by observing what the user says. The same mechanism exists for the AI peer — but only if `observe_me=True` is set for the agent peer. Without it, the agent peer accumulates nothing.
+
+Additionally, existing persona files (SOUL.md, IDENTITY.md) should seed the AI peer's Honcho representation at first activation.
+
+### Part A: observe_me=True for agent peer
+
+```typescript
+await session.addPeers([
+  [ownerPeer.id, { observeMe: true,  observeOthers: false }],
+  [agentPeer.id, { observeMe: true,  observeOthers: true  }], // was false
+]);
+```
+
+One-line change. Foundational. Without it, the AI peer representation stays empty regardless of what the agent says.
+
+### Part B: seedAiIdentity()
+
+```typescript
+async function seedAiIdentity(
+  agentPeer: Peer,
+  content: string,
+  source: string
+): Promise<boolean> {
+  const wrapped = [
+    `<ai_identity_seed>`,
+    `<source>${source}</source>`,
+    ``,
+    content.trim(),
+    `</ai_identity_seed>`,
+  ].join("\n");
+
+  await agentPeer.addMessage("assistant", wrapped);
+  return true;
+}
+```
+
+### Part C: migrate agent files at setup
+
+During `honcho setup`, upload agent-self files (SOUL.md, IDENTITY.md, AGENTS.md) to the agent peer via `seedAiIdentity()` instead of `session.uploadFile()`. This routes content through Honcho's observation pipeline.
+
+### Part D: AI peer name in identity
+
+When the agent has a configured name, prepend it to the injected system prompt:
+
+```typescript
+const namePrefix = agentName ? `You are ${agentName}.\n\n` : "";
+return { systemPrompt: namePrefix + "## User Memory Context\n\n" + sections };
+```
+
+### CLI surface
+
+```
+honcho identity <file>    # seed from file
+honcho identity --show    # show current AI peer representation
+```
+
+---
+
+## Spec: session naming strategies
+
+### Problem
+
+A single global session means every project shares the same Honcho context. Per-directory sessions provide isolation without requiring users to name sessions manually.
+
+### Strategies
+
+| Strategy | Session key | When to use |
+|---|---|---|
+| `per-directory` | basename of CWD | Default. Each project gets its own session. |
+| `global` | fixed string `"global"` | Single cross-project session. |
+| manual map | user-configured per path | `sessions` config map overrides directory basename. |
+| title-based | sanitized session title | When agent supports named sessions set mid-conversation. |
+
+### Config schema
+
+```json
+{
+  "sessionStrategy": "per-directory",
+  "sessionPeerPrefix": false,
+  "sessions": {
+    "/home/user/projects/foo": "foo-project"
+  }
+}
+```
+
+### CLI surface
+
+```
+honcho sessions              # list all mappings
+honcho map <name>            # map cwd to session name
+honcho map                   # no-arg = list mappings
+```
+
+Resolution order: manual map → session title → directory basename → platform key.
+
+---
+
+## Spec: CLI surface injection
+
+### Problem
+
+When a user asks "how do I change my memory settings?" the agent either hallucinates or says it doesn't know. The agent should know its own management interface.
+
+### Pattern
+
+When Honcho is active, append a compact command reference to the system prompt. Keep it under 300 chars.
+
+```
+# Honcho memory integration
+Active. Session: {sessionKey}. Mode: {mode}.
+Management commands:
+  honcho status                    — show config + connection
+  honcho mode [hybrid|honcho|local] — show or set memory mode
+  honcho sessions                  — list session mappings
+  honcho map <name>                — map directory to session
+  honcho identity [file] [--show]  — seed or show AI identity
+  honcho setup                     — full interactive wizard
+```
+
+---
+
+## openclaw-honcho checklist
+
+Ordered by impact:
+
+- [ ] **Async prefetch** — move `session.context()` out of `before_prompt_build` into post-`agent_end` background Promise
+- [ ] **observe_me=True for agent peer** — one-line change in `session.addPeers()`
+- [ ] **Dynamic reasoning level** — add helper; apply in `honcho_recall` and `honcho_analyze`; add `dialecticReasoningLevel` to config
+- [ ] **Per-peer memory modes** — add `userMemoryMode` / `agentMemoryMode` to config; gate Honcho sync and local writes
+- [ ] **seedAiIdentity()** — add helper; use during setup migration for SOUL.md / IDENTITY.md
+- [ ] **Session naming strategies** — add `sessionStrategy`, `sessions` map, `sessionPeerPrefix`
+- [ ] **CLI surface injection** — append command reference to `before_prompt_build` return value
+- [ ] **honcho identity subcommand** — seed from file or `--show` current representation
+- [ ] **AI peer name injection** — if `aiPeer` name configured, prepend to injected system prompt
+- [ ] **honcho mode / sessions / map** — CLI parity with Hermes
+
+Already done in openclaw-honcho (do not re-implement): `lastSavedIndex` dedup, platform metadata stripping, multi-agent parent observer, `peerPerspective` on `context()`, tiered tool surface, workspace `agentPeerMap`, QMD passthrough, self-hosted Honcho.
+
+---
+
+## nanobot-honcho checklist
+
+Greenfield integration. Start from openclaw-honcho's architecture and apply all Hermes patterns from day one.
+
+### Phase 1 — core correctness
+
+- [ ] Dual peer model (owner + agent peer), both with `observe_me=True`
+- [ ] Message capture at turn end with `lastSavedIndex` dedup
+- [ ] Platform metadata stripping before Honcho storage
+- [ ] Async prefetch from day one — do not implement blocking context injection
+- [ ] Legacy file migration at first activation (USER.md → owner peer, SOUL.md → `seedAiIdentity()`)
+
+### Phase 2 — configuration
+
+- [ ] Config schema: `apiKey`, `workspaceId`, `baseUrl`, `memoryMode`, `userMemoryMode`, `agentMemoryMode`, `dialecticReasoningLevel`, `sessionStrategy`, `sessions`
+- [ ] Per-peer memory mode gating
+- [ ] Dynamic reasoning level
+- [ ] Session naming strategies
+
+### Phase 3 — tools and CLI
+
+- [ ] Tool surface: `honcho_profile`, `honcho_recall`, `honcho_analyze`, `honcho_search`, `honcho_context`
+- [ ] CLI: `setup`, `status`, `sessions`, `map`, `mode`, `identity`
+- [ ] CLI surface injection into system prompt
+- [ ] AI peer name wired into agent identity
@@ -0,0 +1,110 @@
+# Migrating from OpenClaw to Hermes Agent
+
+This guide covers how to import your OpenClaw settings, memories, skills, and API keys into Hermes Agent.
+
+## Three Ways to Migrate
+
+### 1. Automatic (during first-time setup)
+
+When you run `hermes setup` for the first time and Hermes detects `~/.openclaw`, it automatically offers to import your OpenClaw data before configuration begins. Just accept the prompt and everything is handled for you.
+
+### 2. CLI Command (quick, scriptable)
+
+```bash
+hermes claw migrate                      # Full migration with confirmation prompt
+hermes claw migrate --dry-run            # Preview what would happen
+hermes claw migrate --preset user-data   # Migrate without API keys/secrets
+hermes claw migrate --yes                # Skip confirmation prompt
+```
+
+**All options:**
+
+| Flag | Description |
+|------|-------------|
+| `--source PATH` | Path to OpenClaw directory (default: `~/.openclaw`) |
+| `--dry-run` | Preview only — no files are modified |
+| `--preset {user-data,full}` | Migration preset (default: `full`). `user-data` excludes secrets |
+| `--overwrite` | Overwrite existing files (default: skip conflicts) |
+| `--migrate-secrets` | Include allowlisted secrets (auto-enabled with `full` preset) |
+| `--workspace-target PATH` | Copy workspace instructions (AGENTS.md) to this absolute path |
+| `--skill-conflict {skip,overwrite,rename}` | How to handle skill name conflicts (default: `skip`) |
+| `--yes`, `-y` | Skip confirmation prompts |
+
+### 3. Agent-Guided (interactive, with previews)
+
+Ask the agent to run the migration for you:
+
+```
+> Migrate my OpenClaw setup to Hermes
+```
+
+The agent will use the `openclaw-migration` skill to:
+1. Run a dry-run first to preview changes
+2. Ask about conflict resolution (SOUL.md, skills, etc.)
+3. Let you choose between `user-data` and `full` presets
+4. Execute the migration with your choices
+5. Print a detailed summary of what was migrated
+
+## What Gets Migrated
+
+### `user-data` preset
+| Item | Source | Destination |
+|------|--------|-------------|
+| SOUL.md | `~/.openclaw/workspace/SOUL.md` | `~/.hermes/SOUL.md` |
+| Memory entries | `~/.openclaw/workspace/MEMORY.md` | `~/.hermes/memories/MEMORY.md` |
+| User profile | `~/.openclaw/workspace/USER.md` | `~/.hermes/memories/USER.md` |
+| Skills | `~/.openclaw/workspace/skills/` | `~/.hermes/skills/openclaw-imports/` |
+| Command allowlist | `~/.openclaw/workspace/exec_approval_patterns.yaml` | Merged into `~/.hermes/config.yaml` |
+| Messaging settings | `~/.openclaw/config.yaml` (TELEGRAM_ALLOWED_USERS, MESSAGING_CWD) | `~/.hermes/.env` |
+| TTS assets | `~/.openclaw/workspace/tts/` | `~/.hermes/tts/` |
+
+### `full` preset (adds to `user-data`)
+| Item | Source | Destination |
+|------|--------|-------------|
+| Telegram bot token | `~/.openclaw/config.yaml` | `~/.hermes/.env` |
+| OpenRouter API key | `~/.openclaw/.env` or config | `~/.hermes/.env` |
+| OpenAI API key | `~/.openclaw/.env` or config | `~/.hermes/.env` |
+| Anthropic API key | `~/.openclaw/.env` or config | `~/.hermes/.env` |
+| ElevenLabs API key | `~/.openclaw/.env` or config | `~/.hermes/.env` |
+
+Only these 6 allowlisted secrets are ever imported. Other credentials are skipped and reported.
+
+## Conflict Handling
+
+By default, the migration **will not overwrite** existing Hermes data:
+
+- **SOUL.md** — skipped if one already exists in `~/.hermes/`
+- **Memory entries** — skipped if memories already exist (to avoid duplicates)
+- **Skills** — skipped if a skill with the same name already exists
+- **API keys** — skipped if the key is already set in `~/.hermes/.env`
+
+To overwrite conflicts, use `--overwrite`. The migration creates backups before overwriting.
+
+For skills, you can also use `--skill-conflict rename` to import conflicting skills under a new name (e.g., `skill-name-imported`).
+
+## Migration Report
+
+Every migration (including dry runs) produces a report showing:
+- **Migrated items** — what was successfully imported
+- **Conflicts** — items skipped because they already exist
+- **Skipped items** — items not found in the source
+- **Errors** — items that failed to import
+
+For execute runs, the full report is saved to `~/.hermes/migration/openclaw/<timestamp>/`.
+
+## Troubleshooting
+
+### "OpenClaw directory not found"
+The migration looks for `~/.openclaw` by default. If your OpenClaw is installed elsewhere, use `--source`:
+```bash
+hermes claw migrate --source /path/to/.openclaw
+```
+
+### "Migration script not found"
+The migration script ships with Hermes Agent. If you installed via pip (not git clone), the `optional-skills/` directory may not be present. Install the skill from the Skills Hub:
+```bash
+hermes skills install openclaw-migration
+```
+
+### Memory overflow
+If your OpenClaw MEMORY.md or USER.md exceeds Hermes' character limits, excess entries are exported to an overflow file in the migration report directory. You can manually review and add the most important ones.
@@ -0,0 +1,608 @@
+# Pricing Accuracy Architecture
+
+Date: 2026-03-16
+
+## Goal
+
+Hermes should only show dollar costs when they are backed by an official source for the user's actual billing path.
+
+This design replaces the current static, heuristic pricing flow in:
+
+- `run_agent.py`
+- `agent/usage_pricing.py`
+- `agent/insights.py`
+- `cli.py`
+
+with a provider-aware pricing system that:
+
+- handles cache billing correctly
+- distinguishes `actual` vs `estimated` vs `included` vs `unknown`
+- reconciles post-hoc costs when providers expose authoritative billing data
+- supports direct providers, OpenRouter, subscriptions, enterprise pricing, and custom endpoints
+
+## Problems In The Current Design
+
+Current Hermes behavior has four structural issues:
+
+1. It stores only `prompt_tokens` and `completion_tokens`, which is insufficient for providers that bill cache reads and cache writes separately.
+2. It uses a static model price table and fuzzy heuristics, which can drift from current official pricing.
+3. It assumes public API list pricing matches the user's real billing path.
+4. It has no distinction between live estimates and reconciled billed cost.
+
+## Design Principles
+
+1. Normalize usage before pricing.
+2. Never fold cached tokens into plain input cost.
+3. Track certainty explicitly.
+4. Treat the billing path as part of the model identity.
+5. Prefer official machine-readable sources over scraped docs.
+6. Use post-hoc provider cost APIs when available.
+7. Show `n/a` rather than inventing precision.
+
+## High-Level Architecture
+
+The new system has four layers:
+
+1. `usage_normalization`
+   Converts raw provider usage into a canonical usage record.
+2. `pricing_source_resolution`
+   Determines the billing path, source of truth, and applicable pricing source.
+3. `cost_estimation_and_reconciliation`
+   Produces an immediate estimate when possible, then replaces or annotates it with actual billed cost later.
+4. `presentation`
+   `/usage`, `/insights`, and the status bar display cost with certainty metadata.
+
+## Canonical Usage Record
+
+Add a canonical usage model that every provider path maps into before any pricing math happens.
+
+Suggested structure:
+
+```python
+@dataclass
+class CanonicalUsage:
+    provider: str
+    billing_provider: str
+    model: str
+    billing_route: str
+
+    input_tokens: int = 0
+    output_tokens: int = 0
+    cache_read_tokens: int = 0
+    cache_write_tokens: int = 0
+    reasoning_tokens: int = 0
+    request_count: int = 1
+
+    raw_usage: dict[str, Any] | None = None
+    raw_usage_fields: dict[str, str] | None = None
+    computed_fields: set[str] | None = None
+
+    provider_request_id: str | None = None
+    provider_generation_id: str | None = None
+    provider_response_id: str | None = None
+```
+
+Rules:
+
+- `input_tokens` means non-cached input only.
+- `cache_read_tokens` and `cache_write_tokens` are never merged into `input_tokens`.
+- `output_tokens` excludes cache metrics.
+- `reasoning_tokens` is telemetry unless a provider officially bills it separately.
+
+This is the same normalization pattern used by `opencode`, extended with provenance and reconciliation ids.
+
+## Provider Normalization Rules
+
+### OpenAI Direct
+
+Source usage fields:
+
+- `prompt_tokens`
+- `completion_tokens`
+- `prompt_tokens_details.cached_tokens`
+
+Normalization:
+
+- `cache_read_tokens = cached_tokens`
+- `input_tokens = prompt_tokens - cached_tokens`
+- `cache_write_tokens = 0` unless OpenAI exposes it in the relevant route
+- `output_tokens = completion_tokens`
+
+### Anthropic Direct
+
+Source usage fields:
+
+- `input_tokens`
+- `output_tokens`
+- `cache_read_input_tokens`
+- `cache_creation_input_tokens`
+
+Normalization:
+
+- `input_tokens = input_tokens`
+- `output_tokens = output_tokens`
+- `cache_read_tokens = cache_read_input_tokens`
+- `cache_write_tokens = cache_creation_input_tokens`
+
+### OpenRouter
+
+Estimate-time usage normalization should use the response usage payload with the same rules as the underlying provider when possible.
+
+Reconciliation-time records should also store:
+
+- OpenRouter generation id
+- native token fields when available
+- `total_cost`
+- `cache_discount`
+- `upstream_inference_cost`
+- `is_byok`
+
+### Gemini / Vertex
+
+Use official Gemini or Vertex usage fields where available.
+
+If cached content tokens are exposed:
+
+- map them to `cache_read_tokens`
+
+If a route exposes no cache creation metric:
+
+- store `cache_write_tokens = 0`
+- preserve the raw usage payload for later extension
+
+### DeepSeek And Other Direct Providers
+
+Normalize only the fields that are officially exposed.
+
+If a provider does not expose cache buckets:
+
+- do not infer them unless the provider explicitly documents how to derive them
+
+### Subscription / Included-Cost Routes
+
+These still use the canonical usage model.
+
+Tokens are tracked normally. Cost depends on billing mode, not on whether usage exists.
+
+## Billing Route Model
+
+Hermes must stop keying pricing solely by `model`.
+
+Introduce a billing route descriptor:
+
+```python
+@dataclass
+class BillingRoute:
+    provider: str
+    base_url: str | None
+    model: str
+    billing_mode: str
+    organization_hint: str | None = None
+```
+
+`billing_mode` values:
+
+- `official_cost_api`
+- `official_generation_api`
+- `official_models_api`
+- `official_docs_snapshot`
+- `subscription_included`
+- `user_override`
+- `custom_contract`
+- `unknown`
+
+Examples:
+
+- OpenAI direct API with Costs API access: `official_cost_api`
+- Anthropic direct API with Usage & Cost API access: `official_cost_api`
+- OpenRouter request before reconciliation: `official_models_api`
+- OpenRouter request after generation lookup: `official_generation_api`
+- GitHub Copilot style subscription route: `subscription_included`
+- local OpenAI-compatible server: `unknown`
+- enterprise contract with configured rates: `custom_contract`
+
+## Cost Status Model
+
+Every displayed cost should have:
+
+```python
+@dataclass
+class CostResult:
+    amount_usd: Decimal | None
+    status: Literal["actual", "estimated", "included", "unknown"]
+    source: Literal[
+        "provider_cost_api",
+        "provider_generation_api",
+        "provider_models_api",
+        "official_docs_snapshot",
+        "user_override",
+        "custom_contract",
+        "none",
+    ]
+    label: str
+    fetched_at: datetime | None
+    pricing_version: str | None
+    notes: list[str]
+```
+
+Presentation rules:
+
+- `actual`: show dollar amount as final
+- `estimated`: show dollar amount with estimate labeling
+- `included`: show `included` or `$0.00 (included)` depending on UX choice
+- `unknown`: show `n/a`
+
+## Official Source Hierarchy
+
+Resolve cost using this order:
+
+1. Request-level or account-level official billed cost
+2. Official machine-readable model pricing
+3. Official docs snapshot
+4. User override or custom contract
+5. Unknown
+
+The system must never skip to a lower level if a higher-confidence source exists for the current billing route.
+
+## Provider-Specific Truth Rules
+
+### OpenAI Direct
+
+Preferred truth:
+
+1. Costs API for reconciled spend
+2. Official pricing page for live estimate
+
+### Anthropic Direct
+
+Preferred truth:
+
+1. Usage & Cost API for reconciled spend
+2. Official pricing docs for live estimate
+
+### OpenRouter
+
+Preferred truth:
+
+1. `GET /api/v1/generation` for reconciled `total_cost`
+2. `GET /api/v1/models` pricing for live estimate
+
+Do not use underlying provider public pricing as the source of truth for OpenRouter billing.
+
+### Gemini / Vertex
+
+Preferred truth:
+
+1. official billing export or billing API for reconciled spend when available for the route
+2. official pricing docs for estimate
+
+### DeepSeek
+
+Preferred truth:
+
+1. official machine-readable cost source if available in the future
+2. official pricing docs snapshot today
+
+### Subscription-Included Routes
+
+Preferred truth:
+
+1. explicit route config marking the model as included in subscription
+
+These should display `included`, not an API list-price estimate.
+
+### Custom Endpoint / Local Model
+
+Preferred truth:
+
+1. user override
+2. custom contract config
+3. unknown
+
+These should default to `unknown`.
+
+## Pricing Catalog
+
+Replace the current `MODEL_PRICING` dict with a richer pricing catalog.
+
+Suggested record:
+
+```python
+@dataclass
+class PricingEntry:
+    provider: str
+    route_pattern: str
+    model_pattern: str
+
+    input_cost_per_million: Decimal | None = None
+    output_cost_per_million: Decimal | None = None
+    cache_read_cost_per_million: Decimal | None = None
+    cache_write_cost_per_million: Decimal | None = None
+    request_cost: Decimal | None = None
+    image_cost: Decimal | None = None
+
+    source: str = "official_docs_snapshot"
+    source_url: str | None = None
+    fetched_at: datetime | None = None
+    pricing_version: str | None = None
+```
+
+The catalog should be route-aware:
+
+- `openai:gpt-5`
+- `anthropic:claude-opus-4-6`
+- `openrouter:anthropic/claude-opus-4.6`
+- `copilot:gpt-4o`
+
+This avoids conflating direct-provider billing with aggregator billing.
+
+## Pricing Sync Architecture
+
+Introduce a pricing sync subsystem instead of manually maintaining a single hardcoded table.
+
+Suggested modules:
+
+- `agent/pricing/catalog.py`
+- `agent/pricing/sources.py`
+- `agent/pricing/sync.py`
+- `agent/pricing/reconcile.py`
+- `agent/pricing/types.py`
+
+### Sync Sources
+
+- OpenRouter models API
+- official provider docs snapshots where no API exists
+- user overrides from config
+
+### Sync Output
+
+Cache pricing entries locally with:
+
+- source URL
+- fetch timestamp
+- version/hash
+- confidence/source type
+
+### Sync Frequency
+
+- startup warm cache
+- background refresh every 6 to 24 hours depending on source
+- manual `hermes pricing sync`
+
+## Reconciliation Architecture
+
+Live requests may produce only an estimate initially. Hermes should reconcile them later when a provider exposes actual billed cost.
+
+Suggested flow:
+
+1. Agent call completes.
+2. Hermes stores canonical usage plus reconciliation ids.
+3. Hermes computes an immediate estimate if a pricing source exists.
+4. A reconciliation worker fetches actual cost when supported.
+5. Session and message records are updated with `actual` cost.
+
+This can run:
+
+- inline for cheap lookups
+- asynchronously for delayed provider accounting
+
+## Persistence Changes
+
+Session storage should stop storing only aggregate prompt/completion totals.
+
+Add fields for both usage and cost certainty:
+
+- `input_tokens`
+- `output_tokens`
+- `cache_read_tokens`
+- `cache_write_tokens`
+- `reasoning_tokens`
+- `estimated_cost_usd`
+- `actual_cost_usd`
+- `cost_status`
+- `cost_source`
+- `pricing_version`
+- `billing_provider`
+- `billing_mode`
+
+If schema expansion is too large for one PR, add a new pricing events table:
+
+```text
+session_cost_events
+  id
+  session_id
+  request_id
+  provider
+  model
+  billing_mode
+  input_tokens
+  output_tokens
+  cache_read_tokens
+  cache_write_tokens
+  estimated_cost_usd
+  actual_cost_usd
+  cost_status
+  cost_source
+  pricing_version
+  created_at
+  updated_at
+```
+
+## Hermes Touchpoints
+
+### `run_agent.py`
+
+Current responsibility:
+
+- parse raw provider usage
+- update session token counters
+
+New responsibility:
+
+- build `CanonicalUsage`
+- update canonical counters
+- store reconciliation ids
+- emit usage event to pricing subsystem
+
+### `agent/usage_pricing.py`
+
+Current responsibility:
+
+- static lookup table
+- direct cost arithmetic
+
+New responsibility:
+
+- move or replace with pricing catalog facade
+- no fuzzy model-family heuristics
+- no direct pricing without billing-route context
+
+### `cli.py`
+
+Current responsibility:
+
+- compute session cost directly from prompt/completion totals
+
+New responsibility:
+
+- display `CostResult`
+- show status badges:
+  - `actual`
+  - `estimated`
+  - `included`
+  - `n/a`
+
+### `agent/insights.py`
+
+Current responsibility:
+
+- recompute historical estimates from static pricing
+
+New responsibility:
+
+- aggregate stored pricing events
+- prefer actual cost over estimate
+- surface estimates only when reconciliation is unavailable
+
+## UX Rules
+
+### Status Bar
+
+Show one of:
+
+- `$1.42`
+- `~$1.42`
+- `included`
+- `cost n/a`
+
+Where:
+
+- `$1.42` means `actual`
+- `~$1.42` means `estimated`
+- `included` means subscription-backed or explicitly zero-cost route
+- `cost n/a` means unknown
+
+### `/usage`
+
+Show:
+
+- token buckets
+- estimated cost
+- actual cost if available
+- cost status
+- pricing source
+
+### `/insights`
+
+Aggregate:
+
+- actual cost totals
+- estimated-only totals
+- unknown-cost sessions count
+- included-cost sessions count
+
+## Config And Overrides
+
+Add user-configurable pricing overrides in config:
+
+```yaml
+pricing:
+  mode: hybrid
+  sync_on_startup: true
+  sync_interval_hours: 12
+  overrides:
+    - provider: openrouter
+      model: anthropic/claude-opus-4.6
+      billing_mode: custom_contract
+      input_cost_per_million: 4.25
+      output_cost_per_million: 22.0
+      cache_read_cost_per_million: 0.5
+      cache_write_cost_per_million: 6.0
+  included_routes:
+    - provider: copilot
+      model: "*"
+    - provider: codex-subscription
+      model: "*"
+```
+
+Overrides must win over catalog defaults for the matching billing route.
+
+## Rollout Plan
+
+### Phase 1
+
+- add canonical usage model
+- split cache token buckets in `run_agent.py`
+- stop pricing cache-inflated prompt totals
+- preserve current UI with improved backend math
+
+### Phase 2
+
+- add route-aware pricing catalog
+- integrate OpenRouter models API sync
+- add `estimated` vs `included` vs `unknown`
+
+### Phase 3
+
+- add reconciliation for OpenRouter generation cost
+- add actual cost persistence
+- update `/insights` to prefer actual cost
+
+### Phase 4
+
+- add direct OpenAI and Anthropic reconciliation paths
+- add user overrides and contract pricing
+- add pricing sync CLI command
+
+## Testing Strategy
+
+Add tests for:
+
+- OpenAI cached token subtraction
+- Anthropic cache read/write separation
+- OpenRouter estimated vs actual reconciliation
+- subscription-backed models showing `included`
+- custom endpoints showing `n/a`
+- override precedence
+- stale catalog fallback behavior
+
+Current tests that assume heuristic pricing should be replaced with route-aware expectations.
+
+## Non-Goals
+
+- exact enterprise billing reconstruction without an official source or user override
+- backfilling perfect historical cost for old sessions that lack cache bucket data
+- scraping arbitrary provider web pages at request time
+
+## Recommendation
+
+Do not expand the existing `MODEL_PRICING` dict.
+
+That path cannot satisfy the product requirement. Hermes should instead migrate to:
+
+- canonical usage normalization
+- route-aware pricing sources
+- estimate-then-reconcile cost lifecycle
+- explicit certainty states in the UI
+
+This is the minimum architecture that makes the statement "Hermes pricing is backed by official sources where possible, and otherwise clearly labeled" defensible.
@@ -0,0 +1,89 @@
+# ============================================================================
+# Hermes Agent — Example Skin Template
+# ============================================================================
+#
+# Copy this file to ~/.hermes/skins/<name>.yaml to create a custom skin.
+# All fields are optional — missing values inherit from the default skin.
+# Activate with: /skin <name>  or  display.skin: <name> in config.yaml
+#
+# See hermes_cli/skin_engine.py for the full schema reference.
+# ============================================================================
+
+# Required: unique skin name (used in /skin command and config)
+name: example
+description: An example custom skin — copy and modify this template
+
+# ── Colors ──────────────────────────────────────────────────────────────────
+# Hex color values for Rich markup. These control the CLI's visual palette.
+colors:
+  # Banner panel (the startup welcome box)
+  banner_border: "#CD7F32"        # Panel border
+  banner_title: "#FFD700"         # Panel title text
+  banner_accent: "#FFBF00"        # Section headers (Available Tools, Skills, etc.)
+  banner_dim: "#B8860B"           # Dim/muted text (separators, model info)
+  banner_text: "#FFF8DC"          # Body text (tool names, skill names)
+
+  # UI elements
+  ui_accent: "#FFBF00"            # General accent color
+  ui_label: "#4dd0e1"             # Labels
+  ui_ok: "#4caf50"                # Success indicators
+  ui_error: "#ef5350"             # Error indicators
+  ui_warn: "#ffa726"              # Warning indicators
+
+  # Input area
+  prompt: "#FFF8DC"               # Prompt text color
+  input_rule: "#CD7F32"           # Horizontal rule around input
+
+  # Response box
+  response_border: "#FFD700"      # Response box border (ANSI color)
+
+  # Session display
+  session_label: "#DAA520"        # Session label
+  session_border: "#8B8682"       # Session ID dim color
+
+# ── Spinner ─────────────────────────────────────────────────────────────────
+# Customize the animated spinner shown during API calls and tool execution.
+spinner:
+  # Faces shown while waiting for the API response
+  waiting_faces:
+    - "(｡◕‿◕｡)"
+    - "(◕‿◕✿)"
+    - "٩(◕‿◕｡)۶"
+
+  # Faces shown during extended thinking/reasoning
+  thinking_faces:
+    - "(｡•́︿•̀｡)"
+    - "(◔_◔)"
+    - "(¬‿¬)"
+
+  # Verbs used in spinner messages (e.g., "pondering your request...")
+  thinking_verbs:
+    - "pondering"
+    - "contemplating"
+    - "musing"
+    - "ruminating"
+
+  # Optional: left/right decorations around the spinner
+  # Each entry is a [left, right] pair. Omit entirely for no wings.
+  # wings:
+  #   - ["⟪⚔", "⚔⟫"]
+  #   - ["⟪▲", "▲⟫"]
+
+# ── Branding ────────────────────────────────────────────────────────────────
+# Text strings used throughout the CLI interface.
+branding:
+  agent_name: "Hermes Agent"          # Banner title, about display
+  welcome: "Welcome! Type your message or /help for commands."
+  goodbye: "Goodbye! ⚕"              # Exit message
+  response_label: " ⚕ Hermes "       # Response box header label
+  prompt_symbol: "❯ "                 # Input prompt symbol
+  help_header: "(^_^)? Available Commands"  # /help header text
+
+# ── Tool Output ─────────────────────────────────────────────────────────────
+# Character used as the prefix for tool output lines.
+# Default is "┊" (thin dotted vertical line). Some alternatives:
+#   "╎" (light triple dash vertical)
+#   "▏" (left one-eighth block)
+#   "│" (box drawing light vertical)
+#   "┃" (box drawing heavy vertical)
+tool_prefix: "┊"
@@ -138,6 +138,7 @@ class HermesAgentLoop:
        max_turns: int = 30,
        task_id: Optional[str] = None,
        temperature: float = 1.0,
+        top_p: Optional[float] = None,
        max_tokens: Optional[int] = None,
        extra_body: Optional[Dict[str, Any]] = None,
        budget_config: Optional["BudgetConfig"] = None,
@@ -153,6 +154,7 @@ class HermesAgentLoop:
            max_turns: Maximum number of LLM calls before stopping
            task_id: Unique ID for terminal/browser session isolation
            temperature: Sampling temperature for generation
+            top_p: Nucleus sampling top_p (None = omit, use provider default)
            max_tokens: Max tokens per generation (None for server default)
            extra_body: Extra parameters passed to the OpenAI client's create() call.
                        Used for OpenRouter provider preferences, transforms, etc.
@@ -168,6 +170,7 @@ class HermesAgentLoop:
        self.max_turns = max_turns
        self.task_id = task_id or str(uuid.uuid4())
        self.temperature = temperature
+        self.top_p = top_p
        self.max_tokens = max_tokens
        self.extra_body = extra_body
        self.budget_config = budget_config or DEFAULT_BUDGET
@@ -211,6 +214,9 @@ class HermesAgentLoop:
                "temperature": self.temperature,
            }

+            if self.top_p is not None:
+                chat_kwargs["top_p"] = self.top_p
+
            # Only pass tools if we have them
            if self.tool_schemas:
                chat_kwargs["tools"] = self.tool_schemas
@@ -225,20 +231,35 @@ class HermesAgentLoop:
                chat_kwargs["extra_body"] = self.extra_body

            # Make the API call -- standard OpenAI spec
+            # Retry on timeout/connection errors (provider queuing, rate limits)
            api_start = _time.monotonic()
-            try:
-                response = await self.server.chat_completion(**chat_kwargs)
-            except Exception as e:
-                api_elapsed = _time.monotonic() - api_start
-                logger.error("API call failed on turn %d (%.1fs): %s", turn + 1, api_elapsed, e)
-                return AgentResult(
-                    messages=messages,
-                    managed_state=self._get_managed_state(),
-                    turns_used=turn + 1,
-                    finished_naturally=False,
-                    reasoning_per_turn=reasoning_per_turn,
-                    tool_errors=tool_errors,
-                )
+            response = None
+            max_retries = 3
+            for attempt in range(max_retries):
+                try:
+                    response = await self.server.chat_completion(**chat_kwargs)
+                    break
+                except Exception as e:
+                    api_elapsed = _time.monotonic() - api_start
+                    is_retryable = "timeout" in type(e).__name__.lower() or "connection" in type(e).__name__.lower()
+                    if is_retryable and attempt < max_retries - 1:
+                        wait = 2 ** attempt
+                        logger.warning(
+                            "[%s] API call timed out on turn %d attempt %d (%.1fs), retrying in %ds: %s",
+                            self.task_id[:8], turn + 1, attempt + 1, api_elapsed, wait, e,
+                        )
+                        await asyncio.sleep(wait)
+                        api_start = _time.monotonic()
+                        continue
+                    logger.error("API call failed on turn %d (%.1fs): %s", turn + 1, api_elapsed, e)
+                    return AgentResult(
+                        messages=messages,
+                        managed_state=self._get_managed_state(),
+                        turns_used=turn + 1,
+                        finished_naturally=False,
+                        reasoning_per_turn=reasoning_per_turn,
+                        tool_errors=tool_errors,
+                    )

            api_elapsed = _time.monotonic() - api_start

@@ -15,15 +15,15 @@

 env:
  enabled_toolsets: ["terminal", "file"]
-  max_agent_turns: 60
+  max_agent_turns: 100
  max_token_length: 32000
-  agent_temperature: 0.8
+  agent_temperature: 1.0
  terminal_backend: "modal"
-  terminal_timeout: 300        # 5 min per command (builds, pip install)
-  tool_pool_size: 128          # thread pool for 89 parallel tasks
-  dataset_name: "NousResearch/terminal-bench-2"
+  terminal_timeout: 300 # 5 min per command (builds, pip install)
+  tool_pool_size: 128 # thread pool for 89 parallel tasks
+  dataset_name: "NousResearch/terminal-bench-2-verified-flattened"
  test_timeout: 600
-  task_timeout: 1800           # 30 min wall-clock per task, auto-FAIL if exceeded
+  task_timeout: 900 # 15 min wall-clock per task, auto-FAIL if exceeded
  tokenizer_name: "NousResearch/Hermes-3-Llama-3.1-8B"
  use_wandb: true
  wandb_name: "terminal-bench-2"
@@ -33,10 +33,15 @@ env:
  # Modal's blocking calls (App.lookup, etc.) deadlock when too many sandboxes
  # are created simultaneously inside thread pool workers via asyncio.run().
  max_concurrent_tasks: 8
+  extra_body:
+    provider:
+      order: ["DeepInfra"]
+      allow_fallbacks: false

 openai:
  base_url: "https://openrouter.ai/api/v1"
-  model_name: "anthropic/claude-opus-4.6"
+  model_name: "nvidia/nemotron-3-super-120b-a12b"
  server_type: "openai"
  health_check: false
+  timeout: 300 # 5 min per API call (default 1200s causes 20min stalls)
  # api_key loaded from OPENROUTER_API_KEY in .env
@@ -32,8 +32,8 @@ export PYTHONUNBUFFERED=1
 # These go to the log file; tqdm + [START]/[PASS]/[FAIL] go to terminal
 export LOGLEVEL=INFO

-python terminalbench2_env.py evaluate \
-  --config default.yaml \
+uv run python environments/benchmarks/terminalbench_2/terminalbench2_env.py evaluate \
+  --config environments/benchmarks/terminalbench_2/default.yaml \
  "$@" \
  2>&1 | tee "$LOG_FILE"

@@ -52,18 +52,18 @@ _repo_root = Path(__file__).resolve().parent.parent.parent.parent
 if str(_repo_root) not in sys.path:
    sys.path.insert(0, str(_repo_root))

-from pydantic import Field
-
 from atroposlib.envs.base import EvalHandlingEnum
 from atroposlib.envs.server_handling.server_manager import APIServerConfig
+from pydantic import Field

+from agent.prompt_builder import DEFAULT_AGENT_IDENTITY
 from environments.agent_loop import AgentResult, HermesAgentLoop
 from environments.hermes_base_env import HermesAgentBaseEnv, HermesAgentEnvConfig
 from environments.tool_context import ToolContext
 from tools.terminal_tool import (
-    register_task_env_overrides,
-    clear_task_env_overrides,
    cleanup_vm,
+    clear_task_env_overrides,
+    register_task_env_overrides,
 )

 logger = logging.getLogger(__name__)
@@ -73,6 +73,7 @@ logger = logging.getLogger(__name__)
 # Configuration
 # =============================================================================

+
 class TerminalBench2EvalConfig(HermesAgentEnvConfig):
    """
    Configuration for the Terminal-Bench 2.0 evaluation environment.
@@ -138,11 +139,27 @@ class TerminalBench2EvalConfig(HermesAgentEnvConfig):

 # Tasks that cannot run properly on Modal and are excluded from scoring.
 MODAL_INCOMPATIBLE_TASKS = {
-    "qemu-startup",        # Needs KVM/hardware virtualization
-    "qemu-alpine-ssh",     # Needs KVM/hardware virtualization
-    "crack-7z-hash",       # Password brute-force -- too slow for cloud sandbox timeouts
+    "qemu-startup",  # Needs KVM/hardware virtualization
+    "qemu-alpine-ssh",  # Needs KVM/hardware virtualization
+    "crack-7z-hash",  # Password brute-force -- too slow for cloud sandbox timeouts
 }

+# Injected as a user message when the model responds with plain text instead of
+# calling a tool or including a <task_status> tag.
+_FORMAT_NUDGE_MESSAGE = (
+    "You wrote a plain text response instead of using your tools. "
+    "Plain text responses do not affect the environment — nothing was executed or saved.\n\n"
+    "You MUST use your tools (terminal, read_file, write_file) to actually complete the task. "
+    "Do not describe what you would do — execute it now by making tool calls.\n\n"
+    "If you have already completed all required work using tools in previous turns, "
+    "respond with exactly: <task_status>DONE</task_status>\n"
+    "If you have exhausted all approaches and cannot make further progress, "
+    "respond with exactly: <task_status>UNFINISHED</task_status>"
+)
+
+# Maximum number of format nudges before giving up and moving on to scoring.
+_MAX_FORMAT_NUDGES = 3
+

 # =============================================================================
 # Tar extraction helper
@@ -203,7 +220,6 @@ def _safe_extract_tar(tar: tarfile.TarFile, target_dir: Path) -> None:
        except OSError:
            pass

-
 def _extract_base64_tar(b64_data: str, target_dir: Path):
    """Extract a base64-encoded tar.gz archive into target_dir."""
    if not b64_data:
@@ -218,6 +234,7 @@ def _extract_base64_tar(b64_data: str, target_dir: Path):
 # Main Environment
 # =============================================================================

+
 class TerminalBench2EvalEnv(HermesAgentBaseEnv):
    """
    Terminal-Bench 2.0 evaluation environment (eval-only, no training).
@@ -262,23 +279,18 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
            enabled_toolsets=["terminal", "file"],
            disabled_toolsets=None,
            distribution=None,
-
            # Agent settings -- TB2 tasks are complex, need many turns
            max_agent_turns=60,
            max_token_length=16000,
            agent_temperature=0.6,
-            system_prompt=None,
-
+            system_prompt=DEFAULT_AGENT_IDENTITY,
            # Modal backend for per-task cloud-isolated sandboxes
            terminal_backend="modal",
-            terminal_timeout=300,   # 5 min per command (builds, pip install, etc.)
-
+            terminal_timeout=300,  # 5 min per command (builds, pip install, etc.)
            # Test execution timeout (TB2 test scripts can install deps like pytest)
            test_timeout=180,
-
            # 89 tasks run in parallel, each needs a thread for tool calls
            tool_pool_size=128,
-
            # --- Eval-only Atropos settings ---
            # These settings make the env work as an eval-only environment:
            #   - STOP_TRAIN: pauses training during eval (standard for eval envs)
@@ -288,7 +300,6 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
            group_size=1,
            steps_per_eval=1,
            total_steps=1,
-
            tokenizer_name="NousResearch/Hermes-3-Llama-3.1-8B",
            use_wandb=True,
            wandb_name="terminal-bench-2",
@@ -336,7 +347,11 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):

        # Skip tasks incompatible with the current backend (e.g., QEMU on Modal)
        # plus any user-specified skip_tasks
-        skip = set(MODAL_INCOMPATIBLE_TASKS) if self.config.terminal_backend == "modal" else set()
+        skip = (
+            set(MODAL_INCOMPATIBLE_TASKS)
+            if self.config.terminal_backend == "modal"
+            else set()
+        )
        if self.config.skip_tasks:
            skip |= {name.strip() for name in self.config.skip_tasks.split(",")}
        if skip:
@@ -344,7 +359,9 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
            tasks = [t for t in tasks if t["task_name"] not in skip]
            skipped = before - len(tasks)
            if skipped > 0:
-                print(f"  Skipped {skipped} incompatible tasks: {sorted(skip & {t['task_name'] for t in ds})}")
+                print(
+                    f"  Skipped {skipped} incompatible tasks: {sorted(skip & {t['task_name'] for t in ds})}"
+                )

        self.all_eval_items = tasks
        self.iter = 0
@@ -354,6 +371,16 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
        for i, task in enumerate(self.all_eval_items):
            self.category_index[task.get("category", "unknown")].append(i)

+        # Pre-compute which tasks need Modal's add_python (avoids re-decoding
+        # multi-MB environment_tar blobs during per-task rollouts).
+        self._needs_add_python: Dict[str, bool] = {
+            task["task_name"]: self._image_needs_add_python(task)
+            for task in self.all_eval_items
+        }
+        add_py_count = sum(self._needs_add_python.values())
+        if add_py_count:
+            print(f"  {add_py_count} tasks need add_python (non-python base image)")
+
        # Reward tracking for wandb logging
        self.eval_metrics: List[Tuple[str, float]] = []

@@ -361,15 +388,30 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
        # immediately on completion so data is preserved even on Ctrl+C.
        # Timestamped filename so each run produces a unique file.
        import datetime
+
        log_dir = os.path.join(os.path.dirname(__file__), "logs")
        os.makedirs(log_dir, exist_ok=True)
        run_ts = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
-        self._streaming_path = os.path.join(log_dir, f"samples_{run_ts}.jsonl")
+        model_name = self.server.servers[0].config.model_name
+        model_slug = model_name.replace("/", "_").replace(":", "_")
+        self._streaming_path = os.path.join(
+            log_dir, f"samples_{run_ts}_{model_slug}.jsonl"
+        )
        self._streaming_file = open(self._streaming_path, "w")
        self._streaming_lock = __import__("threading").Lock()
+        self._run_meta = {
+            "model_name": model_name,
+            "temperature": self.config.agent_temperature,
+            "top_p": self.config.agent_top_p,
+            "max_agent_turns": self.config.max_agent_turns,
+            "task_timeout": self.config.task_timeout,
+            "terminal_backend": self.config.terminal_backend,
+        }
        print(f"  Streaming results to: {self._streaming_path}")

-        print(f"TB2 ready: {len(self.all_eval_items)} tasks across {len(self.category_index)} categories")
+        print(
+            f"TB2 ready: {len(self.all_eval_items)} tasks across {len(self.category_index)} categories"
+        )
        for cat, indices in sorted(self.category_index.items()):
            print(f"  {cat}: {len(indices)} tasks")

@@ -378,7 +420,9 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
        if not hasattr(self, "_streaming_file") or self._streaming_file.closed:
            return
        with self._streaming_lock:
-            self._streaming_file.write(json.dumps(result, ensure_ascii=False, default=str) + "\n")
+            self._streaming_file.write(
+                json.dumps(result, ensure_ascii=False, default=str) + "\n"
+            )
            self._streaming_file.flush()

    # =========================================================================
@@ -414,6 +458,36 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
    # Docker image resolution
    # =========================================================================

+    @staticmethod
+    def _image_needs_add_python(item: Dict[str, Any]) -> bool:
+        """Check if the task's base image lacks `python` on PATH.
+
+        Parses the Dockerfile FROM line in environment_tar. Returns True
+        for non-python base images (ubuntu, debian, etc.) that need
+        Modal's add_python parameter.
+        """
+        environment_tar = item.get("environment_tar", "")
+        if not environment_tar:
+            return False
+        try:
+            raw = base64.b64decode(environment_tar)
+            buf = io.BytesIO(raw)
+            with tarfile.open(fileobj=buf, mode="r:gz") as tar:
+                for member in tar:
+                    if not member.isfile() or "Dockerfile" not in member.name:
+                        continue
+                    f = tar.extractfile(member)
+                    if not f:
+                        continue
+                    for line in f.read().decode("utf-8", errors="ignore").splitlines():
+                        stripped = line.strip()
+                        if stripped.upper().startswith("FROM "):
+                            base = stripped.split()[1].lower()
+                            return not base.startswith("python:")
+        except Exception:
+            pass
+        return False
+
    def _resolve_task_image(
        self, item: Dict[str, Any], task_name: str
    ) -> Tuple[str, Optional[Path]]:
@@ -446,7 +520,9 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
            if dockerfile_path.exists():
                logger.info(
                    "Task %s: building from Dockerfile (force_build=%s, docker_image=%s)",
-                    task_name, self.config.force_build, bool(docker_image),
+                    task_name,
+                    self.config.force_build,
+                    bool(docker_image),
                )
                return str(dockerfile_path), task_dir

@@ -454,12 +530,80 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
        if docker_image:
            logger.warning(
                "Task %s: force_build=True but no environment_tar, "
-                "falling back to docker_image %s", task_name, docker_image,
+                "falling back to docker_image %s",
+                task_name,
+                docker_image,
            )
            return docker_image, None

        return "", None

+    # =========================================================================
+    # Agent loop with format nudging
+    # =========================================================================
+
+    async def _run_with_nudges(
+        self,
+        server,
+        tools: List[Dict[str, Any]],
+        valid_names: set,
+        messages: List[Dict[str, Any]],
+        task_id: str,
+        task_name: str,
+    ) -> Tuple["AgentResult", int]:
+        """Run the agent loop, nudging if the model returns plain text without task_status tag."""
+        total_turns_used = 0
+        nudge_count = 0
+        result = None
+
+        while total_turns_used < self.config.max_agent_turns:
+            remaining = self.config.max_agent_turns - total_turns_used
+            agent = HermesAgentLoop(
+                server=server,
+                tool_schemas=tools,
+                valid_tool_names=valid_names,
+                max_turns=remaining,
+                task_id=task_id,
+                temperature=self.config.agent_temperature,
+                top_p=self.config.agent_top_p,
+                max_tokens=self.config.max_token_length,
+                extra_body=self.config.extra_body,
+            )
+            result = await agent.run(messages)
+            total_turns_used += result.turns_used
+
+            if not result.finished_naturally:
+                break
+
+            last_content = next(
+                (
+                    m.get("content", "") or ""
+                    for m in reversed(messages)
+                    if m.get("role") == "assistant"
+                ),
+                "",
+            )
+            if "<task_status>" in last_content:
+                break
+
+            if nudge_count >= _MAX_FORMAT_NUDGES:
+                logger.warning(
+                    "Task %s: model ignored %d format nudges; stopping.",
+                    task_name,
+                    nudge_count,
+                )
+                break
+            nudge_count += 1
+            logger.info(
+                "Task %s: nudging model (nudge %d/%d) — no tool calls and no task_status",
+                task_name,
+                nudge_count,
+                _MAX_FORMAT_NUDGES,
+            )
+            messages.append({"role": "user", "content": _FORMAT_NUDGE_MESSAGE})
+
+        return result, total_turns_used
+
    # =========================================================================
    # Per-task evaluation -- agent loop + test verification
    # =========================================================================
@@ -488,6 +632,7 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
        task_dir = None  # Set if we extract a Dockerfile (needs cleanup)

        from tqdm import tqdm
+
        tqdm.write(f"  [START] {task_name} (task_id={task_id[:8]})")
        task_start = time.time()

@@ -495,24 +640,32 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
            # --- 1. Resolve Docker image ---
            modal_image, task_dir = self._resolve_task_image(eval_item, task_name)
            if not modal_image:
-                logger.error("Task %s: no docker_image or environment_tar, skipping", task_name)
+                logger.error(
+                    "Task %s: no docker_image or environment_tar, skipping", task_name
+                )
                return {
-                    "passed": False, "reward": 0.0,
-                    "task_name": task_name, "category": category,
+                    "passed": False,
+                    "reward": 0.0,
+                    "task_name": task_name,
+                    "category": category,
                    "error": "no_image",
                }

            # --- 2. Register per-task image override ---
            # Set both modal_image and docker_image so the task image is used
            # regardless of which backend is configured.
-            register_task_env_overrides(task_id, {
+            overrides = {
                "modal_image": modal_image,
                "docker_image": modal_image,
                "cwd": "/app",
-            })
+            }
+            if self._needs_add_python.get(task_name, False):
+                overrides["add_python"] = "3.12"
+            register_task_env_overrides(task_id, overrides)
            logger.info(
                "Task %s: registered image override for task_id %s",
-                task_name, task_id[:8],
+                task_name,
+                task_id[:8],
            )

            # --- 3. Resolve tools and build messages ---
@@ -520,53 +673,48 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):

            messages: List[Dict[str, Any]] = []
            if self.config.system_prompt:
-                messages.append({"role": "system", "content": self.config.system_prompt})
+                messages.append(
+                    {"role": "system", "content": self.config.system_prompt}
+                )
            messages.append({"role": "user", "content": self.format_prompt(eval_item)})

-            # --- 4. Run agent loop ---
-            # Use ManagedServer (Phase 2) for vLLM/SGLang backends to get
-            # token-level tracking via /generate. Falls back to direct
-            # ServerManager (Phase 1) for OpenAI endpoints.
+            # --- 4. Run agent loop with format enforcement ---
+            # The model must either call a tool or end with <task_status>DONE/UNFINISHED</task_status>.
+            # If it returns plain text without the tag, inject a nudge user message and
+            # continue with the remaining turn budget (up to _MAX_FORMAT_NUDGES times).
            if self._use_managed_server():
                async with self.server.managed_server(
                    tokenizer=self.tokenizer,
                    preserve_think_blocks=bool(self.config.thinking_mode),
                ) as managed:
-                    agent = HermesAgentLoop(
+                    result, total_turns_used = await self._run_with_nudges(
                        server=managed,
-                        tool_schemas=tools,
-                        valid_tool_names=valid_names,
-                        max_turns=self.config.max_agent_turns,
+                        tools=tools,
+                        valid_names=valid_names,
+                        messages=messages,
                        task_id=task_id,
-                        temperature=self.config.agent_temperature,
-                        max_tokens=self.config.max_token_length,
-                        extra_body=self.config.extra_body,
-                        budget_config=self.config.build_budget_config(),
+                        task_name=task_name,
                    )
-                    result = await agent.run(messages)
            else:
-                agent = HermesAgentLoop(
+                result, total_turns_used = await self._run_with_nudges(
                    server=self.server,
-                    tool_schemas=tools,
-                    valid_tool_names=valid_names,
-                    max_turns=self.config.max_agent_turns,
+                    tools=tools,
+                    valid_names=valid_names,
+                    messages=messages,
                    task_id=task_id,
-                    temperature=self.config.agent_temperature,
-                    max_tokens=self.config.max_token_length,
-                    extra_body=self.config.extra_body,
-                    budget_config=self.config.build_budget_config(),
+                    task_name=task_name,
                )
-                result = await agent.run(messages)

            # --- 5. Verify -- run test suite in the agent's sandbox ---
            # Skip verification if the agent produced no meaningful output
            only_system_and_user = all(
-                msg.get("role") in ("system", "user") for msg in result.messages
+                msg.get("role") in ("system", "user") for msg in messages
            )
-            if result.turns_used == 0 or only_system_and_user:
+            if total_turns_used == 0 or only_system_and_user:
                logger.warning(
                    "Task %s: agent produced no output (turns=%d). Reward=0.",
-                    task_name, result.turns_used,
+                    task_name,
+                    total_turns_used,
                )
                reward = 0.0
            else:
@@ -578,7 +726,10 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
                    loop = asyncio.get_event_loop()
                    reward = await loop.run_in_executor(
                        None,  # default thread pool
-                        self._run_tests, eval_item, ctx, task_name,
+                        self._run_tests,
+                        eval_item,
+                        ctx,
+                        task_name,
                    )
                except Exception as e:
                    logger.error("Task %s: test verification failed: %s", task_name, e)
@@ -589,20 +740,26 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
            passed = reward == 1.0
            status = "PASS" if passed else "FAIL"
            elapsed = time.time() - task_start
-            tqdm.write(f"  [{status}] {task_name} (turns={result.turns_used}, {elapsed:.0f}s)")
+            tqdm.write(
+                f"  [{status}] {task_name} (turns={total_turns_used}, {elapsed:.0f}s)"
+            )
            logger.info(
                "Task %s: reward=%.1f, turns=%d, finished=%s",
-                task_name, reward, result.turns_used, result.finished_naturally,
+                task_name,
+                reward,
+                total_turns_used,
+                result.finished_naturally,
            )

            out = {
+                **self._run_meta,
                "passed": passed,
                "reward": reward,
                "task_name": task_name,
                "category": category,
-                "turns_used": result.turns_used,
+                "turns_used": total_turns_used,
                "finished_naturally": result.finished_naturally,
-                "messages": result.messages,
+                "messages": messages,
            }
            self._save_result(out)
            return out
@@ -612,8 +769,11 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
            logger.error("Task %s: rollout failed: %s", task_name, e, exc_info=True)
            tqdm.write(f"  [ERROR] {task_name}: {e} ({elapsed:.0f}s)")
            out = {
-                "passed": False, "reward": 0.0,
-                "task_name": task_name, "category": category,
+                **self._run_meta,
+                "passed": False,
+                "reward": 0.0,
+                "task_name": task_name,
+                "category": category,
                "error": str(e),
            }
            self._save_result(out)
@@ -686,7 +846,8 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
        # Execute the test suite
        logger.info(
            "Task %s: running test suite (timeout=%ds)",
-            task_name, self.config.test_timeout,
+            task_name,
+            self.config.test_timeout,
        )
        test_result = ctx.terminal(
            "bash /tests/test.sh",
@@ -719,7 +880,9 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
                        logger.warning(
                            "Task %s: reward.txt content unexpected (%r), "
                            "falling back to exit_code=%d",
-                            task_name, content, exit_code,
+                            task_name,
+                            content,
+                            exit_code,
                        )
                        reward = 1.0 if exit_code == 0 else 0.0
            else:
@@ -727,14 +890,17 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
                logger.warning(
                    "Task %s: reward.txt not found after download, "
                    "falling back to exit_code=%d",
-                    task_name, exit_code,
+                    task_name,
+                    exit_code,
                )
                reward = 1.0 if exit_code == 0 else 0.0
        except Exception as e:
            logger.warning(
                "Task %s: failed to download verifier dir: %s, "
                "falling back to exit_code=%d",
-                task_name, e, exit_code,
+                task_name,
+                e,
+                exit_code,
            )
            reward = 1.0 if exit_code == 0 else 0.0
        finally:
@@ -745,7 +911,9 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
            output_preview = output[-500:] if output else "(no output)"
            logger.info(
                "Task %s: FAIL (exit_code=%d)\n%s",
-                task_name, exit_code, output_preview,
+                task_name,
+                exit_code,
+                output_preview,
            )

        return reward
@@ -770,12 +938,18 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
            )
        except asyncio.TimeoutError:
            from tqdm import tqdm
+
            elapsed = self.config.task_timeout
-            tqdm.write(f"  [TIMEOUT] {task_name} (exceeded {elapsed}s wall-clock limit)")
+            tqdm.write(
+                f"  [TIMEOUT] {task_name} (exceeded {elapsed}s wall-clock limit)"
+            )
            logger.error("Task %s: wall-clock timeout after %ds", task_name, elapsed)
            out = {
-                "passed": False, "reward": 0.0,
-                "task_name": task_name, "category": category,
+                **self._run_meta,
+                "passed": False,
+                "reward": 0.0,
+                "task_name": task_name,
+                "category": category,
                "error": f"timeout ({elapsed}s)",
            }
            self._save_result(out)
@@ -809,23 +983,25 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
                    self.handleError(record)

        handler = _TqdmHandler()
-        handler.setFormatter(logging.Formatter(
-            "%(asctime)s [%(name)s] %(levelname)s: %(message)s",
-            datefmt="%H:%M:%S",
-        ))
+        handler.setFormatter(
+            logging.Formatter(
+                "%(asctime)s [%(name)s] %(levelname)s: %(message)s",
+                datefmt="%H:%M:%S",
+            )
+        )
        root = logging.getLogger()
        root.handlers = [handler]  # Replace any existing handlers
        root.setLevel(logging.INFO)

        # Silence noisy third-party loggers that flood the output
-        logging.getLogger("httpx").setLevel(logging.WARNING)      # Every HTTP request
-        logging.getLogger("openai").setLevel(logging.WARNING)     # OpenAI client retries
-        logging.getLogger("rex-deploy").setLevel(logging.WARNING) # Swerex deployment
+        logging.getLogger("httpx").setLevel(logging.WARNING)  # Every HTTP request
+        logging.getLogger("openai").setLevel(logging.WARNING)  # OpenAI client retries
+        logging.getLogger("rex-deploy").setLevel(logging.WARNING)  # Swerex deployment
        logging.getLogger("rex_image_builder").setLevel(logging.WARNING)  # Image builds

-        print(f"\n{'='*60}")
+        print(f"\n{'=' * 60}")
        print("Starting Terminal-Bench 2.0 Evaluation")
-        print(f"{'='*60}")
+        print(f"{'=' * 60}")
        print(f"  Dataset: {self.config.dataset_name}")
        print(f"  Total tasks: {len(self.all_eval_items)}")
        print(f"  Max agent turns: {self.config.max_agent_turns}")
@@ -833,9 +1009,11 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
        print(f"  Terminal backend: {self.config.terminal_backend}")
        print(f"  Tool thread pool: {self.config.tool_pool_size}")
        print(f"  Terminal timeout: {self.config.terminal_timeout}s/cmd")
-        print(f"  Terminal lifetime: {self.config.terminal_lifetime}s (auto: task_timeout + 120)")
+        print(
+            f"  Terminal lifetime: {self.config.terminal_lifetime}s (auto: task_timeout + 120)"
+        )
        print(f"  Max concurrent tasks: {self.config.max_concurrent_tasks}")
-        print(f"{'='*60}\n")
+        print(f"{'=' * 60}\n")

        # Semaphore to limit concurrent Modal sandbox creations.
        # Without this, all 86 tasks fire simultaneously, each creating a Modal
@@ -877,6 +1055,7 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
            await asyncio.gather(*eval_tasks, return_exceptions=True)
            # Belt-and-suspenders: clean up any remaining sandboxes
            from tools.terminal_tool import cleanup_all_environments
+
            cleanup_all_environments()
            print("All sandboxes cleaned up.")
            return
@@ -922,9 +1101,9 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
        self.eval_metrics = [(k, v) for k, v in eval_metrics.items()]

        # ---- Print summary ----
-        print(f"\n{'='*60}")
+        print(f"\n{'=' * 60}")
        print("Terminal-Bench 2.0 Evaluation Results")
-        print(f"{'='*60}")
+        print(f"{'=' * 60}")
        print(f"Overall Pass Rate: {overall_pass_rate:.4f} ({passed}/{total})")
        print(f"Evaluation Time: {end_time - start_time:.1f} seconds")

@@ -944,7 +1123,7 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
            extra = f" (error: {error})" if error else ""
            print(f"  [{status}] {r['task_name']} (turns={turns}){extra}")

-        print(f"{'='*60}\n")
+        print(f"{'=' * 60}\n")

        # Build sample records for evaluate_log (includes full conversations)
        samples = [
@@ -969,6 +1148,7 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
                end_time=end_time,
                generation_parameters={
                    "temperature": self.config.agent_temperature,
+                    "top_p": self.config.agent_top_p,
                    "max_tokens": self.config.max_token_length,
                    "max_agent_turns": self.config.max_agent_turns,
                    "terminal_backend": self.config.terminal_backend,
@@ -985,6 +1165,7 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
        # Kill all remaining sandboxes. Timed-out tasks leave orphaned thread
        # pool workers still executing commands -- cleanup_all stops them.
        from tools.terminal_tool import cleanup_all_environments
+
        print("\nCleaning up all sandboxes...")
        cleanup_all_environments()

@@ -992,6 +1173,7 @@ class TerminalBench2EvalEnv(HermesAgentBaseEnv):
        # tasks are killed immediately instead of retrying against dead
        # sandboxes and spamming the console with TimeoutError warnings.
        from environments.agent_loop import _tool_executor
+
        _tool_executor.shutdown(wait=False, cancel_futures=True)
        print("Done.")

@@ -115,6 +115,10 @@ class HermesAgentEnvConfig(BaseEnvConfig):
        default=1.0,
        description="Sampling temperature for agent generation during rollouts.",
    )
+    agent_top_p: Optional[float] = Field(
+        default=None,
+        description="Nucleus sampling top_p for agent generation. None = provider default.",
+    )

    # --- Terminal backend ---
    terminal_backend: str = Field(
@@ -529,6 +533,7 @@ class HermesAgentBaseEnv(BaseEnv):
                        max_turns=self.config.max_agent_turns,
                        task_id=task_id,
                        temperature=self.config.agent_temperature,
+                        top_p=self.config.agent_top_p,
                        max_tokens=self.config.max_token_length,
                        extra_body=self.config.extra_body,
                        budget_config=self.config.build_budget_config(),
@@ -547,6 +552,7 @@ class HermesAgentBaseEnv(BaseEnv):
                    max_turns=self.config.max_agent_turns,
                    task_id=task_id,
                    temperature=self.config.agent_temperature,
+                    top_p=self.config.agent_top_p,
                    max_tokens=self.config.max_token_length,
                    extra_body=self.config.extra_body,
                    budget_config=self.config.build_budget_config(),
@@ -561,6 +567,7 @@ class HermesAgentBaseEnv(BaseEnv):
                max_turns=self.config.max_agent_turns,
                task_id=task_id,
                temperature=self.config.agent_temperature,
+                top_p=self.config.agent_top_p,
                max_tokens=self.config.max_token_length,
                extra_body=self.config.extra_body,
                budget_config=self.config.build_budget_config(),
@@ -49,8 +49,6 @@ class HermesToolCallParser(ToolCallParser):
                    continue

                tc_data = json.loads(raw_json)
-                if "name" not in tc_data:
-                    continue
                tool_calls.append(
                    ChatCompletionMessageToolCall(
                        id=f"call_{uuid.uuid4().hex[:8]}",
@@ -89,8 +89,6 @@ class MistralToolCallParser(ToolCallParser):
                        parsed = [parsed]

                    for tc in parsed:
-                        if "name" not in tc:
-                            continue
                        args = tc.get("arguments", {})
                        if isinstance(args, dict):
                            args = json.dumps(args, ensure_ascii=False)
@@ -22,40 +22,20 @@
    },
    "nixpkgs": {
      "locked": {
-        "lastModified": 1775036866,
-        "narHash": "sha256-ZojAnPuCdy657PbTq5V0Y+AHKhZAIwSIT2cb8UgAz/U=",
+        "lastModified": 1751274312,
+        "narHash": "sha256-/bVBlRpECLVzjV19t5KMdMFWSwKLtb5RyXdjz3LJT+g=",
        "owner": "NixOS",
        "repo": "nixpkgs",
-        "rev": "6201e203d09599479a3b3450ed24fa81537ebc4e",
+        "rev": "50ab793786d9de88ee30ec4e4c24fb4236fc2674",
        "type": "github"
      },
      "original": {
        "owner": "NixOS",
-        "ref": "nixos-unstable",
+        "ref": "nixos-24.11",
        "repo": "nixpkgs",
        "type": "github"
      }
    },
-    "npm-lockfile-fix": {
-      "inputs": {
-        "nixpkgs": [
-          "nixpkgs"
-        ]
-      },
-      "locked": {
-        "lastModified": 1775903712,
-        "narHash": "sha256-2GV79U6iVH4gKAPWYrxUReB0S41ty/Y3dBLquU8AlaA=",
-        "owner": "jeslie0",
-        "repo": "npm-lockfile-fix",
-        "rev": "c6093acb0c0548e0f9b8b3d82918823721930fe8",
-        "type": "github"
-      },
-      "original": {
-        "owner": "jeslie0",
-        "repo": "npm-lockfile-fix",
-        "type": "github"
-      }
-    },
    "pyproject-build-systems": {
      "inputs": {
        "nixpkgs": [
@@ -144,7 +124,6 @@
      "inputs": {
        "flake-parts": "flake-parts",
        "nixpkgs": "nixpkgs",
-        "npm-lockfile-fix": "npm-lockfile-fix",
        "pyproject-build-systems": "pyproject-build-systems",
        "pyproject-nix": "pyproject-nix_2",
        "uv2nix": "uv2nix_2"
@@ -2,7 +2,7 @@
  description = "Hermes Agent - AI agent framework by Nous Research";

  inputs = {
-    nixpkgs.url = "github:NixOS/nixpkgs/nixos-unstable";
+    nixpkgs.url = "github:NixOS/nixpkgs/nixos-24.11";
    flake-parts = {
      url = "github:hercules-ci/flake-parts";
      inputs.nixpkgs-lib.follows = "nixpkgs";
@@ -19,20 +19,11 @@
      url = "github:pyproject-nix/build-system-pkgs";
      inputs.nixpkgs.follows = "nixpkgs";
    };
-    npm-lockfile-fix = {
-      url = "github:jeslie0/npm-lockfile-fix";
-      inputs.nixpkgs.follows = "nixpkgs";
-    };
  };

-  outputs =
-    inputs:
+  outputs = inputs:
    inputs.flake-parts.lib.mkFlake { inherit inputs; } {
-      systems = [
-        "x86_64-linux"
-        "aarch64-linux"
-        "aarch64-darwin"
-      ];
+      systems = [ "x86_64-linux" "aarch64-linux" "aarch64-darwin" ];

      imports = [
        ./nix/packages.nix
@@ -18,7 +18,9 @@ suppress delivery.
 """

 import logging
+import os
 import threading
+from pathlib import Path

 logger = logging.getLogger("hooks.boot-md")

@@ -76,15 +76,10 @@ def build_channel_directory(adapters: Dict[Any, Any]) -> Dict[str, Any]:
        except Exception as e:
            logger.warning("Channel directory: failed to build %s: %s", platform.value, e)

-    # Platforms that don't support direct channel enumeration get session-based
-    # discovery automatically.  Skip infrastructure entries that aren't messaging
-    # platforms — everything else falls through to _build_from_sessions().
-    _SKIP_SESSION_DISCOVERY = frozenset({"local", "api_server", "webhook"})
-    for plat in Platform:
-        plat_name = plat.value
-        if plat_name in _SKIP_SESSION_DISCOVERY or plat_name in platforms:
-            continue
-        platforms[plat_name] = _build_from_sessions(plat_name)
+    # Telegram, WhatsApp & Signal can't enumerate chats -- pull from session history
+    for plat_name in ("telegram", "whatsapp", "signal", "email", "sms"):
+        if plat_name not in platforms:
+            platforms[plat_name] = _build_from_sessions(plat_name)

    directory = {
        "updated_at": datetime.now().isoformat(),
@@ -100,7 +95,7 @@ def build_channel_directory(adapters: Dict[Any, Any]) -> Dict[str, Any]:


 def _build_discord(adapter) -> List[Dict[str, str]]:
-    """Enumerate all text channels and forum channels the Discord bot can see."""
+    """Enumerate all text channels the Discord bot can see."""
    channels = []
    client = getattr(adapter, "_client", None)
    if not client:
@@ -119,15 +114,6 @@ def _build_discord(adapter) -> List[Dict[str, str]]:
                "guild": guild.name,
                "type": "channel",
            })
-        # Forum channels (type 15) — creating a message auto-spawns a thread post.
-        forums = getattr(guild, "forum_channels", None) or []
-        for ch in forums:
-            channels.append({
-                "id": str(ch.id),
-                "name": ch.name,
-                "guild": guild.name,
-                "type": "forum",
-            })
        # Also include DM-capable users we've interacted with is not
        # feasible via guild enumeration; those come from sessions.

@@ -200,15 +186,6 @@ def load_directory() -> Dict[str, Any]:
        return {"updated_at": None, "platforms": {}}


-def lookup_channel_type(platform_name: str, chat_id: str) -> Optional[str]:
-    """Return the channel ``type`` string (e.g. ``"channel"``, ``"forum"``) for *chat_id*, or *None* if unknown."""
-    directory = load_directory()
-    for ch in directory.get("platforms", {}).get(platform_name, []):
-        if ch.get("id") == chat_id:
-            return ch.get("type")
-    return None
-
-
 def resolve_channel_name(platform_name: str, name: str) -> Optional[str]:
    """
    Resolve a human-friendly channel name to a numeric ID.
@@ -63,10 +63,6 @@ class Platform(Enum):
    WEBHOOK = "webhook"
    FEISHU = "feishu"
    WECOM = "wecom"
-    WECOM_CALLBACK = "wecom_callback"
-    WEIXIN = "weixin"
-    BLUEBUBBLES = "bluebubbles"
-    QQBOT = "qqbot"


@dataclass
@@ -192,7 +188,7 @@ class StreamingConfig:
    """Configuration for real-time token streaming to messaging platforms."""
    enabled: bool = False
    transport: str = "edit"       # "edit" (progressive editMessageText) or "off"
-    edit_interval: float = 1.0    # Seconds between message edits (Telegram rate-limits at ~1/s)
+    edit_interval: float = 0.3    # Seconds between message edits
    buffer_threshold: int = 40    # Chars before forcing an edit
    cursor: str = " ▉"           # Cursor shown during streaming

@@ -212,7 +208,7 @@ class StreamingConfig:
        return cls(
            enabled=data.get("enabled", False),
            transport=data.get("transport", "edit"),
-            edit_interval=float(data.get("edit_interval", 1.0)),
+            edit_interval=float(data.get("edit_interval", 0.3)),
            buffer_threshold=int(data.get("buffer_threshold", 40)),
            cursor=data.get("cursor", " ▉"),
        )
@@ -258,24 +254,12 @@ class GatewayConfig:
    # Streaming configuration
    streaming: StreamingConfig = field(default_factory=StreamingConfig)

-    # Session store pruning: drop SessionEntry records older than this many
-    # days from the in-memory dict and sessions.json.  Keeps the store from
-    # growing unbounded in gateways serving many chats/threads/users over
-    # months.  Pruning is invisible to users — if they resume, they get a
-    # fresh session exactly as if the reset policy had fired.  0 = disabled.
-    session_store_max_age_days: int = 90
-
    def get_connected_platforms(self) -> List[Platform]:
        """Return list of platforms that are enabled and configured."""
        connected = []
        for platform, config in self.platforms.items():
            if not config.enabled:
                continue
-            # Weixin requires both a token and an account_id
-            if platform == Platform.WEIXIN:
-                if config.extra.get("account_id") and (config.token or config.extra.get("token")):
-                    connected.append(platform)
-                continue
            # Platforms that use token/api_key auth
            if config.token or config.api_key:
                connected.append(platform)
@@ -300,28 +284,9 @@ class GatewayConfig:
            # Feishu uses extra dict for app credentials
            elif platform == Platform.FEISHU and config.extra.get("app_id"):
                connected.append(platform)
-            # WeCom bot mode uses extra dict for bot credentials
+            # WeCom uses extra dict for bot credentials
            elif platform == Platform.WECOM and config.extra.get("bot_id"):
                connected.append(platform)
-            # WeCom callback mode uses corp_id or apps list
-            elif platform == Platform.WECOM_CALLBACK and (
-                config.extra.get("corp_id") or config.extra.get("apps")
-            ):
-                connected.append(platform)
-            # BlueBubbles uses extra dict for local server config
-            elif platform == Platform.BLUEBUBBLES and config.extra.get("server_url") and config.extra.get("password"):
-                connected.append(platform)
-            # QQBot uses extra dict for app credentials
-            elif platform == Platform.QQBOT and config.extra.get("app_id") and config.extra.get("client_secret"):
-                connected.append(platform)
-            # DingTalk uses client_id/client_secret from config.extra or env vars
-            elif platform == Platform.DINGTALK and (
-                config.extra.get("client_id") or os.getenv("DINGTALK_CLIENT_ID")
-            ) and (
-                config.extra.get("client_secret") or os.getenv("DINGTALK_CLIENT_SECRET")
-            ):
-                connected.append(platform)
-        
        return connected
    
    def get_home_channel(self, platform: Platform) -> Optional[HomeChannel]:
@@ -372,7 +337,6 @@ class GatewayConfig:
            "thread_sessions_per_user": self.thread_sessions_per_user,
            "unauthorized_dm_behavior": self.unauthorized_dm_behavior,
            "streaming": self.streaming.to_dict(),
-            "session_store_max_age_days": self.session_store_max_age_days,
        }
    
    @classmethod
@@ -420,13 +384,6 @@ class GatewayConfig:
            "pair",
        )

-        try:
-            session_store_max_age_days = int(data.get("session_store_max_age_days", 90))
-            if session_store_max_age_days < 0:
-                session_store_max_age_days = 0
-        except (TypeError, ValueError):
-            session_store_max_age_days = 90
-
        return cls(
            platforms=platforms,
            default_reset_policy=default_policy,
@@ -441,7 +398,6 @@ class GatewayConfig:
            thread_sessions_per_user=_coerce_bool(thread_sessions_per_user, False),
            unauthorized_dm_behavior=unauthorized_dm_behavior,
            streaming=StreamingConfig.from_dict(data.get("streaming", {})),
-            session_store_max_age_days=session_store_max_age_days,
        )

    def get_unauthorized_dm_behavior(self, platform: Optional[Platform] = None) -> str:
@@ -572,18 +528,8 @@ def load_gateway_config() -> GatewayConfig:
                    bridged["reply_prefix"] = platform_cfg["reply_prefix"]
                if "require_mention" in platform_cfg:
                    bridged["require_mention"] = platform_cfg["require_mention"]
-                if "free_response_channels" in platform_cfg:
-                    bridged["free_response_channels"] = platform_cfg["free_response_channels"]
                if "mention_patterns" in platform_cfg:
                    bridged["mention_patterns"] = platform_cfg["mention_patterns"]
-                if plat == Platform.DISCORD and "channel_skill_bindings" in platform_cfg:
-                    bridged["channel_skill_bindings"] = platform_cfg["channel_skill_bindings"]
-                if "channel_prompts" in platform_cfg:
-                    channel_prompts = platform_cfg["channel_prompts"]
-                    if isinstance(channel_prompts, dict):
-                        bridged["channel_prompts"] = {str(k): v for k, v in channel_prompts.items()}
-                    else:
-                        bridged["channel_prompts"] = channel_prompts
                if not bridged:
                    continue
                plat_data = platforms_data.setdefault(plat.value, {})
@@ -596,19 +542,6 @@ def load_gateway_config() -> GatewayConfig:
                    plat_data["extra"] = extra
                extra.update(bridged)

-            # Slack settings → env vars (env vars take precedence)
-            slack_cfg = yaml_cfg.get("slack", {})
-            if isinstance(slack_cfg, dict):
-                if "require_mention" in slack_cfg and not os.getenv("SLACK_REQUIRE_MENTION"):
-                    os.environ["SLACK_REQUIRE_MENTION"] = str(slack_cfg["require_mention"]).lower()
-                if "allow_bots" in slack_cfg and not os.getenv("SLACK_ALLOW_BOTS"):
-                    os.environ["SLACK_ALLOW_BOTS"] = str(slack_cfg["allow_bots"]).lower()
-                frc = slack_cfg.get("free_response_channels")
-                if frc is not None and not os.getenv("SLACK_FREE_RESPONSE_CHANNELS"):
-                    if isinstance(frc, list):
-                        frc = ",".join(str(v) for v in frc)
-                    os.environ["SLACK_FREE_RESPONSE_CHANNELS"] = str(frc)
-
            # Discord settings → env vars (env vars take precedence)
            discord_cfg = yaml_cfg.get("discord", {})
            if isinstance(discord_cfg, dict):
@@ -629,32 +562,12 @@ def load_gateway_config() -> GatewayConfig:
                    if isinstance(ic, list):
                        ic = ",".join(str(v) for v in ic)
                    os.environ["DISCORD_IGNORED_CHANNELS"] = str(ic)
-                # allowed_channels: if set, bot ONLY responds in these channels (whitelist)
-                ac = discord_cfg.get("allowed_channels")
-                if ac is not None and not os.getenv("DISCORD_ALLOWED_CHANNELS"):
-                    if isinstance(ac, list):
-                        ac = ",".join(str(v) for v in ac)
-                    os.environ["DISCORD_ALLOWED_CHANNELS"] = str(ac)
                # no_thread_channels: channels where bot responds directly without creating thread
                ntc = discord_cfg.get("no_thread_channels")
                if ntc is not None and not os.getenv("DISCORD_NO_THREAD_CHANNELS"):
                    if isinstance(ntc, list):
                        ntc = ",".join(str(v) for v in ntc)
                    os.environ["DISCORD_NO_THREAD_CHANNELS"] = str(ntc)
-                # allow_mentions: granular control over what the bot can ping.
-                # Safe defaults (no @everyone/roles) are applied in the adapter;
-                # these YAML keys only override when set and let users opt back
-                # into unsafe modes (e.g. roles=true) if they actually want it.
-                allow_mentions_cfg = discord_cfg.get("allow_mentions")
-                if isinstance(allow_mentions_cfg, dict):
-                    for yaml_key, env_key in (
-                        ("everyone", "DISCORD_ALLOW_MENTION_EVERYONE"),
-                        ("roles", "DISCORD_ALLOW_MENTION_ROLES"),
-                        ("users", "DISCORD_ALLOW_MENTION_USERS"),
-                        ("replied_user", "DISCORD_ALLOW_MENTION_REPLIED_USER"),
-                    ):
-                        if yaml_key in allow_mentions_cfg and not os.getenv(env_key):
-                            os.environ[env_key] = str(allow_mentions_cfg[yaml_key]).lower()

            # Telegram settings → env vars (env vars take precedence)
            telegram_cfg = yaml_cfg.get("telegram", {})
@@ -669,25 +582,8 @@ def load_gateway_config() -> GatewayConfig:
                    if isinstance(frc, list):
                        frc = ",".join(str(v) for v in frc)
                    os.environ["TELEGRAM_FREE_RESPONSE_CHATS"] = str(frc)
-                ignored_threads = telegram_cfg.get("ignored_threads")
-                if ignored_threads is not None and not os.getenv("TELEGRAM_IGNORED_THREADS"):
-                    if isinstance(ignored_threads, list):
-                        ignored_threads = ",".join(str(v) for v in ignored_threads)
-                    os.environ["TELEGRAM_IGNORED_THREADS"] = str(ignored_threads)
                if "reactions" in telegram_cfg and not os.getenv("TELEGRAM_REACTIONS"):
                    os.environ["TELEGRAM_REACTIONS"] = str(telegram_cfg["reactions"]).lower()
-                if "proxy_url" in telegram_cfg and not os.getenv("TELEGRAM_PROXY"):
-                    os.environ["TELEGRAM_PROXY"] = str(telegram_cfg["proxy_url"]).strip()
-                if "disable_link_previews" in telegram_cfg:
-                    plat_data = platforms_data.setdefault(Platform.TELEGRAM.value, {})
-                    if not isinstance(plat_data, dict):
-                        plat_data = {}
-                        platforms_data[Platform.TELEGRAM.value] = plat_data
-                    extra = plat_data.setdefault("extra", {})
-                    if not isinstance(extra, dict):
-                        extra = {}
-                        plat_data["extra"] = extra
-                    extra["disable_link_previews"] = telegram_cfg["disable_link_previews"]

            whatsapp_cfg = yaml_cfg.get("whatsapp", {})
            if isinstance(whatsapp_cfg, dict):
@@ -701,24 +597,6 @@ def load_gateway_config() -> GatewayConfig:
                        frc = ",".join(str(v) for v in frc)
                    os.environ["WHATSAPP_FREE_RESPONSE_CHATS"] = str(frc)

-            # DingTalk settings → env vars (env vars take precedence)
-            dingtalk_cfg = yaml_cfg.get("dingtalk", {})
-            if isinstance(dingtalk_cfg, dict):
-                if "require_mention" in dingtalk_cfg and not os.getenv("DINGTALK_REQUIRE_MENTION"):
-                    os.environ["DINGTALK_REQUIRE_MENTION"] = str(dingtalk_cfg["require_mention"]).lower()
-                if "mention_patterns" in dingtalk_cfg and not os.getenv("DINGTALK_MENTION_PATTERNS"):
-                    os.environ["DINGTALK_MENTION_PATTERNS"] = json.dumps(dingtalk_cfg["mention_patterns"])
-                frc = dingtalk_cfg.get("free_response_chats")
-                if frc is not None and not os.getenv("DINGTALK_FREE_RESPONSE_CHATS"):
-                    if isinstance(frc, list):
-                        frc = ",".join(str(v) for v in frc)
-                    os.environ["DINGTALK_FREE_RESPONSE_CHATS"] = str(frc)
-                allowed = dingtalk_cfg.get("allowed_users")
-                if allowed is not None and not os.getenv("DINGTALK_ALLOWED_USERS"):
-                    if isinstance(allowed, list):
-                        allowed = ",".join(str(v) for v in allowed)
-                    os.environ["DINGTALK_ALLOWED_USERS"] = str(allowed)
-
            # Matrix settings → env vars (env vars take precedence)
            matrix_cfg = yaml_cfg.get("matrix", {})
            if isinstance(matrix_cfg, dict):
@@ -731,8 +609,6 @@ def load_gateway_config() -> GatewayConfig:
                    os.environ["MATRIX_FREE_RESPONSE_ROOMS"] = str(frc)
                if "auto_thread" in matrix_cfg and not os.getenv("MATRIX_AUTO_THREAD"):
                    os.environ["MATRIX_AUTO_THREAD"] = str(matrix_cfg["auto_thread"]).lower()
-                if "dm_mention_threads" in matrix_cfg and not os.getenv("MATRIX_DM_MENTION_THREADS"):
-                    os.environ["MATRIX_DM_MENTION_THREADS"] = str(matrix_cfg["dm_mention_threads"]).lower()

    except Exception as e:
        logger.warning(
@@ -748,17 +624,6 @@ def load_gateway_config() -> GatewayConfig:
    _apply_env_overrides(config)
    
    # --- Validate loaded values ---
-    _validate_gateway_config(config)
-
-    return config
-
-
-def _validate_gateway_config(config: "GatewayConfig") -> None:
-    """Validate and sanitize a loaded GatewayConfig in place.
-
-    Called by ``load_gateway_config()`` after all config sources are merged.
-    Extracted as a separate function for testability.
-    """
    policy = config.default_reset_policy

    if not (0 <= policy.at_hour <= 23):
@@ -782,7 +647,6 @@ def _validate_gateway_config(config: "GatewayConfig") -> None:
        Platform.SLACK: "SLACK_BOT_TOKEN",
        Platform.MATTERMOST: "MATTERMOST_TOKEN",
        Platform.MATRIX: "MATRIX_ACCESS_TOKEN",
-        Platform.WEIXIN: "WEIXIN_TOKEN",
    }
    for platform, pconfig in config.platforms.items():
        if not pconfig.enabled:
@@ -795,31 +659,7 @@ def _validate_gateway_config(config: "GatewayConfig") -> None:
                platform.value, env_name,
            )

-    # Reject known-weak placeholder tokens.
-    # Ported from openclaw/openclaw#64586: users who copy .env.example
-    # without changing placeholder values get a clear startup error instead
-    # of a confusing "auth failed" from the platform API.
-    try:
-        from hermes_cli.auth import has_usable_secret
-    except ImportError:
-        has_usable_secret = None  # type: ignore[assignment]
-
-    if has_usable_secret is not None:
-        for platform, pconfig in config.platforms.items():
-            if not pconfig.enabled:
-                continue
-            env_name = _token_env_names.get(platform)
-            if not env_name:
-                continue
-            token = pconfig.token
-            if token and token.strip() and not has_usable_secret(token, min_length=4):
-                logger.error(
-                    "%s is enabled but %s is set to a placeholder value ('%s'). "
-                    "Set a real bot token before starting the gateway. "
-                    "The adapter will NOT be started.",
-                    platform.value, env_name, token.strip()[:6] + "...",
-                )
-                pconfig.enabled = False
+    return config


 def _apply_env_overrides(config: GatewayConfig) -> None:
@@ -1042,9 +882,6 @@ def _apply_env_overrides(config: GatewayConfig) -> None:
                pass
        if api_server_host:
            config.platforms[Platform.API_SERVER].extra["host"] = api_server_host
-        api_server_model_name = os.getenv("API_SERVER_MODEL_NAME", "")
-        if api_server_model_name:
-            config.platforms[Platform.API_SERVER].extra["model_name"] = api_server_model_name

    # Webhook platform
    webhook_enabled = os.getenv("WEBHOOK_ENABLED", "").lower() in ("true", "1", "yes")
@@ -1062,25 +899,6 @@ def _apply_env_overrides(config: GatewayConfig) -> None:
        if webhook_secret:
            config.platforms[Platform.WEBHOOK].extra["secret"] = webhook_secret

-    # DingTalk
-    dingtalk_client_id = os.getenv("DINGTALK_CLIENT_ID")
-    dingtalk_client_secret = os.getenv("DINGTALK_CLIENT_SECRET")
-    if dingtalk_client_id and dingtalk_client_secret:
-        if Platform.DINGTALK not in config.platforms:
-            config.platforms[Platform.DINGTALK] = PlatformConfig()
-        config.platforms[Platform.DINGTALK].enabled = True
-        config.platforms[Platform.DINGTALK].extra.update({
-            "client_id": dingtalk_client_id,
-            "client_secret": dingtalk_client_secret,
-        })
-        dingtalk_home = os.getenv("DINGTALK_HOME_CHANNEL")
-        if dingtalk_home:
-            config.platforms[Platform.DINGTALK].home_channel = HomeChannel(
-                platform=Platform.DINGTALK,
-                chat_id=dingtalk_home,
-                name=os.getenv("DINGTALK_HOME_CHANNEL_NAME", "Home"),
-            )
-
    # Feishu / Lark
    feishu_app_id = os.getenv("FEISHU_APP_ID")
    feishu_app_secret = os.getenv("FEISHU_APP_SECRET")
@@ -1130,125 +948,6 @@ def _apply_env_overrides(config: GatewayConfig) -> None:
                name=os.getenv("WECOM_HOME_CHANNEL_NAME", "Home"),
            )

-    # WeCom callback mode (self-built apps)
-    wecom_callback_corp_id = os.getenv("WECOM_CALLBACK_CORP_ID")
-    wecom_callback_corp_secret = os.getenv("WECOM_CALLBACK_CORP_SECRET")
-    if wecom_callback_corp_id and wecom_callback_corp_secret:
-        if Platform.WECOM_CALLBACK not in config.platforms:
-            config.platforms[Platform.WECOM_CALLBACK] = PlatformConfig()
-        config.platforms[Platform.WECOM_CALLBACK].enabled = True
-        config.platforms[Platform.WECOM_CALLBACK].extra.update({
-            "corp_id": wecom_callback_corp_id,
-            "corp_secret": wecom_callback_corp_secret,
-            "agent_id": os.getenv("WECOM_CALLBACK_AGENT_ID", ""),
-            "token": os.getenv("WECOM_CALLBACK_TOKEN", ""),
-            "encoding_aes_key": os.getenv("WECOM_CALLBACK_ENCODING_AES_KEY", ""),
-            "host": os.getenv("WECOM_CALLBACK_HOST", "0.0.0.0"),
-            "port": int(os.getenv("WECOM_CALLBACK_PORT", "8645")),
-        })
-
-    # Weixin (personal WeChat via iLink Bot API)
-    weixin_token = os.getenv("WEIXIN_TOKEN")
-    weixin_account_id = os.getenv("WEIXIN_ACCOUNT_ID")
-    if weixin_token or weixin_account_id:
-        if Platform.WEIXIN not in config.platforms:
-            config.platforms[Platform.WEIXIN] = PlatformConfig()
-        config.platforms[Platform.WEIXIN].enabled = True
-        if weixin_token:
-            config.platforms[Platform.WEIXIN].token = weixin_token
-        extra = config.platforms[Platform.WEIXIN].extra
-        if weixin_account_id:
-            extra["account_id"] = weixin_account_id
-        weixin_base_url = os.getenv("WEIXIN_BASE_URL", "").strip()
-        if weixin_base_url:
-            extra["base_url"] = weixin_base_url.rstrip("/")
-        weixin_cdn_base_url = os.getenv("WEIXIN_CDN_BASE_URL", "").strip()
-        if weixin_cdn_base_url:
-            extra["cdn_base_url"] = weixin_cdn_base_url.rstrip("/")
-        weixin_dm_policy = os.getenv("WEIXIN_DM_POLICY", "").strip().lower()
-        if weixin_dm_policy:
-            extra["dm_policy"] = weixin_dm_policy
-        weixin_group_policy = os.getenv("WEIXIN_GROUP_POLICY", "").strip().lower()
-        if weixin_group_policy:
-            extra["group_policy"] = weixin_group_policy
-        weixin_allowed_users = os.getenv("WEIXIN_ALLOWED_USERS", "").strip()
-        if weixin_allowed_users:
-            extra["allow_from"] = weixin_allowed_users
-        weixin_group_allowed_users = os.getenv("WEIXIN_GROUP_ALLOWED_USERS", "").strip()
-        if weixin_group_allowed_users:
-            extra["group_allow_from"] = weixin_group_allowed_users
-        weixin_split_multiline = os.getenv("WEIXIN_SPLIT_MULTILINE_MESSAGES", "").strip()
-        if weixin_split_multiline:
-            extra["split_multiline_messages"] = weixin_split_multiline
-        weixin_home = os.getenv("WEIXIN_HOME_CHANNEL", "").strip()
-        if weixin_home:
-            config.platforms[Platform.WEIXIN].home_channel = HomeChannel(
-                platform=Platform.WEIXIN,
-                chat_id=weixin_home,
-                name=os.getenv("WEIXIN_HOME_CHANNEL_NAME", "Home"),
-            )
-
-    # BlueBubbles (iMessage)
-    bluebubbles_server_url = os.getenv("BLUEBUBBLES_SERVER_URL")
-    bluebubbles_password = os.getenv("BLUEBUBBLES_PASSWORD")
-    if bluebubbles_server_url and bluebubbles_password:
-        if Platform.BLUEBUBBLES not in config.platforms:
-            config.platforms[Platform.BLUEBUBBLES] = PlatformConfig()
-        config.platforms[Platform.BLUEBUBBLES].enabled = True
-        config.platforms[Platform.BLUEBUBBLES].extra.update({
-            "server_url": bluebubbles_server_url.rstrip("/"),
-            "password": bluebubbles_password,
-            "webhook_host": os.getenv("BLUEBUBBLES_WEBHOOK_HOST", "127.0.0.1"),
-            "webhook_port": int(os.getenv("BLUEBUBBLES_WEBHOOK_PORT", "8645")),
-            "webhook_path": os.getenv("BLUEBUBBLES_WEBHOOK_PATH", "/bluebubbles-webhook"),
-            "send_read_receipts": os.getenv("BLUEBUBBLES_SEND_READ_RECEIPTS", "true").lower() in ("true", "1", "yes"),
-        })
-    bluebubbles_home = os.getenv("BLUEBUBBLES_HOME_CHANNEL")
-    if bluebubbles_home and Platform.BLUEBUBBLES in config.platforms:
-        config.platforms[Platform.BLUEBUBBLES].home_channel = HomeChannel(
-            platform=Platform.BLUEBUBBLES,
-            chat_id=bluebubbles_home,
-            name=os.getenv("BLUEBUBBLES_HOME_CHANNEL_NAME", "Home"),
-        )
-
-    # QQ (Official Bot API v2)
-    qq_app_id = os.getenv("QQ_APP_ID")
-    qq_client_secret = os.getenv("QQ_CLIENT_SECRET")
-    if qq_app_id or qq_client_secret:
-        if Platform.QQBOT not in config.platforms:
-            config.platforms[Platform.QQBOT] = PlatformConfig()
-        config.platforms[Platform.QQBOT].enabled = True
-        extra = config.platforms[Platform.QQBOT].extra
-        if qq_app_id:
-            extra["app_id"] = qq_app_id
-        if qq_client_secret:
-            extra["client_secret"] = qq_client_secret
-        qq_allowed_users = os.getenv("QQ_ALLOWED_USERS", "").strip()
-        if qq_allowed_users:
-            extra["allow_from"] = qq_allowed_users
-        qq_group_allowed = os.getenv("QQ_GROUP_ALLOWED_USERS", "").strip()
-        if qq_group_allowed:
-            extra["group_allow_from"] = qq_group_allowed
-        qq_home = os.getenv("QQBOT_HOME_CHANNEL", "").strip()
-        qq_home_name_env = "QQBOT_HOME_CHANNEL_NAME"
-        if not qq_home:
-            # Back-compat: accept the pre-rename name and log a one-time warning.
-            legacy_home = os.getenv("QQ_HOME_CHANNEL", "").strip()
-            if legacy_home:
-                qq_home = legacy_home
-                qq_home_name_env = "QQ_HOME_CHANNEL_NAME"
-                import logging
-                logging.getLogger(__name__).warning(
-                    "QQ_HOME_CHANNEL is deprecated; rename to QQBOT_HOME_CHANNEL "
-                    "in your .env for consistency with the platform key."
-                )
-        if qq_home:
-            config.platforms[Platform.QQBOT].home_channel = HomeChannel(
-                platform=Platform.QQBOT,
-                chat_id=qq_home,
-                name=os.getenv("QQBOT_HOME_CHANNEL_NAME") or os.getenv(qq_home_name_env, "Home"),
-            )
-
    # Session settings
    idle_minutes = os.getenv("SESSION_IDLE_MINUTES")
    if idle_minutes:
@@ -12,7 +12,7 @@ import logging
 from pathlib import Path
 from datetime import datetime
 from dataclasses import dataclass
-from typing import Dict, List, Optional, Any
+from typing import Dict, List, Optional, Any, Union

 from hermes_cli.config import get_hermes_home

@@ -124,6 +124,53 @@ class DeliveryRouter:
        self.adapters = adapters or {}
        self.output_dir = get_hermes_home() / "cron" / "output"
    
+    def resolve_targets(
+        self,
+        deliver: Union[str, List[str]],
+        origin: Optional[SessionSource] = None
+    ) -> List[DeliveryTarget]:
+        """
+        Resolve delivery specification to concrete targets.
+        
+        Args:
+            deliver: Delivery spec - "origin", "telegram", ["local", "discord"], etc.
+            origin: The source where the request originated (for "origin" target)
+        
+        Returns:
+            List of resolved delivery targets
+        """
+        if isinstance(deliver, str):
+            deliver = [deliver]
+        
+        targets = []
+        seen_platforms = set()
+        
+        for target_str in deliver:
+            target = DeliveryTarget.parse(target_str, origin)
+            
+            # Resolve home channel if needed
+            if target.chat_id is None and target.platform != Platform.LOCAL:
+                home = self.config.get_home_channel(target.platform)
+                if home:
+                    target.chat_id = home.chat_id
+                else:
+                    # No home channel configured, skip this platform
+                    continue
+            
+            # Deduplicate
+            key = (target.platform, target.chat_id, target.thread_id)
+            if key not in seen_platforms:
+                seen_platforms.add(key)
+                targets.append(target)
+        
+        # Always include local if configured
+        if self.config.always_log_local:
+            local_key = (Platform.LOCAL, None, None)
+            if local_key not in seen_platforms:
+                targets.append(DeliveryTarget(platform=Platform.LOCAL))
+        
+        return targets
+    
    async def deliver(
        self,
        content: str,
@@ -252,5 +299,19 @@ class DeliveryRouter:
        return await adapter.send(target.chat_id, content, metadata=send_metadata or None)


+def parse_deliver_spec(
+    deliver: Optional[Union[str, List[str]]],
+    origin: Optional[SessionSource] = None,
+    default: str = "origin"
+) -> Union[str, List[str]]:
+    """
+    Normalize a delivery specification.
+    
+    If None or empty, returns the default.
+    """
+    if not deliver:
+        return default
+    return deliver
+


@@ -1,194 +0,0 @@
-"""Per-platform display/verbosity configuration resolver.
-
-Provides ``resolve_display_setting()`` — the single entry-point for reading
-display settings with platform-specific overrides and sensible defaults.
-
-Resolution order (first non-None wins):
-    1. ``display.platforms.<platform>.<key>``  — explicit per-platform user override
-    2. ``display.<key>``                       — global user setting
-    3. ``_PLATFORM_DEFAULTS[<platform>][<key>]``  — built-in sensible default
-    4. ``_GLOBAL_DEFAULTS[<key>]``              — built-in global default
-
-Exception: ``display.streaming`` is CLI-only.  Gateway streaming follows the
-top-level ``streaming`` config unless ``display.platforms.<platform>.streaming``
-sets an explicit per-platform override.
-
-Backward compatibility: ``display.tool_progress_overrides`` is still read as a
-fallback for ``tool_progress`` when no ``display.platforms`` entry exists.  A
-config migration (version bump) automatically moves the old format into the new
-``display.platforms`` structure.
-"""
-
-from __future__ import annotations
-
-from typing import Any
-
-# ---------------------------------------------------------------------------
-# Overrideable display settings and their global defaults
-# ---------------------------------------------------------------------------
-# These are the settings that can be configured per-platform.
-# Other display settings (compact, personality, skin, etc.) are CLI-only
-# and don't participate in per-platform resolution.
-
-_GLOBAL_DEFAULTS: dict[str, Any] = {
-    "tool_progress": "all",
-    "show_reasoning": False,
-    "tool_preview_length": 0,
-    "streaming": None,  # None = follow top-level streaming config
-}
-
-# ---------------------------------------------------------------------------
-# Sensible per-platform defaults — tiered by platform capability
-# ---------------------------------------------------------------------------
-# Tier 1 (high): Supports message editing, typically personal/team use
-# Tier 2 (medium): Supports editing but often workspace/customer-facing
-# Tier 3 (low): No edit support — each progress msg is permanent
-# Tier 4 (minimal): Batch/non-interactive delivery
-
-_TIER_HIGH = {
-    "tool_progress": "all",
-    "show_reasoning": False,
-    "tool_preview_length": 40,
-    "streaming": None,  # follow global
-}
-
-_TIER_MEDIUM = {
-    "tool_progress": "new",
-    "show_reasoning": False,
-    "tool_preview_length": 40,
-    "streaming": None,
-}
-
-_TIER_LOW = {
-    "tool_progress": "off",
-    "show_reasoning": False,
-    "tool_preview_length": 40,
-    "streaming": False,
-}
-
-_TIER_MINIMAL = {
-    "tool_progress": "off",
-    "show_reasoning": False,
-    "tool_preview_length": 0,
-    "streaming": False,
-}
-
-_PLATFORM_DEFAULTS: dict[str, dict[str, Any]] = {
-    # Tier 1 — full edit support, personal/team use
-    "telegram":    _TIER_HIGH,
-    "discord":     _TIER_HIGH,
-
-    # Tier 2 — edit support, often customer/workspace channels
-    "slack":           _TIER_MEDIUM,
-    "mattermost":      _TIER_MEDIUM,
-    "matrix":          _TIER_MEDIUM,
-    "feishu":          _TIER_MEDIUM,
-
-    # Tier 3 — no edit support, progress messages are permanent
-    "signal":          _TIER_LOW,
-    "whatsapp":        _TIER_MEDIUM,  # Baileys bridge supports /edit
-    "bluebubbles":     _TIER_LOW,
-    "weixin":          _TIER_LOW,
-    "wecom":           _TIER_LOW,
-    "wecom_callback":  _TIER_LOW,
-    "dingtalk":        _TIER_LOW,
-
-    # Tier 4 — batch or non-interactive delivery
-    "email":           _TIER_MINIMAL,
-    "sms":             _TIER_MINIMAL,
-    "webhook":         _TIER_MINIMAL,
-    "homeassistant":   _TIER_MINIMAL,
-    "api_server":      {**_TIER_HIGH, "tool_preview_length": 0},
-}
-
-# Canonical set of per-platform overrideable keys (for validation).
-OVERRIDEABLE_KEYS = frozenset(_GLOBAL_DEFAULTS.keys())
-
-
-def resolve_display_setting(
-    user_config: dict,
-    platform_key: str,
-    setting: str,
-    fallback: Any = None,
-) -> Any:
-    """Resolve a display setting with per-platform override support.
-
-    Parameters
-    ----------
-    user_config : dict
-        The full parsed config.yaml dict.
-    platform_key : str
-        Platform config key (e.g. ``"telegram"``, ``"slack"``).  Use
-        ``_platform_config_key(source.platform)`` from gateway/run.py.
-    setting : str
-        Display setting name (e.g. ``"tool_progress"``, ``"show_reasoning"``).
-    fallback : Any
-        Fallback value when the setting isn't found anywhere.
-
-    Returns
-    -------
-    The resolved value, or *fallback* if nothing is configured.
-    """
-    display_cfg = user_config.get("display") or {}
-
-    # 1. Explicit per-platform override (display.platforms.<platform>.<key>)
-    platforms = display_cfg.get("platforms") or {}
-    plat_overrides = platforms.get(platform_key)
-    if isinstance(plat_overrides, dict):
-        val = plat_overrides.get(setting)
-        if val is not None:
-            return _normalise(setting, val)
-
-    # 1b. Backward compat: display.tool_progress_overrides.<platform>
-    if setting == "tool_progress":
-        legacy = display_cfg.get("tool_progress_overrides")
-        if isinstance(legacy, dict):
-            val = legacy.get(platform_key)
-            if val is not None:
-                return _normalise(setting, val)
-
-    # 2. Global user setting (display.<key>).  Skip display.streaming because
-    # that key controls only CLI terminal streaming; gateway token streaming is
-    # governed by the top-level streaming config plus per-platform overrides.
-    if setting != "streaming":
-        val = display_cfg.get(setting)
-        if val is not None:
-            return _normalise(setting, val)
-
-    # 3. Built-in platform default
-    plat_defaults = _PLATFORM_DEFAULTS.get(platform_key)
-    if plat_defaults:
-        val = plat_defaults.get(setting)
-        if val is not None:
-            return val
-
-    # 4. Built-in global default
-    val = _GLOBAL_DEFAULTS.get(setting)
-    if val is not None:
-        return val
-
-    return fallback
-
-
-# ---------------------------------------------------------------------------
-# Helpers
-# ---------------------------------------------------------------------------
-
-def _normalise(setting: str, value: Any) -> Any:
-    """Normalise YAML quirks (bare ``off`` → False in YAML 1.1)."""
-    if setting == "tool_progress":
-        if value is False:
-            return "off"
-        if value is True:
-            return "all"
-        return str(value).lower()
-    if setting in ("show_reasoning", "streaming"):
-        if isinstance(value, str):
-            return value.lower() in ("true", "1", "yes", "on")
-        return bool(value)
-    if setting == "tool_preview_length":
-        try:
-            return int(value)
-        except (TypeError, ValueError):
-            return 0
-    return value
@@ -9,11 +9,9 @@ Each adapter handles:
 """

 from .base import BasePlatformAdapter, MessageEvent, SendResult
-from .qqbot import QQAdapter

 __all__ = [
    "BasePlatformAdapter",
    "MessageEvent",
    "SendResult",
-    "QQAdapter",
 ]
@@ -6,236 +6,22 @@ and implement the required methods.
 """

 import asyncio
-import ipaddress
 import logging
 import os
 import random
 import re
-import socket as _socket
-import subprocess
-import sys
 import uuid
 from abc import ABC, abstractmethod
 from urllib.parse import urlsplit

 logger = logging.getLogger(__name__)
-
-
-def utf16_len(s: str) -> int:
-    """Count UTF-16 code units in *s*.
-
-    Telegram's message-length limit (4 096) is measured in UTF-16 code units,
-    **not** Unicode code-points.  Characters outside the Basic Multilingual
-    Plane (emoji like 😀, CJK Extension B, musical symbols, …) are encoded as
-    surrogate pairs and therefore consume **two** UTF-16 code units each, even
-    though Python's ``len()`` counts them as one.
-
-    Ported from nearai/ironclaw#2304 which discovered the same discrepancy in
-    Rust's ``chars().count()``.
-    """
-    return len(s.encode("utf-16-le")) // 2
-
-
-def _prefix_within_utf16_limit(s: str, limit: int) -> str:
-    """Return the longest prefix of *s* whose UTF-16 length ≤ *limit*.
-
-    Unlike a plain ``s[:limit]``, this respects surrogate-pair boundaries so
-    we never slice a multi-code-unit character in half.
-    """
-    if utf16_len(s) <= limit:
-        return s
-    # Binary search for the longest safe prefix
-    lo, hi = 0, len(s)
-    while lo < hi:
-        mid = (lo + hi + 1) // 2
-        if utf16_len(s[:mid]) <= limit:
-            lo = mid
-        else:
-            hi = mid - 1
-    return s[:lo]
-
-
-def _custom_unit_to_cp(s: str, budget: int, len_fn) -> int:
-    """Return the largest codepoint offset *n* such that ``len_fn(s[:n]) <= budget``.
-
-    Used by :meth:`BasePlatformAdapter.truncate_message` when *len_fn* measures
-    length in units different from Python codepoints (e.g. UTF-16 code units).
-    Falls back to binary search which is O(log n) calls to *len_fn*.
-    """
-    if len_fn(s) <= budget:
-        return len(s)
-    lo, hi = 0, len(s)
-    while lo < hi:
-        mid = (lo + hi + 1) // 2
-        if len_fn(s[:mid]) <= budget:
-            lo = mid
-        else:
-            hi = mid - 1
-    return lo
-
-
-def is_network_accessible(host: str) -> bool:
-    """Return True if *host* would expose the server beyond loopback.
-
-    Loopback addresses (127.0.0.1, ::1, IPv4-mapped ::ffff:127.0.0.1)
-    are local-only.  Unspecified addresses (0.0.0.0, ::) bind all
-    interfaces.  Hostnames are resolved; DNS failure fails closed.
-    """
-    try:
-        addr = ipaddress.ip_address(host)
-        if addr.is_loopback:
-            return False
-        # ::ffff:127.0.0.1 — Python reports is_loopback=False for mapped
-        # addresses, so check the underlying IPv4 explicitly.
-        if getattr(addr, "ipv4_mapped", None) and addr.ipv4_mapped.is_loopback:
-            return False
-        return True
-    except ValueError:
-        # when host variable is a hostname, we should try to resolve below
-        pass
-
-    try:
-        resolved = _socket.getaddrinfo(
-            host, None, _socket.AF_UNSPEC, _socket.SOCK_STREAM,
-        )
-        # if the hostname resolves into at least one non-loopback address,
-        # then we consider it to be network accessible
-        for _family, _type, _proto, _canonname, sockaddr in resolved:
-            addr = ipaddress.ip_address(sockaddr[0])
-            if not addr.is_loopback:
-                return True
-        return False
-    except (_socket.gaierror, OSError):
-        return True
-
-
-def _detect_macos_system_proxy() -> str | None:
-    """Read the macOS system HTTP(S) proxy via ``scutil --proxy``.
-
-    Returns an ``http://host:port`` URL string if an HTTP or HTTPS proxy is
-    enabled, otherwise *None*.  Falls back silently on non-macOS or on any
-    subprocess error.
-    """
-    if sys.platform != "darwin":
-        return None
-    try:
-        out = subprocess.check_output(
-            ["scutil", "--proxy"], timeout=3, text=True, stderr=subprocess.DEVNULL,
-        )
-    except Exception:
-        return None
-
-    props: dict[str, str] = {}
-    for line in out.splitlines():
-        line = line.strip()
-        if " : " in line:
-            key, _, val = line.partition(" : ")
-            props[key.strip()] = val.strip()
-
-    # Prefer HTTPS, fall back to HTTP
-    for enable_key, host_key, port_key in (
-        ("HTTPSEnable", "HTTPSProxy", "HTTPSPort"),
-        ("HTTPEnable", "HTTPProxy", "HTTPPort"),
-    ):
-        if props.get(enable_key) == "1":
-            host = props.get(host_key)
-            port = props.get(port_key)
-            if host and port:
-                return f"http://{host}:{port}"
-    return None
-
-
-def resolve_proxy_url(platform_env_var: str | None = None) -> str | None:
-    """Return a proxy URL from env vars, or macOS system proxy.
-
-    Check order:
-      0. *platform_env_var* (e.g. ``DISCORD_PROXY``) — highest priority
-      1. HTTPS_PROXY / HTTP_PROXY / ALL_PROXY (and lowercase variants)
-      2. macOS system proxy via ``scutil --proxy`` (auto-detect)
-
-    Returns *None* if no proxy is found.
-    """
-    if platform_env_var:
-        value = (os.environ.get(platform_env_var) or "").strip()
-        if value:
-            return value
-    for key in ("HTTPS_PROXY", "HTTP_PROXY", "ALL_PROXY",
-                "https_proxy", "http_proxy", "all_proxy"):
-        value = (os.environ.get(key) or "").strip()
-        if value:
-            return value
-    return _detect_macos_system_proxy()
-
-
-def proxy_kwargs_for_bot(proxy_url: str | None) -> dict:
-    """Build kwargs for ``commands.Bot()`` / ``discord.Client()`` with proxy.
-
-    Returns:
-      - SOCKS URL  → ``{"connector": ProxyConnector(..., rdns=True)}``
-      - HTTP URL   → ``{"proxy": url}``
-      - *None*     → ``{}``
-
-    ``rdns=True`` forces remote DNS resolution through the proxy — required
-    by many SOCKS implementations (Shadowrocket, Clash) and essential for
-    bypassing DNS pollution behind the GFW.
-    """
-    if not proxy_url:
-        return {}
-    if proxy_url.lower().startswith("socks"):
-        try:
-            from aiohttp_socks import ProxyConnector
-
-            connector = ProxyConnector.from_url(proxy_url, rdns=True)
-            return {"connector": connector}
-        except ImportError:
-            logger.warning(
-                "aiohttp_socks not installed — SOCKS proxy %s ignored. "
-                "Run: pip install aiohttp-socks",
-                proxy_url,
-            )
-            return {}
-    return {"proxy": proxy_url}
-
-
-def proxy_kwargs_for_aiohttp(proxy_url: str | None) -> tuple[dict, dict]:
-    """Build kwargs for standalone ``aiohttp.ClientSession`` with proxy.
-
-    Returns ``(session_kwargs, request_kwargs)`` where:
-      - SOCKS → ``({"connector": ProxyConnector(...)}, {})``
-      - HTTP  → ``({}, {"proxy": url})``
-      - None  → ``({}, {})``
-
-    Usage::
-
-        sess_kw, req_kw = proxy_kwargs_for_aiohttp(proxy_url)
-        async with aiohttp.ClientSession(**sess_kw) as session:
-            async with session.get(url, **req_kw) as resp:
-                ...
-    """
-    if not proxy_url:
-        return {}, {}
-    if proxy_url.lower().startswith("socks"):
-        try:
-            from aiohttp_socks import ProxyConnector
-
-            connector = ProxyConnector.from_url(proxy_url, rdns=True)
-            return {"connector": connector}, {}
-        except ImportError:
-            logger.warning(
-                "aiohttp_socks not installed — SOCKS proxy %s ignored. "
-                "Run: pip install aiohttp-socks",
-                proxy_url,
-            )
-            return {}, {}
-    return {}, {"proxy": proxy_url}
-
-
 from dataclasses import dataclass, field
 from datetime import datetime
 from pathlib import Path
 from typing import Dict, List, Optional, Any, Callable, Awaitable, Tuple
 from enum import Enum

+import sys
 from pathlib import Path as _Path
 sys.path.insert(0, str(_Path(__file__).resolve().parents[2]))

@@ -250,7 +36,7 @@ GATEWAY_SECRET_CAPTURE_UNSUPPORTED_MESSAGE = (
 )


-def safe_url_for_log(url: str, max_len: int = 80) -> str:
+def _safe_url_for_log(url: str, max_len: int = 80) -> str:
    """Return a URL string safe for logs (no query/fragment/userinfo)."""
    if max_len <= 0:
        return ""
@@ -287,23 +73,6 @@ def safe_url_for_log(url: str, max_len: int = 80) -> str:
    return f"{safe[:max_len - 3]}..."


-async def _ssrf_redirect_guard(response):
-    """Re-validate each redirect target to prevent redirect-based SSRF.
-
-    Without this, an attacker can host a public URL that 302-redirects to
-    http://169.254.169.254/ and bypass the pre-flight is_safe_url() check.
-
-    Must be async because httpx.AsyncClient awaits response event hooks.
-    """
-    if response.is_redirect and response.next_request:
-        redirect_url = str(response.next_request.url)
-        from tools.url_safety import is_safe_url
-        if not is_safe_url(redirect_url):
-            raise ValueError(
-                f"Blocked redirect to private/internal address: {safe_url_for_log(redirect_url)}"
-            )
-
-
 # ---------------------------------------------------------------------------
 # Image cache utilities
 #
@@ -323,23 +92,6 @@ def get_image_cache_dir() -> Path:
    return IMAGE_CACHE_DIR


-def _looks_like_image(data: bytes) -> bool:
-    """Return True if *data* starts with a known image magic-byte sequence."""
-    if len(data) < 4:
-        return False
-    if data[:8] == b"\x89PNG\r\n\x1a\n":
-        return True
-    if data[:3] == b"\xff\xd8\xff":
-        return True
-    if data[:6] in (b"GIF87a", b"GIF89a"):
-        return True
-    if data[:2] == b"BM":
-        return True
-    if data[:4] == b"RIFF" and len(data) >= 12 and data[8:12] == b"WEBP":
-        return True
-    return False
-
-
 def cache_image_from_bytes(data: bytes, ext: str = ".jpg") -> str:
    """
    Save raw image bytes to the cache and return the absolute file path.
@@ -350,17 +102,7 @@ def cache_image_from_bytes(data: bytes, ext: str = ".jpg") -> str:

    Returns:
        Absolute path to the cached image file as a string.
-
-    Raises:
-        ValueError: If *data* does not look like a valid image (e.g. an HTML
-            error page returned by the upstream server).
    """
-    if not _looks_like_image(data):
-        snippet = data[:80].decode("utf-8", errors="replace")
-        raise ValueError(
-            f"Refusing to cache non-image data as {ext} "
-            f"(starts with: {snippet!r})"
-        )
    cache_dir = get_image_cache_dir()
    filename = f"img_{uuid.uuid4().hex[:12]}{ext}"
    filepath = cache_dir / filename
@@ -388,7 +130,7 @@ async def cache_image_from_url(url: str, ext: str = ".jpg", retries: int = 2) ->
    """
    from tools.url_safety import is_safe_url
    if not is_safe_url(url):
-        raise ValueError(f"Blocked unsafe URL (SSRF protection): {safe_url_for_log(url)}")
+        raise ValueError(f"Blocked unsafe URL (SSRF protection): {_safe_url_for_log(url)}")

    import asyncio
    import httpx
@@ -396,11 +138,7 @@ async def cache_image_from_url(url: str, ext: str = ".jpg", retries: int = 2) ->
    _log = _logging.getLogger(__name__)

    last_exc = None
-    async with httpx.AsyncClient(
-        timeout=30.0,
-        follow_redirects=True,
-        event_hooks={"response": [_ssrf_redirect_guard]},
-    ) as client:
+    async with httpx.AsyncClient(timeout=30.0, follow_redirects=True) as client:
        for attempt in range(retries + 1):
            try:
                response = await client.get(
@@ -422,7 +160,7 @@ async def cache_image_from_url(url: str, ext: str = ".jpg", retries: int = 2) ->
                        "Media cache retry %d/%d for %s (%.1fs): %s",
                        attempt + 1,
                        retries,
-                        safe_url_for_log(url),
+                        _safe_url_for_log(url),
                        wait,
                        exc,
                    )
@@ -507,7 +245,7 @@ async def cache_audio_from_url(url: str, ext: str = ".ogg", retries: int = 2) ->
    """
    from tools.url_safety import is_safe_url
    if not is_safe_url(url):
-        raise ValueError(f"Blocked unsafe URL (SSRF protection): {safe_url_for_log(url)}")
+        raise ValueError(f"Blocked unsafe URL (SSRF protection): {_safe_url_for_log(url)}")

    import asyncio
    import httpx
@@ -515,11 +253,7 @@ async def cache_audio_from_url(url: str, ext: str = ".ogg", retries: int = 2) ->
    _log = _logging.getLogger(__name__)

    last_exc = None
-    async with httpx.AsyncClient(
-        timeout=30.0,
-        follow_redirects=True,
-        event_hooks={"response": [_ssrf_redirect_guard]},
-    ) as client:
+    async with httpx.AsyncClient(timeout=30.0, follow_redirects=True) as client:
        for attempt in range(retries + 1):
            try:
                response = await client.get(
@@ -541,7 +275,7 @@ async def cache_audio_from_url(url: str, ext: str = ".ogg", retries: int = 2) ->
                        "Audio cache retry %d/%d for %s (%.1fs): %s",
                        attempt + 1,
                        retries,
-                        safe_url_for_log(url),
+                        _safe_url_for_log(url),
                        wait,
                        exc,
                    )
@@ -564,7 +298,6 @@ SUPPORTED_DOCUMENT_TYPES = {
    ".pdf": "application/pdf",
    ".md": "text/markdown",
    ".txt": "text/plain",
-    ".log": "text/plain",
    ".zip": "application/zip",
    ".docx": "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
    ".xlsx": "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
@@ -644,14 +377,6 @@ class MessageType(Enum):
    COMMAND = "command"  # /command style


-class ProcessingOutcome(Enum):
-    """Result classification for message-processing lifecycle hooks."""
-
-    SUCCESS = "success"
-    FAILURE = "failure"
-    CANCELLED = "cancelled"
-
-
@dataclass
 class MessageEvent:
    """
@@ -669,15 +394,6 @@ class MessageEvent:
    # Original platform data
    raw_message: Any = None
    message_id: Optional[str] = None
-
-    # Platform-specific update identifier.  For Telegram this is the
-    # ``update_id`` from the PTB Update wrapper; other platforms currently
-    # ignore it.  Used by ``/restart`` to record the triggering update so the
-    # new gateway can advance the Telegram offset past it and avoid processing
-    # the same ``/restart`` twice if PTB's graceful-shutdown ACK times out
-    # ("Error while calling `get_updates` one more time to mark all fetched
-    # updates" in gateway.log).
-    platform_update_id: Optional[int] = None
    
    # Media attachments
    # media_urls: local file paths (for vision tool access)
@@ -688,18 +404,9 @@ class MessageEvent:
    reply_to_message_id: Optional[str] = None
    reply_to_text: Optional[str] = None  # Text of the replied-to message (for context injection)
    
-    # Auto-loaded skill(s) for topic/channel bindings (e.g., Telegram DM Topics,
-    # Discord channel_skill_bindings).  A single name or ordered list.
-    auto_skill: Optional[str | list[str]] = None
-
-    # Per-channel ephemeral system prompt (e.g. Discord channel_prompts).
-    # Applied at API call time and never persisted to transcript history.
-    channel_prompt: Optional[str] = None
+    # Auto-loaded skill for topic/channel bindings (e.g., Telegram DM Topics)
+    auto_skill: Optional[str] = None
    
-    # Internal flag — set for synthetic events (e.g. background process
-    # completion notifications) that must bypass user authorization checks.
-    internal: bool = False
-
    # Timestamps
    timestamp: datetime = field(default_factory=datetime.now)
    
@@ -716,9 +423,6 @@ class MessageEvent:
        raw = parts[0][1:].lower() if parts else None
        if raw and "@" in raw:
            raw = raw.split("@", 1)[0]
-        # Reject file paths: valid command names never contain /
-        if raw and "/" in raw:
-            return None
        return raw
    
    def get_command_args(self) -> str:
@@ -739,63 +443,6 @@ class SendResult:
    retryable: bool = False  # True for transient connection errors — base will retry automatically


-def merge_pending_message_event(
-    pending_messages: Dict[str, MessageEvent],
-    session_key: str,
-    event: MessageEvent,
-    *,
-    merge_text: bool = False,
-) -> None:
-    """Store or merge a pending event for a session.
-
-    Photo bursts/albums often arrive as multiple near-simultaneous PHOTO
-    events. Merge those into the existing queued event so the next turn sees
-    the whole burst.
-
-    When ``merge_text`` is enabled, rapid follow-up TEXT events are appended
-    instead of replacing the pending turn. This is used for Telegram bursty
-    follow-ups so a multi-part user thought is not silently truncated to only
-    the last queued fragment.
-    """
-    existing = pending_messages.get(session_key)
-    if existing:
-        existing_is_photo = getattr(existing, "message_type", None) == MessageType.PHOTO
-        incoming_is_photo = event.message_type == MessageType.PHOTO
-        existing_has_media = bool(existing.media_urls)
-        incoming_has_media = bool(event.media_urls)
-
-        if existing_is_photo and incoming_is_photo:
-            existing.media_urls.extend(event.media_urls)
-            existing.media_types.extend(event.media_types)
-            if event.text:
-                existing.text = BasePlatformAdapter._merge_caption(existing.text, event.text)
-            return
-
-        if existing_has_media or incoming_has_media:
-            if incoming_has_media:
-                existing.media_urls.extend(event.media_urls)
-                existing.media_types.extend(event.media_types)
-            if event.text:
-                if existing.text:
-                    existing.text = BasePlatformAdapter._merge_caption(existing.text, event.text)
-                else:
-                    existing.text = event.text
-            if existing_is_photo or incoming_is_photo:
-                existing.message_type = MessageType.PHOTO
-            return
-
-        if (
-            merge_text
-            and getattr(existing, "message_type", None) == MessageType.TEXT
-            and event.message_type == MessageType.TEXT
-        ):
-            if event.text:
-                existing.text = f"{existing.text}\n{event.text}" if existing.text else event.text
-            return
-
-    pending_messages[session_key] = event
-
-
 # Error substrings that indicate a transient *connection* failure worth retrying.
 # "timeout" / "timed out" / "readtimeout" / "writetimeout" are intentionally
 # excluded: a read/write timeout on a non-idempotent call (e.g. send_message)
@@ -820,36 +467,6 @@ _RETRYABLE_ERROR_PATTERNS = (
 MessageHandler = Callable[[MessageEvent], Awaitable[Optional[str]]]


-def resolve_channel_prompt(
-    config_extra: dict,
-    channel_id: str,
-    parent_id: str | None = None,
-) -> str | None:
-    """Resolve a per-channel ephemeral prompt from platform config.
-
-    Looks up ``channel_prompts`` in the adapter's ``config.extra`` dict.
-    Prefers an exact match on *channel_id*; falls back to *parent_id*
-    (useful for forum threads / child channels inheriting a parent prompt).
-
-    Returns the prompt string, or None if no match is found.  Blank/whitespace-
-    only prompts are treated as absent.
-    """
-    prompts = config_extra.get("channel_prompts") or {}
-    if not isinstance(prompts, dict):
-        return None
-
-    for key in (channel_id, parent_id):
-        if not key:
-            continue
-        prompt = prompts.get(key)
-        if prompt is None:
-            continue
-        prompt = str(prompt).strip()
-        if prompt:
-            return prompt
-    return None
-
-
 class BasePlatformAdapter(ABC):
    """
    Base class for platform adapters.
@@ -879,13 +496,6 @@ class BasePlatformAdapter(ABC):
        # Gateway shutdown cancels these so an old gateway instance doesn't keep
        # working on a task after --replace or manual restarts.
        self._background_tasks: set[asyncio.Task] = set()
-        # One-shot callbacks to fire after the main response is delivered.
-        # Keyed by session_key.  GatewayRunner uses this to defer
-        # background-review notifications ("💾 Skill created") until the
-        # primary reply has been sent.
-        self._post_delivery_callbacks: Dict[str, Callable] = {}
-        self._expected_cancelled_tasks: set[asyncio.Task] = set()
-        self._busy_session_handler: Optional[Callable[[MessageEvent, str], Awaitable[bool]]] = None
        # Chats where auto-TTS on voice input is disabled (set by /voice off)
        self._auto_tts_disabled_chats: set = set()
        # Chats where typing indicator is paused (e.g. during approval waits).
@@ -955,36 +565,7 @@ class BasePlatformAdapter(ABC):
        result = handler(self)
        if asyncio.iscoroutine(result):
            await result
-
-    def _acquire_platform_lock(self, scope: str, identity: str, resource_desc: str) -> bool:
-        """Acquire a scoped lock for this adapter. Returns True on success."""
-        from gateway.status import acquire_scoped_lock
-        self._platform_lock_scope = scope
-        self._platform_lock_identity = identity
-        acquired, existing = acquire_scoped_lock(
-            scope, identity, metadata={'platform': self.platform.value}
-        )
-        if acquired:
-            return True
-        owner_pid = existing.get('pid') if isinstance(existing, dict) else None
-        message = (
-            f'{resource_desc} already in use'
-            + (f' (PID {owner_pid})' if owner_pid else '')
-            + '. Stop the other gateway first.'
-        )
-        logger.error('[%s] %s', self.name, message)
-        self._set_fatal_error(f'{scope}_lock', message, retryable=False)
-        return False
-
-    def _release_platform_lock(self) -> None:
-        """Release the scoped lock acquired by _acquire_platform_lock."""
-        identity = getattr(self, '_platform_lock_identity', None)
-        if not identity:
-            return
-        from gateway.status import release_scoped_lock
-        release_scoped_lock(self._platform_lock_scope, identity)
-        self._platform_lock_identity = None
-
+    
    @property
    def name(self) -> str:
        """Human-readable name for this adapter."""
@@ -1003,10 +584,6 @@ class BasePlatformAdapter(ABC):
        an optional response string.
        """
        self._message_handler = handler
-
-    def set_busy_session_handler(self, handler: Optional[Callable[[MessageEvent, str], Awaitable[bool]]]) -> None:
-        """Set an optional handler for messages arriving during active sessions."""
-        self._busy_session_handler = handler
    
    def set_session_store(self, session_store: Any) -> None:
        """
@@ -1054,40 +631,16 @@ class BasePlatformAdapter(ABC):
        """
        pass

-    # Default: the adapter treats ``finalize=True`` on edit_message as a
-    # no-op and is happy to have the stream consumer skip redundant final
-    # edits.  Subclasses that *require* an explicit finalize call to close
-    # out the message lifecycle (e.g. rich card / AI assistant surfaces
-    # such as DingTalk AI Cards) override this to True (class attribute or
-    # property) so the stream consumer knows not to short-circuit.
-    REQUIRES_EDIT_FINALIZE: bool = False
-
    async def edit_message(
        self,
        chat_id: str,
        message_id: str,
        content: str,
-        *,
-        finalize: bool = False,
    ) -> SendResult:
        """
        Edit a previously sent message. Optional — platforms that don't
        support editing return success=False and callers fall back to
        sending a new message.
-
-        ``finalize`` signals that this is the last edit in a streaming
-        sequence.  Most platforms (Telegram, Slack, Discord, Matrix,
-        etc.) treat it as a no-op because their edit APIs have no notion
-        of message lifecycle state — an edit is an edit.  Platforms that
-        render streaming updates with a distinct "in progress" state and
-        require explicit closure (e.g. rich card / AI assistant surfaces
-        such as DingTalk AI Cards) use it to finalize the message and
-        transition the UI out of the streaming indicator — those should
-        also set ``REQUIRES_EDIT_FINALIZE = True`` so callers route a
-        final edit through even when content is unchanged.  Callers
-        should set ``finalize=True`` on the final edit of a streamed
-        response (typically when ``got_done`` fires in the stream
-        consumer) and leave it ``False`` on intermediate edits.
        """
        return SendResult(success=False, error="Not supported")

@@ -1324,7 +877,7 @@ class BasePlatformAdapter(ABC):
                path = path[1:-1].strip()
            path = path.lstrip("`\"'").rstrip("`\"',.;:)}]")
            if path:
-                media.append((os.path.expanduser(path), has_voice_tag))
+                media.append((path, has_voice_tag))

        # Remove MEDIA tags from content (including surrounding quote/backtick wrappers)
        if media:
@@ -1451,7 +1004,7 @@ class BasePlatformAdapter(ABC):
    async def on_processing_start(self, event: MessageEvent) -> None:
        """Hook called when background processing begins."""

-    async def on_processing_complete(self, event: MessageEvent, outcome: ProcessingOutcome) -> None:
+    async def on_processing_complete(self, event: MessageEvent, success: bool) -> None:
        """Hook called when background processing completes."""

    async def _run_processing_hook(self, hook_name: str, *args: Any, **kwargs: Any) -> None:
@@ -1612,9 +1165,7 @@ class BasePlatformAdapter(ABC):
            # session lifecycle and its cleanup races with the running task
            # (see PR #4926).
            cmd = event.get_command()
-            from hermes_cli.commands import should_bypass_active_session
-
-            if should_bypass_active_session(cmd):
+            if cmd in ("approve", "deny", "status", "stop", "new", "reset"):
                logger.debug(
                    "[%s] Command '/%s' bypassing active-session guard for %s",
                    self.name, cmd, session_key,
@@ -1633,19 +1184,19 @@ class BasePlatformAdapter(ABC):
                    logger.error("[%s] Command '/%s' dispatch failed: %s", self.name, cmd, e, exc_info=True)
                return

-            if self._busy_session_handler is not None:
-                try:
-                    if await self._busy_session_handler(event, session_key):
-                        return
-                except Exception as e:
-                    logger.error("[%s] Busy-session handler failed: %s", self.name, e, exc_info=True)
-
            # Special case: photo bursts/albums frequently arrive as multiple near-
            # simultaneous messages. Queue them without interrupting the active run,
            # then process them immediately after the current task finishes.
            if event.message_type == MessageType.PHOTO:
                logger.debug("[%s] Queuing photo follow-up for session %s without interrupt", self.name, session_key)
-                merge_pending_message_event(self._pending_messages, session_key, event)
+                existing = self._pending_messages.get(session_key)
+                if existing and existing.message_type == MessageType.PHOTO:
+                    existing.media_urls.extend(event.media_urls)
+                    existing.media_types.extend(event.media_types)
+                    if event.text:
+                        existing.text = self._merge_caption(existing.text, event.text)
+                else:
+                    self._pending_messages[session_key] = event
                return  # Don't interrupt now - will run after current task completes

            # Default behavior for non-photo follow-ups: interrupt the running agent
@@ -1672,7 +1223,6 @@ class BasePlatformAdapter(ABC):
            return
        if hasattr(task, "add_done_callback"):
            task.add_done_callback(self._background_tasks.discard)
-            task.add_done_callback(self._expected_cancelled_tasks.discard)
    
    @staticmethod
    def _get_human_delay() -> float:
@@ -1729,21 +1279,6 @@ class BasePlatformAdapter(ABC):
            # streaming already delivered the text (already_sent=True) or
            # when the message was queued behind an active agent.  Log at
            # DEBUG to avoid noisy warnings for expected behavior.
-            #
-            # Suppress stale response when the session was interrupted by a
-            # new message that hasn't been consumed yet.  The pending message
-            # is processed by the pending-message handler below (#8221/#2483).
-            if (
-                response
-                and interrupt_event.is_set()
-                and session_key in self._pending_messages
-            ):
-                logger.info(
-                    "[%s] Suppressing stale response for interrupted session %s",
-                    self.name,
-                    session_key,
-                )
-                response = None
            if not response:
                logger.debug("[%s] Handler returned empty/None response for %s", self.name, event.source.chat_id)
            if response:
@@ -1824,7 +1359,7 @@ class BasePlatformAdapter(ABC):
                        logger.info(
                            "[%s] Sending image: %s (alt=%s)",
                            self.name,
-                            safe_url_for_log(image_url),
+                            _safe_url_for_log(image_url),
                            alt_text[:30] if alt_text else "",
                        )
                        # Route animated GIFs through send_animation for proper playback
@@ -1916,28 +1451,15 @@ class BasePlatformAdapter(ABC):

            # Determine overall success for the processing hook
            processing_ok = delivery_succeeded if delivery_attempted else not bool(response)
-            await self._run_processing_hook(
-                "on_processing_complete",
-                event,
-                ProcessingOutcome.SUCCESS if processing_ok else ProcessingOutcome.FAILURE,
-            )
+            await self._run_processing_hook("on_processing_complete", event, processing_ok)

            # Check if there's a pending message that was queued during our processing
            if session_key in self._pending_messages:
                pending_event = self._pending_messages.pop(session_key)
                logger.debug("[%s] Processing queued message from interrupt", self.name)
-                # Keep the _active_sessions entry live across the turn chain
-                # and only CLEAR the interrupt Event — do NOT delete the entry.
-                # If we deleted here, a concurrent inbound message arriving
-                # during the awaits below would pass the Level-1 guard, spawn
-                # its own _process_message_background, and run simultaneously
-                # with the recursive drain below.  Two agents on one
-                # session_key = duplicate responses, duplicate tool calls.
-                # Clearing the Event keeps the guard live so follow-ups take
-                # the busy-handler path (queue + interrupt) as intended.
-                _active = self._active_sessions.get(session_key)
-                if _active is not None:
-                    _active.clear()
+                # Clean up current session before processing pending
+                if session_key in self._active_sessions:
+                    del self._active_sessions[session_key]
                typing_task.cancel()
                try:
                    await typing_task
@@ -1948,14 +1470,10 @@ class BasePlatformAdapter(ABC):
                return  # Already cleaned up
                
        except asyncio.CancelledError:
-            current_task = asyncio.current_task()
-            outcome = ProcessingOutcome.CANCELLED
-            if current_task is None or current_task not in self._expected_cancelled_tasks:
-                outcome = ProcessingOutcome.FAILURE
-            await self._run_processing_hook("on_processing_complete", event, outcome)
+            await self._run_processing_hook("on_processing_complete", event, False)
            raise
        except Exception as e:
-            await self._run_processing_hook("on_processing_complete", event, ProcessingOutcome.FAILURE)
+            await self._run_processing_hook("on_processing_complete", event, False)
            logger.error("[%s] Error handling message: %s", self.name, e, exc_info=True)
            # Send the error to the user so they aren't left with radio silence
            try:
@@ -1974,14 +1492,6 @@ class BasePlatformAdapter(ABC):
            except Exception:
                pass  # Last resort — don't let error reporting crash the handler
        finally:
-            # Fire any one-shot post-delivery callback registered for this
-            # session (e.g. deferred background-review notifications).
-            _post_cb = getattr(self, "_post_delivery_callbacks", {}).pop(session_key, None)
-            if callable(_post_cb):
-                try:
-                    _post_cb()
-                except Exception:
-                    pass
            # Stop typing indicator
            typing_task.cancel()
            try:
@@ -1995,34 +1505,6 @@ class BasePlatformAdapter(ABC):
                    await self.stop_typing(event.source.chat_id)
            except Exception:
                pass
-            # Late-arrival drain: a message may have arrived during the
-            # cleanup awaits above (typing_task cancel, stop_typing).  Such
-            # messages passed the Level-1 guard (entry still live, Event
-            # possibly set) and landed in _pending_messages via the
-            # busy-handler path.  Without this block, we would delete the
-            # active-session entry and the queued message would be silently
-            # dropped (user never gets a reply).
-            late_pending = self._pending_messages.pop(session_key, None)
-            if late_pending is not None:
-                logger.debug(
-                    "[%s] Late-arrival pending message during cleanup — spawning drain task",
-                    self.name,
-                )
-                _active = self._active_sessions.get(session_key)
-                if _active is not None:
-                    _active.clear()
-                drain_task = asyncio.create_task(
-                    self._process_message_background(late_pending, session_key)
-                )
-                try:
-                    self._background_tasks.add(drain_task)
-                    drain_task.add_done_callback(self._background_tasks.discard)
-                except TypeError:
-                    # Tests stub create_task() with non-hashable sentinels; tolerate.
-                    pass
-                # Leave _active_sessions[session_key] populated — the drain
-                # task's own lifecycle will clean it up.
-                return
            # Clean up session tracking
            if session_key in self._active_sessions:
                del self._active_sessions[session_key]
@@ -2035,12 +1517,10 @@ class BasePlatformAdapter(ABC):
        """
        tasks = [task for task in self._background_tasks if not task.done()]
        for task in tasks:
-            self._expected_cancelled_tasks.add(task)
            task.cancel()
        if tasks:
            await asyncio.gather(*tasks, return_exceptions=True)
        self._background_tasks.clear()
-        self._expected_cancelled_tasks.clear()
        self._pending_messages.clear()
        self._active_sessions.clear()

@@ -2063,7 +1543,6 @@ class BasePlatformAdapter(ABC):
        chat_topic: Optional[str] = None,
        user_id_alt: Optional[str] = None,
        chat_id_alt: Optional[str] = None,
-        is_bot: bool = False,
    ) -> SessionSource:
        """Helper to build a SessionSource for this platform."""
        # Normalize empty topic to None
@@ -2080,7 +1559,6 @@ class BasePlatformAdapter(ABC):
            chat_topic=chat_topic.strip() if chat_topic else None,
            user_id_alt=user_id_alt,
            chat_id_alt=chat_id_alt,
-            is_bot=is_bot,
        )
    
    @abstractmethod
@@ -2106,11 +1584,7 @@ class BasePlatformAdapter(ABC):
        return content
    
    @staticmethod
-    def truncate_message(
-        content: str,
-        max_length: int = 4096,
-        len_fn: Optional["Callable[[str], int]"] = None,
-    ) -> List[str]:
+    def truncate_message(content: str, max_length: int = 4096) -> List[str]:
        """
        Split a long message into chunks, preserving code block boundaries.

@@ -2122,16 +1596,11 @@ class BasePlatformAdapter(ABC):
        Args:
            content: The full message content
            max_length: Maximum length per chunk (platform-specific)
-            len_fn: Optional length function for measuring string length.
-                     Defaults to ``len`` (Unicode code-points).  Pass
-                     ``utf16_len`` for platforms that measure message
-                     length in UTF-16 code units (e.g. Telegram).

        Returns:
            List of message chunks
        """
-        _len = len_fn or len
-        if _len(content) <= max_length:
+        if len(content) <= max_length:
            return [content]

        INDICATOR_RESERVE = 10   # room for " (XX/XX)"
@@ -2150,33 +1619,22 @@ class BasePlatformAdapter(ABC):

            # How much body text we can fit after accounting for the prefix,
            # a potential closing fence, and the chunk indicator.
-            headroom = max_length - INDICATOR_RESERVE - _len(prefix) - _len(FENCE_CLOSE)
+            headroom = max_length - INDICATOR_RESERVE - len(prefix) - len(FENCE_CLOSE)
            if headroom < 1:
                headroom = max_length // 2

            # Everything remaining fits in one final chunk
-            if _len(prefix) + _len(remaining) <= max_length - INDICATOR_RESERVE:
+            if len(prefix) + len(remaining) <= max_length - INDICATOR_RESERVE:
                chunks.append(prefix + remaining)
                break

-            # Find a natural split point (prefer newlines, then spaces).
-            # When _len != len (e.g. utf16_len for Telegram), headroom is
-            # measured in the custom unit.  We need codepoint-based slice
-            # positions that stay within the custom-unit budget.
-            #
-            # _safe_slice_pos() maps a custom-unit budget to the largest
-            # codepoint offset whose custom length ≤ budget.
-            if _len is not len:
-                # Map headroom (custom units) → codepoint slice length
-                _cp_limit = _custom_unit_to_cp(remaining, headroom, _len)
-            else:
-                _cp_limit = headroom
-            region = remaining[:_cp_limit]
+            # Find a natural split point (prefer newlines, then spaces)
+            region = remaining[:headroom]
            split_at = region.rfind("\n")
-            if split_at < _cp_limit // 2:
+            if split_at < headroom // 2:
                split_at = region.rfind(" ")
            if split_at < 1:
-                split_at = _cp_limit
+                split_at = headroom

            # Avoid splitting inside an inline code span (`...`).
            # If the text before split_at has an odd number of unescaped
@@ -2196,7 +1654,7 @@ class BasePlatformAdapter(ABC):
                    safe_split = candidate.rfind(" ", 0, last_bt)
                    nl_split = candidate.rfind("\n", 0, last_bt)
                    safe_split = max(safe_split, nl_split)
-                    if safe_split > _cp_limit // 4:
+                    if safe_split > headroom // 4:
                        split_at = safe_split

            chunk_body = remaining[:split_at]
@@ -1,918 +0,0 @@
-"""BlueBubbles iMessage platform adapter.
-
-Uses the local BlueBubbles macOS server for outbound REST sends and inbound
-webhooks.  Supports text messaging, media attachments (images, voice, video,
-documents), tapback reactions, typing indicators, and read receipts.
-
-Architecture based on PR #5869 (benjaminsehl) with inbound attachment
-downloading from PR #4588 (YuhangLin).
-"""
-
-import asyncio
-import json
-import logging
-import os
-import re
-import uuid
-from datetime import datetime
-from typing import Any, Dict, List, Optional
-from urllib.parse import quote
-
-import httpx
-
-from gateway.config import Platform, PlatformConfig
-from gateway.platforms.base import (
-    BasePlatformAdapter,
-    MessageEvent,
-    MessageType,
-    SendResult,
-    cache_image_from_bytes,
-    cache_audio_from_bytes,
-    cache_document_from_bytes,
-)
-from gateway.platforms.helpers import strip_markdown
-
-logger = logging.getLogger(__name__)
-
-# ---------------------------------------------------------------------------
-# Constants
-# ---------------------------------------------------------------------------
-
-DEFAULT_WEBHOOK_HOST = "127.0.0.1"
-DEFAULT_WEBHOOK_PORT = 8645
-DEFAULT_WEBHOOK_PATH = "/bluebubbles-webhook"
-MAX_TEXT_LENGTH = 4000
-
-# Tapback reaction codes (BlueBubbles associatedMessageType values)
-_TAPBACK_ADDED = {
-    2000: "love", 2001: "like", 2002: "dislike",
-    2003: "laugh", 2004: "emphasize", 2005: "question",
-}
-_TAPBACK_REMOVED = {
-    3000: "love", 3001: "like", 3002: "dislike",
-    3003: "laugh", 3004: "emphasize", 3005: "question",
-}
-
-# Webhook event types that carry user messages
-_MESSAGE_EVENTS = {"new-message", "message", "updated-message"}
-
-# Log redaction patterns
-_PHONE_RE = re.compile(r"\+?\d{7,15}")
-_EMAIL_RE = re.compile(r"[\w.+-]+@[\w-]+\.[\w.]+")
-
-
-def _redact(text: str) -> str:
-    """Redact phone numbers and emails from log output."""
-    text = _PHONE_RE.sub("[REDACTED]", text)
-    text = _EMAIL_RE.sub("[REDACTED]", text)
-    return text
-
-
-# ---------------------------------------------------------------------------
-# Helpers
-# ---------------------------------------------------------------------------
-
-def check_bluebubbles_requirements() -> bool:
-    try:
-        import aiohttp  # noqa: F401
-        import httpx as _httpx  # noqa: F401
-    except ImportError:
-        return False
-    return True
-
-
-def _normalize_server_url(raw: str) -> str:
-    value = (raw or "").strip()
-    if not value:
-        return ""
-    if not re.match(r"^https?://", value, flags=re.I):
-        value = f"http://{value}"
-    return value.rstrip("/")
-
-
-
-
-
-# ---------------------------------------------------------------------------
-# Adapter
-# ---------------------------------------------------------------------------
-
-class BlueBubblesAdapter(BasePlatformAdapter):
-    platform = Platform.BLUEBUBBLES
-    MAX_MESSAGE_LENGTH = MAX_TEXT_LENGTH
-
-    def __init__(self, config: PlatformConfig):
-        super().__init__(config, Platform.BLUEBUBBLES)
-        extra = config.extra or {}
-        self.server_url = _normalize_server_url(
-            extra.get("server_url") or os.getenv("BLUEBUBBLES_SERVER_URL", "")
-        )
-        self.password = extra.get("password") or os.getenv("BLUEBUBBLES_PASSWORD", "")
-        self.webhook_host = (
-            extra.get("webhook_host")
-            or os.getenv("BLUEBUBBLES_WEBHOOK_HOST", DEFAULT_WEBHOOK_HOST)
-        )
-        self.webhook_port = int(
-            extra.get("webhook_port")
-            or os.getenv("BLUEBUBBLES_WEBHOOK_PORT", str(DEFAULT_WEBHOOK_PORT))
-        )
-        self.webhook_path = (
-            extra.get("webhook_path")
-            or os.getenv("BLUEBUBBLES_WEBHOOK_PATH", DEFAULT_WEBHOOK_PATH)
-        )
-        if not str(self.webhook_path).startswith("/"):
-            self.webhook_path = f"/{self.webhook_path}"
-        self.send_read_receipts = bool(extra.get("send_read_receipts", True))
-        self.client: Optional[httpx.AsyncClient] = None
-        self._runner = None
-        self._private_api_enabled: Optional[bool] = None
-        self._helper_connected: bool = False
-        self._guid_cache: Dict[str, str] = {}
-
-    # ------------------------------------------------------------------
-    # API helpers
-    # ------------------------------------------------------------------
-
-    def _api_url(self, path: str) -> str:
-        sep = "&" if "?" in path else "?"
-        return f"{self.server_url}{path}{sep}password={quote(self.password, safe='')}"
-
-    async def _api_get(self, path: str) -> Dict[str, Any]:
-        assert self.client is not None
-        res = await self.client.get(self._api_url(path))
-        res.raise_for_status()
-        return res.json()
-
-    async def _api_post(self, path: str, payload: Dict[str, Any]) -> Dict[str, Any]:
-        assert self.client is not None
-        res = await self.client.post(self._api_url(path), json=payload)
-        res.raise_for_status()
-        return res.json()
-
-    # ------------------------------------------------------------------
-    # Lifecycle
-    # ------------------------------------------------------------------
-
-    async def connect(self) -> bool:
-        if not self.server_url or not self.password:
-            logger.error(
-                "[bluebubbles] BLUEBUBBLES_SERVER_URL and BLUEBUBBLES_PASSWORD are required"
-            )
-            return False
-        from aiohttp import web
-
-        self.client = httpx.AsyncClient(timeout=30.0)
-        try:
-            await self._api_get("/api/v1/ping")
-            info = await self._api_get("/api/v1/server/info")
-            server_data = (info or {}).get("data", {})
-            self._private_api_enabled = bool(server_data.get("private_api"))
-            self._helper_connected = bool(server_data.get("helper_connected"))
-            logger.info(
-                "[bluebubbles] connected to %s (private_api=%s, helper=%s)",
-                self.server_url,
-                self._private_api_enabled,
-                self._helper_connected,
-            )
-        except Exception as exc:
-            logger.error(
-                "[bluebubbles] cannot reach server at %s: %s", self.server_url, exc
-            )
-            if self.client:
-                await self.client.aclose()
-                self.client = None
-            return False
-
-        app = web.Application()
-        app.router.add_get("/health", lambda _: web.Response(text="ok"))
-        app.router.add_post(self.webhook_path, self._handle_webhook)
-        self._runner = web.AppRunner(app)
-        await self._runner.setup()
-        site = web.TCPSite(self._runner, self.webhook_host, self.webhook_port)
-        await site.start()
-        self._mark_connected()
-        logger.info(
-            "[bluebubbles] webhook listening on http://%s:%s%s",
-            self.webhook_host,
-            self.webhook_port,
-            self.webhook_path,
-        )
-
-        # Register webhook with BlueBubbles server
-        # This is required for the server to know where to send events
-        await self._register_webhook()
-
-        return True
-
-    async def disconnect(self) -> None:
-        # Unregister webhook before cleaning up
-        await self._unregister_webhook()
-
-        if self.client:
-            await self.client.aclose()
-            self.client = None
-        if self._runner:
-            await self._runner.cleanup()
-            self._runner = None
-        self._mark_disconnected()
-
-    @property
-    def _webhook_url(self) -> str:
-        """Compute the external webhook URL for BlueBubbles registration."""
-        host = self.webhook_host
-        if host in ("0.0.0.0", "127.0.0.1", "localhost", "::"):
-            host = "localhost"
-        return f"http://{host}:{self.webhook_port}{self.webhook_path}"
-
-    @property
-    def _webhook_register_url(self) -> str:
-        """Webhook URL registered with BlueBubbles, including the password as
-        a query param so inbound webhook POSTs carry credentials.
-
-        BlueBubbles posts events to the exact URL registered via
-        ``/api/v1/webhook``. Its webhook registration API does not support
-        custom headers, so embedding the password in the URL is the only
-        way to authenticate inbound webhooks without disabling auth.
-        """
-        base = self._webhook_url
-        if self.password:
-            return f"{base}?password={quote(self.password, safe='')}"
-        return base
-
-    async def _find_registered_webhooks(self, url: str) -> list:
-        """Return list of BB webhook entries matching *url*."""
-        try:
-            res = await self._api_get("/api/v1/webhook")
-            data = res.get("data")
-            if isinstance(data, list):
-                return [wh for wh in data if wh.get("url") == url]
-        except Exception:
-            pass
-        return []
-
-    async def _register_webhook(self) -> bool:
-        """Register this webhook URL with the BlueBubbles server.
-
-        BlueBubbles requires webhooks to be registered via API before
-        it will send events.  Checks for an existing registration first
-        to avoid duplicates (e.g. after a crash without clean shutdown).
-        """
-        if not self.client:
-            return False
-
-        webhook_url = self._webhook_register_url
-
-        # Crash resilience — reuse an existing registration if present
-        existing = await self._find_registered_webhooks(webhook_url)
-        if existing:
-            logger.info(
-                "[bluebubbles] webhook already registered: %s", webhook_url
-            )
-            return True
-
-        payload = {
-            "url": webhook_url,
-            "events": ["new-message", "updated-message"],
-        }
-
-        try:
-            res = await self._api_post("/api/v1/webhook", payload)
-            status = res.get("status", 0)
-            if 200 <= status < 300:
-                logger.info(
-                    "[bluebubbles] webhook registered with server: %s",
-                    webhook_url,
-                )
-                return True
-            else:
-                logger.warning(
-                    "[bluebubbles] webhook registration returned status %s: %s",
-                    status,
-                    res.get("message"),
-                )
-                return False
-        except Exception as exc:
-            logger.warning(
-                "[bluebubbles] failed to register webhook with server: %s",
-                exc,
-            )
-            return False
-
-    async def _unregister_webhook(self) -> bool:
-        """Unregister this webhook URL from the BlueBubbles server.
-
-        Removes *all* matching registrations to clean up any duplicates
-        left by prior crashes.
-        """
-        if not self.client:
-            return False
-
-        webhook_url = self._webhook_register_url
-        removed = False
-
-        try:
-            for wh in await self._find_registered_webhooks(webhook_url):
-                wh_id = wh.get("id")
-                if wh_id:
-                    res = await self.client.delete(
-                        self._api_url(f"/api/v1/webhook/{wh_id}")
-                    )
-                    res.raise_for_status()
-                    removed = True
-            if removed:
-                logger.info(
-                    "[bluebubbles] webhook unregistered: %s", webhook_url
-                )
-        except Exception as exc:
-            logger.debug(
-                "[bluebubbles] failed to unregister webhook (non-critical): %s",
-                exc,
-            )
-        return removed
-
-    # ------------------------------------------------------------------
-    # Chat GUID resolution
-    # ------------------------------------------------------------------
-
-    async def _resolve_chat_guid(self, target: str) -> Optional[str]:
-        """Resolve an email/phone to a BlueBubbles chat GUID.
-
-        If *target* already contains a semicolon (raw GUID format like
-        ``iMessage;-;user@example.com``), it is returned as-is.  Otherwise
-        the adapter queries the BlueBubbles chat list and matches on
-        ``chatIdentifier`` or participant address.
-        """
-        target = (target or "").strip()
-        if not target:
-            return None
-        # Already a raw GUID
-        if ";" in target:
-            return target
-        if target in self._guid_cache:
-            return self._guid_cache[target]
-        try:
-            payload = await self._api_post(
-                "/api/v1/chat/query",
-                {"limit": 100, "offset": 0, "with": ["participants"]},
-            )
-            for chat in payload.get("data", []) or []:
-                guid = chat.get("guid") or chat.get("chatGuid")
-                identifier = chat.get("chatIdentifier") or chat.get("identifier")
-                if identifier == target:
-                    if guid:
-                        self._guid_cache[target] = guid
-                    return guid
-                for part in chat.get("participants", []) or []:
-                    if (part.get("address") or "").strip() == target and guid:
-                        self._guid_cache[target] = guid
-                        return guid
-        except Exception:
-            pass
-        return None
-
-    async def _create_chat_for_handle(
-        self, address: str, message: str
-    ) -> SendResult:
-        """Create a new chat by sending the first message to *address*."""
-        payload = {
-            "addresses": [address],
-            "message": message,
-            "tempGuid": f"temp-{datetime.utcnow().timestamp()}",
-        }
-        try:
-            res = await self._api_post("/api/v1/chat/new", payload)
-            data = res.get("data") or {}
-            msg_id = data.get("guid") or data.get("messageGuid") or "ok"
-            return SendResult(success=True, message_id=str(msg_id), raw_response=res)
-        except Exception as exc:
-            return SendResult(success=False, error=str(exc))
-
-    # ------------------------------------------------------------------
-    # Text sending
-    # ------------------------------------------------------------------
-
-    async def send(
-        self,
-        chat_id: str,
-        content: str,
-        reply_to: Optional[str] = None,
-        metadata: Optional[Dict[str, Any]] = None,
-    ) -> SendResult:
-        text = strip_markdown(content or "")
-        if not text:
-            return SendResult(success=False, error="BlueBubbles send requires text")
-        chunks = self.truncate_message(text, max_length=self.MAX_MESSAGE_LENGTH)
-        last = SendResult(success=True)
-        for chunk in chunks:
-            guid = await self._resolve_chat_guid(chat_id)
-            if not guid:
-                # If the target looks like an address, try creating a new chat
-                if self._private_api_enabled and (
-                    "@" in chat_id or re.match(r"^\+\d+", chat_id)
-                ):
-                    return await self._create_chat_for_handle(chat_id, chunk)
-                return SendResult(
-                    success=False,
-                    error=f"BlueBubbles chat not found for target: {chat_id}",
-                )
-            payload: Dict[str, Any] = {
-                "chatGuid": guid,
-                "tempGuid": f"temp-{datetime.utcnow().timestamp()}",
-                "message": chunk,
-            }
-            if reply_to and self._private_api_enabled and self._helper_connected:
-                payload["method"] = "private-api"
-                payload["selectedMessageGuid"] = reply_to
-                payload["partIndex"] = 0
-            try:
-                res = await self._api_post("/api/v1/message/text", payload)
-                data = res.get("data") or {}
-                msg_id = data.get("guid") or data.get("messageGuid") or "ok"
-                last = SendResult(
-                    success=True, message_id=str(msg_id), raw_response=res
-                )
-            except Exception as exc:
-                return SendResult(success=False, error=str(exc))
-        return last
-
-    # ------------------------------------------------------------------
-    # Media sending (outbound)
-    # ------------------------------------------------------------------
-
-    async def _send_attachment(
-        self,
-        chat_id: str,
-        file_path: str,
-        filename: Optional[str] = None,
-        caption: Optional[str] = None,
-        is_audio_message: bool = False,
-    ) -> SendResult:
-        """Send a file attachment via BlueBubbles multipart upload."""
-        if not self.client:
-            return SendResult(success=False, error="Not connected")
-        if not os.path.isfile(file_path):
-            return SendResult(success=False, error=f"File not found: {file_path}")
-
-        guid = await self._resolve_chat_guid(chat_id)
-        if not guid:
-            return SendResult(success=False, error=f"Chat not found: {chat_id}")
-
-        fname = filename or os.path.basename(file_path)
-        try:
-            with open(file_path, "rb") as f:
-                files = {"attachment": (fname, f, "application/octet-stream")}
-                data: Dict[str, str] = {
-                    "chatGuid": guid,
-                    "name": fname,
-                    "tempGuid": uuid.uuid4().hex,
-                }
-                if is_audio_message:
-                    data["isAudioMessage"] = "true"
-                res = await self.client.post(
-                    self._api_url("/api/v1/message/attachment"),
-                    files=files,
-                    data=data,
-                    timeout=120,
-                )
-                res.raise_for_status()
-                result = res.json()
-
-            if caption:
-                await self.send(chat_id, caption)
-
-            if result.get("status") == 200:
-                rdata = result.get("data") or {}
-                msg_id = rdata.get("guid") if isinstance(rdata, dict) else None
-                return SendResult(
-                    success=True, message_id=msg_id, raw_response=result
-                )
-            return SendResult(
-                success=False,
-                error=result.get("message", "Attachment upload failed"),
-            )
-        except Exception as e:
-            return SendResult(success=False, error=str(e))
-
-    async def send_image(
-        self,
-        chat_id: str,
-        image_url: str,
-        caption: Optional[str] = None,
-        reply_to: Optional[str] = None,
-        metadata: Optional[Dict[str, Any]] = None,
-    ) -> SendResult:
-        try:
-            from gateway.platforms.base import cache_image_from_url
-
-            local_path = await cache_image_from_url(image_url)
-            return await self._send_attachment(chat_id, local_path, caption=caption)
-        except Exception:
-            return await super().send_image(chat_id, image_url, caption, reply_to)
-
-    async def send_image_file(
-        self,
-        chat_id: str,
-        image_path: str,
-        caption: Optional[str] = None,
-        reply_to: Optional[str] = None,
-        **kwargs,
-    ) -> SendResult:
-        return await self._send_attachment(chat_id, image_path, caption=caption)
-
-    async def send_voice(
-        self,
-        chat_id: str,
-        audio_path: str,
-        caption: Optional[str] = None,
-        reply_to: Optional[str] = None,
-        **kwargs,
-    ) -> SendResult:
-        return await self._send_attachment(
-            chat_id, audio_path, caption=caption, is_audio_message=True
-        )
-
-    async def send_video(
-        self,
-        chat_id: str,
-        video_path: str,
-        caption: Optional[str] = None,
-        reply_to: Optional[str] = None,
-        **kwargs,
-    ) -> SendResult:
-        return await self._send_attachment(chat_id, video_path, caption=caption)
-
-    async def send_document(
-        self,
-        chat_id: str,
-        file_path: str,
-        caption: Optional[str] = None,
-        file_name: Optional[str] = None,
-        reply_to: Optional[str] = None,
-        **kwargs,
-    ) -> SendResult:
-        return await self._send_attachment(
-            chat_id, file_path, filename=file_name, caption=caption
-        )
-
-    async def send_animation(
-        self,
-        chat_id: str,
-        animation_url: str,
-        caption: Optional[str] = None,
-        reply_to: Optional[str] = None,
-        metadata: Optional[Dict[str, Any]] = None,
-    ) -> SendResult:
-        return await self.send_image(
-            chat_id, animation_url, caption, reply_to, metadata
-        )
-
-    # ------------------------------------------------------------------
-    # Typing indicators
-    # ------------------------------------------------------------------
-
-    async def send_typing(self, chat_id: str, metadata=None) -> None:
-        if not self._private_api_enabled or not self._helper_connected or not self.client:
-            return
-        try:
-            guid = await self._resolve_chat_guid(chat_id)
-            if guid:
-                encoded = quote(guid, safe="")
-                await self.client.post(
-                    self._api_url(f"/api/v1/chat/{encoded}/typing"), timeout=5
-                )
-        except Exception:
-            pass
-
-    async def stop_typing(self, chat_id: str) -> None:
-        if not self._private_api_enabled or not self._helper_connected or not self.client:
-            return
-        try:
-            guid = await self._resolve_chat_guid(chat_id)
-            if guid:
-                encoded = quote(guid, safe="")
-                await self.client.delete(
-                    self._api_url(f"/api/v1/chat/{encoded}/typing"), timeout=5
-                )
-        except Exception:
-            pass
-
-    # ------------------------------------------------------------------
-    # Read receipts
-    # ------------------------------------------------------------------
-
-    async def mark_read(self, chat_id: str) -> bool:
-        if not self._private_api_enabled or not self._helper_connected or not self.client:
-            return False
-        try:
-            guid = await self._resolve_chat_guid(chat_id)
-            if guid:
-                encoded = quote(guid, safe="")
-                await self.client.post(
-                    self._api_url(f"/api/v1/chat/{encoded}/read"), timeout=5
-                )
-                return True
-        except Exception:
-            pass
-        return False
-
-    # ------------------------------------------------------------------
-    # Tapback reactions
-    # ------------------------------------------------------------------
-
-    # ------------------------------------------------------------------
-    # Chat info
-    # ------------------------------------------------------------------
-
-    async def get_chat_info(self, chat_id: str) -> Dict[str, Any]:
-        is_group = ";+;" in (chat_id or "")
-        info: Dict[str, Any] = {
-            "name": chat_id,
-            "type": "group" if is_group else "dm",
-        }
-        try:
-            guid = await self._resolve_chat_guid(chat_id)
-            if guid:
-                encoded = quote(guid, safe="")
-                res = await self._api_get(
-                    f"/api/v1/chat/{encoded}?with=participants"
-                )
-                data = (res or {}).get("data", {})
-                display_name = (
-                    data.get("displayName")
-                    or data.get("chatIdentifier")
-                    or chat_id
-                )
-                participants = []
-                for p in data.get("participants", []) or []:
-                    addr = (p.get("address") or "").strip()
-                    if addr:
-                        participants.append(addr)
-                info["name"] = display_name
-                if participants:
-                    info["participants"] = participants
-        except Exception:
-            pass
-        return info
-
-    def format_message(self, content: str) -> str:
-        return strip_markdown(content)
-
-    # ------------------------------------------------------------------
-    # Inbound attachment downloading (from #4588)
-    # ------------------------------------------------------------------
-
-    async def _download_attachment(
-        self, att_guid: str, att_meta: Dict[str, Any]
-    ) -> Optional[str]:
-        """Download an attachment from BlueBubbles and cache it locally.
-
-        Returns the local file path on success, None on failure.
-        """
-        if not self.client:
-            return None
-        try:
-            encoded = quote(att_guid, safe="")
-            resp = await self.client.get(
-                self._api_url(f"/api/v1/attachment/{encoded}/download"),
-                timeout=60,
-                follow_redirects=True,
-            )
-            resp.raise_for_status()
-            data = resp.content
-
-            mime = (att_meta.get("mimeType") or "").lower()
-            transfer_name = att_meta.get("transferName", "")
-
-            if mime.startswith("image/"):
-                ext_map = {
-                    "image/jpeg": ".jpg",
-                    "image/png": ".png",
-                    "image/gif": ".gif",
-                    "image/webp": ".webp",
-                    "image/heic": ".jpg",
-                    "image/heif": ".jpg",
-                    "image/tiff": ".jpg",
-                }
-                ext = ext_map.get(mime, ".jpg")
-                return cache_image_from_bytes(data, ext)
-
-            if mime.startswith("audio/"):
-                ext_map = {
-                    "audio/mp3": ".mp3",
-                    "audio/mpeg": ".mp3",
-                    "audio/ogg": ".ogg",
-                    "audio/wav": ".wav",
-                    "audio/x-caf": ".mp3",
-                    "audio/mp4": ".m4a",
-                    "audio/aac": ".m4a",
-                }
-                ext = ext_map.get(mime, ".mp3")
-                return cache_audio_from_bytes(data, ext)
-
-            # Videos, documents, and everything else
-            filename = transfer_name or f"file_{uuid.uuid4().hex[:8]}"
-            return cache_document_from_bytes(data, filename)
-
-        except Exception as exc:
-            logger.warning(
-                "[bluebubbles] failed to download attachment %s: %s",
-                _redact(att_guid),
-                exc,
-            )
-            return None
-
-    # ------------------------------------------------------------------
-    # Webhook handling
-    # ------------------------------------------------------------------
-
-    def _extract_payload_record(
-        self, payload: Dict[str, Any]
-    ) -> Optional[Dict[str, Any]]:
-        data = payload.get("data")
-        if isinstance(data, dict):
-            return data
-        if isinstance(data, list):
-            for item in data:
-                if isinstance(item, dict):
-                    return item
-        if isinstance(payload.get("message"), dict):
-            return payload.get("message")
-        return payload if isinstance(payload, dict) else None
-
-    @staticmethod
-    def _value(*candidates: Any) -> Optional[str]:
-        for candidate in candidates:
-            if isinstance(candidate, str) and candidate.strip():
-                return candidate.strip()
-        return None
-
-    async def _handle_webhook(self, request):
-        from aiohttp import web
-
-        token = (
-            request.query.get("password")
-            or request.query.get("guid")
-            or request.headers.get("x-password")
-            or request.headers.get("x-guid")
-            or request.headers.get("x-bluebubbles-guid")
-        )
-        if token != self.password:
-            return web.json_response({"error": "unauthorized"}, status=401)
-        try:
-            raw = await request.read()
-            body = raw.decode("utf-8", errors="replace")
-            try:
-                payload = json.loads(body)
-            except Exception:
-                from urllib.parse import parse_qs
-
-                form = parse_qs(body)
-                payload_str = (
-                    form.get("payload")
-                    or form.get("data")
-                    or form.get("message")
-                    or [""]
-                )[0]
-                payload = json.loads(payload_str) if payload_str else {}
-        except Exception as exc:
-            logger.error("[bluebubbles] webhook parse error: %s", exc)
-            return web.json_response({"error": "invalid payload"}, status=400)
-
-        event_type = self._value(payload.get("type"), payload.get("event")) or ""
-        # Only process message events; silently acknowledge everything else
-        if event_type and event_type not in _MESSAGE_EVENTS:
-            return web.Response(text="ok")
-
-        record = self._extract_payload_record(payload) or {}
-        is_from_me = bool(
-            record.get("isFromMe")
-            or record.get("fromMe")
-            or record.get("is_from_me")
-        )
-        if is_from_me:
-            return web.Response(text="ok")
-
-        # Skip tapback reactions delivered as messages
-        assoc_type = record.get("associatedMessageType")
-        if isinstance(assoc_type, int) and assoc_type in {
-            **_TAPBACK_ADDED,
-            **_TAPBACK_REMOVED,
-        }:
-            return web.Response(text="ok")
-
-        text = (
-            self._value(
-                record.get("text"), record.get("message"), record.get("body")
-            )
-            or ""
-        )
-
-        # --- Inbound attachment handling ---
-        attachments = record.get("attachments") or []
-        media_urls: List[str] = []
-        media_types: List[str] = []
-        msg_type = MessageType.TEXT
-
-        for att in attachments:
-            att_guid = att.get("guid", "")
-            if not att_guid:
-                continue
-            cached = await self._download_attachment(att_guid, att)
-            if cached:
-                mime = (att.get("mimeType") or "").lower()
-                media_urls.append(cached)
-                media_types.append(mime)
-                if mime.startswith("image/"):
-                    msg_type = MessageType.PHOTO
-                elif mime.startswith("audio/") or (att.get("uti") or "").endswith(
-                    "caf"
-                ):
-                    msg_type = MessageType.VOICE
-                elif mime.startswith("video/"):
-                    msg_type = MessageType.VIDEO
-                else:
-                    msg_type = MessageType.DOCUMENT
-
-        # With multiple attachments, prefer PHOTO if any images present
-        if len(media_urls) > 1:
-            mime_prefixes = {(m or "").split("/")[0] for m in media_types}
-            if "image" in mime_prefixes:
-                msg_type = MessageType.PHOTO
-
-        if not text and media_urls:
-            text = "(attachment)"
-        # --- End attachment handling ---
-
-        chat_guid = self._value(
-            record.get("chatGuid"),
-            payload.get("chatGuid"),
-            record.get("chat_guid"),
-            payload.get("chat_guid"),
-            payload.get("guid"),
-        )
-        # Fallback: BlueBubbles v1.9+ webhook payloads omit top-level chatGuid;
-        # the chat GUID is nested under data.chats[0].guid instead.
-        if not chat_guid:
-            _chats = record.get("chats") or []
-            if _chats and isinstance(_chats[0], dict):
-                chat_guid = _chats[0].get("guid") or _chats[0].get("chatGuid")
-        chat_identifier = self._value(
-            record.get("chatIdentifier"),
-            record.get("identifier"),
-            payload.get("chatIdentifier"),
-            payload.get("identifier"),
-        )
-        sender = (
-            self._value(
-                record.get("handle", {}).get("address")
-                if isinstance(record.get("handle"), dict)
-                else None,
-                record.get("sender"),
-                record.get("from"),
-                record.get("address"),
-            )
-            or chat_identifier
-            or chat_guid
-        )
-        if not (chat_guid or chat_identifier) and sender:
-            chat_identifier = sender
-        if not sender or not (chat_guid or chat_identifier) or not text:
-            return web.json_response({"error": "missing message fields"}, status=400)
-
-        session_chat_id = chat_guid or chat_identifier
-        is_group = bool(record.get("isGroup")) or (";+;" in (chat_guid or ""))
-        source = self.build_source(
-            chat_id=session_chat_id,
-            chat_name=chat_identifier or sender,
-            chat_type="group" if is_group else "dm",
-            user_id=sender,
-            user_name=sender,
-            chat_id_alt=chat_identifier,
-        )
-        event = MessageEvent(
-            text=text,
-            message_type=msg_type,
-            source=source,
-            raw_message=payload,
-            message_id=self._value(
-                record.get("guid"),
-                record.get("messageGuid"),
-                record.get("id"),
-            ),
-            reply_to_message_id=self._value(
-                record.get("threadOriginatorGuid"),
-                record.get("associatedMessageGuid"),
-            ),
-            media_urls=media_urls,
-            media_types=media_types,
-        )
-        task = asyncio.create_task(self.handle_message(event))
-        self._background_tasks.add(task)
-        task.add_done_callback(self._background_tasks.discard)
-
-        # Fire-and-forget read receipt
-        if self.send_read_receipts and session_chat_id:
-            asyncio.create_task(self.mark_read(session_chat_id))
-
-        return web.Response(text="ok")
-
@@ -195,11 +195,7 @@ def _extract_attachments(

        ext = Path(filename).suffix.lower()
        if ext in _IMAGE_EXTS:
-            try:
-                cached_path = cache_image_from_bytes(payload, ext)
-            except ValueError:
-                logger.debug("Skipping non-image attachment %s (invalid magic bytes)", filename)
-                continue
+            cached_path = cache_image_from_bytes(payload, ext)
            attachments.append({
                "path": cached_path,
                "filename": filename,
@@ -34,9 +34,6 @@ from datetime import datetime
 from pathlib import Path
 from types import SimpleNamespace
 from typing import Any, Dict, List, Optional
-from urllib.error import HTTPError, URLError
-from urllib.parse import urlencode
-from urllib.request import Request, urlopen

 # aiohttp/websockets are independent optional deps — import outside lark_oapi
 # so they remain available for tests and webhook mode even if lark_oapi is missing.
@@ -72,10 +69,7 @@ try:
        UpdateMessageRequestBody,
    )
    from lark_oapi.core.const import FEISHU_DOMAIN, LARK_DOMAIN
-    from lark_oapi.event.callback.model.p2_card_action_trigger import (
-        CallBackCard,
-        P2CardActionTriggerResponse,
-    )
+    from lark_oapi.event.callback.model.p2_card_action_trigger import P2CardActionTriggerResponse
    from lark_oapi.event.dispatcher_handler import EventDispatcherHandler
    from lark_oapi.ws import Client as FeishuWSClient

@@ -83,7 +77,6 @@ try:
 except ImportError:
    FEISHU_AVAILABLE = False
    lark = None  # type: ignore[assignment]
-    CallBackCard = None  # type: ignore[assignment]
    P2CardActionTriggerResponse = None  # type: ignore[assignment]
    EventDispatcherHandler = None  # type: ignore[assignment]
    FeishuWSClient = None  # type: ignore[assignment]
@@ -173,35 +166,9 @@ _FEISHU_WEBHOOK_BODY_TIMEOUT_SECONDS = 30          # max seconds to read request
 _FEISHU_WEBHOOK_ANOMALY_THRESHOLD = 25             # consecutive error responses before WARNING log
 _FEISHU_WEBHOOK_ANOMALY_TTL_SECONDS = 6 * 60 * 60  # anomaly tracker TTL (6 hours) — matches openclaw
 _FEISHU_CARD_ACTION_DEDUP_TTL_SECONDS = 15 * 60    # card action token dedup window (15 min)
-
-_APPROVAL_CHOICE_MAP: Dict[str, str] = {
-    "approve_once": "once",
-    "approve_session": "session",
-    "approve_always": "always",
-    "deny": "deny",
-}
-_APPROVAL_LABEL_MAP: Dict[str, str] = {
-    "once": "Approved once",
-    "session": "Approved for session",
-    "always": "Approved permanently",
-    "deny": "Denied",
-}
 _FEISHU_BOT_MSG_TRACK_SIZE = 512                   # LRU size for tracking sent message IDs
 _FEISHU_REPLY_FALLBACK_CODES = frozenset({230011, 231003})  # reply target withdrawn/missing → create fallback
 _FEISHU_ACK_EMOJI = "OK"
-
-# QR onboarding constants
-_ONBOARD_ACCOUNTS_URLS = {
-    "feishu": "https://accounts.feishu.cn",
-    "lark": "https://accounts.larksuite.com",
-}
-_ONBOARD_OPEN_URLS = {
-    "feishu": "https://open.feishu.cn",
-    "lark": "https://open.larksuite.com",
-}
-_REGISTRATION_PATH = "/oauth/v1/app/registration"
-_ONBOARD_REQUEST_TIMEOUT_S = 10
-
 # ---------------------------------------------------------------------------
 # Fallback display strings
 # ---------------------------------------------------------------------------
@@ -297,7 +264,6 @@ class FeishuAdapterSettings:
    bot_name: str
    dedup_cache_size: int
    text_batch_delay_seconds: float
-    text_batch_split_delay_seconds: float
    text_batch_max_messages: int
    text_batch_max_chars: int
    media_batch_delay_seconds: float
@@ -393,21 +359,19 @@ def _render_code_block_element(element: Dict[str, Any]) -> str:


 def _strip_markdown_to_plain_text(text: str) -> str:
-    """Strip markdown formatting to plain text for Feishu text fallbacks.
-
-    Delegates common markdown stripping to the shared helper and adds
-    Feishu-specific patterns (blockquotes, strikethrough, underline tags,
-    horizontal rules, \\r\\n normalisation).
-    """
-    from gateway.platforms.helpers import strip_markdown
    plain = text.replace("\r\n", "\n")
    plain = _MARKDOWN_LINK_RE.sub(lambda m: f"{m.group(1)} ({m.group(2).strip()})", plain)
+    plain = re.sub(r"^#{1,6}\s+", "", plain, flags=re.MULTILINE)
    plain = re.sub(r"^>\s?", "", plain, flags=re.MULTILINE)
    plain = re.sub(r"^\s*---+\s*$", "---", plain, flags=re.MULTILINE)
+    plain = re.sub(r"```(?:[^\n]*\n)?([\s\S]*?)```", lambda m: m.group(1).strip("\n"), plain)
+    plain = re.sub(r"`([^`\n]+)`", r"\1", plain)
+    plain = re.sub(r"\*\*([^*\n]+)\*\*", r"\1", plain)
+    plain = re.sub(r"\*([^*\n]+)\*", r"\1", plain)
    plain = re.sub(r"~~([^~\n]+)~~", r"\1", plain)
    plain = re.sub(r"<u>([\s\S]*?)</u>", r"\1", plain)
-    plain = strip_markdown(plain)
-    return plain
+    plain = re.sub(r"\n{3,}", "\n\n", plain)
+    return plain.strip()


 def _coerce_int(value: Any, default: Optional[int] = None, min_value: int = 0) -> Optional[int]:
@@ -447,6 +411,14 @@ def _build_markdown_post_payload(content: str) -> str:
    )


+def parse_feishu_post_content(raw_content: str) -> FeishuPostParseResult:
+    try:
+        parsed = json.loads(raw_content) if raw_content else {}
+    except json.JSONDecodeError:
+        return FeishuPostParseResult(text_content=FALLBACK_POST_TEXT)
+    return parse_feishu_post_payload(parsed)
+
+
 def parse_feishu_post_payload(payload: Any) -> FeishuPostParseResult:
    resolved = _resolve_post_payload(payload)
    if not resolved:
@@ -1000,8 +972,7 @@ def _run_official_feishu_ws_client(ws_client: Any, adapter: Any) -> None:
        return await original_connect(*args, **kwargs)

    def _configure_with_overrides(conf: Any) -> Any:
-        if original_configure is None:
-            raise RuntimeError("Feishu _configure_with_overrides called but original_configure is None")
+        assert original_configure is not None
        result = original_configure(conf)
        _apply_runtime_ws_overrides()
        return result
@@ -1043,10 +1014,6 @@ class FeishuAdapter(BasePlatformAdapter):
    """Feishu/Lark bot adapter."""

    MAX_MESSAGE_LENGTH = 8000
-    # Threshold for detecting Feishu client-side message splits.
-    # When a chunk is near the ~4096-char practical limit, a continuation
-    # is almost certain.
-    _SPLIT_THRESHOLD = 4000

    # =========================================================================
    # Lifecycle — init / settings / connect / disconnect
@@ -1073,13 +1040,6 @@ class FeishuAdapter(BasePlatformAdapter):
        self._webhook_rate_counts: Dict[str, tuple[int, float]] = {}  # rate_key → (count, window_start)
        self._webhook_anomaly_counts: Dict[str, tuple[int, str, float]] = {}  # ip → (count, last_status, first_seen)
        self._card_action_tokens: Dict[str, float] = {}  # token → first_seen_time
-        # Inbound events that arrived before the adapter loop was ready
-        # (e.g. during startup/restart or network-flap reconnect). A single
-        # drainer thread replays them as soon as the loop becomes available.
-        self._pending_inbound_events: List[Any] = []
-        self._pending_inbound_lock = threading.Lock()
-        self._pending_drain_scheduled = False
-        self._pending_inbound_max_depth = 1000  # cap queue; drop oldest beyond
        self._chat_locks: Dict[str, asyncio.Lock] = {}  # chat_id → lock (per-chat serial processing)
        self._sent_message_ids_to_chat: Dict[str, str] = {}  # message_id → chat_id (for reaction routing)
        self._sent_message_id_order: List[str] = []  # LRU order for _sent_message_ids_to_chat
@@ -1145,9 +1105,6 @@ class FeishuAdapter(BasePlatformAdapter):
            text_batch_delay_seconds=float(
                os.getenv("HERMES_FEISHU_TEXT_BATCH_DELAY_SECONDS", str(_DEFAULT_TEXT_BATCH_DELAY_SECONDS))
            ),
-            text_batch_split_delay_seconds=float(
-                os.getenv("HERMES_FEISHU_TEXT_BATCH_SPLIT_DELAY_SECONDS", "2.0")
-            ),
            text_batch_max_messages=max(
                1,
                int(os.getenv("HERMES_FEISHU_TEXT_BATCH_MAX_MESSAGES", str(_DEFAULT_TEXT_BATCH_MAX_MESSAGES))),
@@ -1195,7 +1152,6 @@ class FeishuAdapter(BasePlatformAdapter):
        self._bot_name = settings.bot_name
        self._dedup_cache_size = settings.dedup_cache_size
        self._text_batch_delay_seconds = settings.text_batch_delay_seconds
-        self._text_batch_split_delay_seconds = settings.text_batch_split_delay_seconds
        self._text_batch_max_messages = settings.text_batch_max_messages
        self._text_batch_max_chars = settings.text_batch_max_chars
        self._media_batch_delay_seconds = settings.media_batch_delay_seconds
@@ -1224,14 +1180,6 @@ class FeishuAdapter(BasePlatformAdapter):
                lambda data: self._on_reaction_event("im.message.reaction.deleted_v1", data)
            )
            .register_p2_card_action_trigger(self._on_card_action_trigger)
-            .register_p2_im_chat_member_bot_added_v1(self._on_bot_added_to_chat)
-            .register_p2_im_chat_member_bot_deleted_v1(self._on_bot_removed_from_chat)
-            .register_p2_im_chat_access_event_bot_p2p_chat_entered_v1(self._on_p2p_chat_entered)
-            .register_p2_im_message_recalled_v1(self._on_message_recalled)
-            .register_p2_customized_event(
-                "drive.notice.comment_add_v1",
-                self._on_drive_comment_event,
-            )
            .build()
        )

@@ -1520,12 +1468,14 @@ class FeishuAdapter(BasePlatformAdapter):
            logger.warning("[Feishu] send_exec_approval failed: %s", exc)
            return SendResult(success=False, error=str(exc))

-    @staticmethod
-    def _build_resolved_approval_card(*, choice: str, user_name: str) -> Dict[str, Any]:
-        """Build raw card JSON for a resolved approval action."""
+    async def _update_approval_card(
+        self, message_id: str, label: str, user_name: str, choice: str,
+    ) -> None:
+        """Replace the approval card with a resolved status card."""
+        if not self._client or not message_id:
+            return
        icon = "❌" if choice == "deny" else "✅"
-        label = _APPROVAL_LABEL_MAP.get(choice, "Resolved")
-        return {
+        card = {
            "config": {"wide_screen_mode": True},
            "header": {
                "title": {"content": f"{icon} {label}", "tag": "plain_text"},
@@ -1538,6 +1488,13 @@ class FeishuAdapter(BasePlatformAdapter):
                },
            ],
        }
+        try:
+            payload = json.dumps(card, ensure_ascii=False)
+            body = self._build_update_message_body(msg_type="interactive", content=payload)
+            request = self._build_update_message_request(message_id=message_id, request_body=body)
+            await asyncio.to_thread(self._client.im.v1.message.update, request)
+        except Exception as exc:
+            logger.warning("[Feishu] Failed to update approval card %s: %s", message_id, exc)

    async def send_voice(
        self,
@@ -1613,18 +1570,13 @@ class FeishuAdapter(BasePlatformAdapter):
            return SendResult(success=False, error=f"Image file not found: {image_path}")

        try:
-            import io as _io
-            with open(image_path, "rb") as f:
-                image_bytes = f.read()
-            # Wrap in BytesIO so lark SDK's MultipartEncoder can read .name and .tell()
-            image_file = _io.BytesIO(image_bytes)
-            image_file.name = os.path.basename(image_path)
-            body = self._build_image_upload_body(
-                image_type=_FEISHU_IMAGE_UPLOAD_TYPE,
-                image=image_file,
-            )
-            request = self._build_image_upload_request(body)
-            upload_response = await asyncio.to_thread(self._client.im.v1.image.create, request)
+            with open(image_path, "rb") as image_file:
+                body = self._build_image_upload_body(
+                    image_type=_FEISHU_IMAGE_UPLOAD_TYPE,
+                    image=image_file,
+                )
+                request = self._build_image_upload_request(body)
+                upload_response = await asyncio.to_thread(self._client.im.v1.image.create, request)
            image_key = self._extract_response_field(upload_response, "image_key")
            if not image_key:
                return self._response_error_result(
@@ -1770,22 +1722,10 @@ class FeishuAdapter(BasePlatformAdapter):
    # =========================================================================

    def _on_message_event(self, data: Any) -> None:
-        """Normalize Feishu inbound events into MessageEvent.
-
-        Called by the lark_oapi SDK's event dispatcher on a background thread.
-        If the adapter loop is not currently accepting callbacks (brief window
-        during startup/restart or network-flap reconnect), the event is queued
-        for replay instead of dropped.
-        """
+        """Normalize Feishu inbound events into MessageEvent."""
        loop = self._loop
-        if not self._loop_accepts_callbacks(loop):
-            start_drainer = self._enqueue_pending_inbound_event(data)
-            if start_drainer:
-                threading.Thread(
-                    target=self._drain_pending_inbound_events,
-                    name="feishu-pending-inbound-drainer",
-                    daemon=True,
-                ).start()
+        if loop is None or bool(getattr(loop, "is_closed", lambda: False)()):
+            logger.warning("[Feishu] Dropping inbound message before adapter loop is ready")
            return
        future = asyncio.run_coroutine_threadsafe(
            self._handle_message_event_data(data),
@@ -1793,124 +1733,6 @@ class FeishuAdapter(BasePlatformAdapter):
        )
        future.add_done_callback(self._log_background_failure)

-    def _enqueue_pending_inbound_event(self, data: Any) -> bool:
-        """Append an event to the pending-inbound queue.
-
-        Returns True if the caller should spawn a drainer thread (no drainer
-        currently scheduled), False if a drainer is already running and will
-        pick up the new event on its next pass.
-        """
-        with self._pending_inbound_lock:
-            if len(self._pending_inbound_events) >= self._pending_inbound_max_depth:
-                # Queue full — drop the oldest to make room. This happens only
-                # if the loop stays unavailable for an extended period AND the
-                # WS keeps firing callbacks. Still better than silent drops.
-                dropped = self._pending_inbound_events.pop(0)
-                try:
-                    event = getattr(dropped, "event", None)
-                    message = getattr(event, "message", None)
-                    message_id = str(getattr(message, "message_id", "") or "unknown")
-                except Exception:
-                    message_id = "unknown"
-                logger.error(
-                    "[Feishu] Pending-inbound queue full (%d); dropped oldest event %s",
-                    self._pending_inbound_max_depth,
-                    message_id,
-                )
-            self._pending_inbound_events.append(data)
-            depth = len(self._pending_inbound_events)
-            should_start = not self._pending_drain_scheduled
-            if should_start:
-                self._pending_drain_scheduled = True
-        logger.warning(
-            "[Feishu] Queued inbound event for replay (loop not ready, queue depth=%d)",
-            depth,
-        )
-        return should_start
-
-    def _drain_pending_inbound_events(self) -> None:
-        """Replay queued inbound events once the adapter loop is ready.
-
-        Runs in a dedicated daemon thread. Polls ``_running`` and
-        ``_loop_accepts_callbacks`` until events can be dispatched or the
-        adapter shuts down. A single drainer handles the entire queue;
-        concurrent ``_on_message_event`` calls just append.
-        """
-        poll_interval = 0.25
-        max_wait_seconds = 120.0  # safety cap: drop queue after 2 minutes
-        waited = 0.0
-        try:
-            while True:
-                if not getattr(self, "_running", True):
-                    # Adapter shutting down — drop queued events rather than
-                    # holding them against a closed loop.
-                    with self._pending_inbound_lock:
-                        dropped = len(self._pending_inbound_events)
-                        self._pending_inbound_events.clear()
-                    if dropped:
-                        logger.warning(
-                            "[Feishu] Dropped %d queued inbound event(s) during shutdown",
-                            dropped,
-                        )
-                    return
-                loop = self._loop
-                if self._loop_accepts_callbacks(loop):
-                    with self._pending_inbound_lock:
-                        batch = self._pending_inbound_events[:]
-                        self._pending_inbound_events.clear()
-                    if not batch:
-                        # Queue emptied between check and grab; done.
-                        with self._pending_inbound_lock:
-                            if not self._pending_inbound_events:
-                                return
-                        continue
-                    dispatched = 0
-                    requeue: List[Any] = []
-                    for event in batch:
-                        try:
-                            fut = asyncio.run_coroutine_threadsafe(
-                                self._handle_message_event_data(event),
-                                loop,
-                            )
-                            fut.add_done_callback(self._log_background_failure)
-                            dispatched += 1
-                        except RuntimeError:
-                            # Loop closed between check and submit — requeue
-                            # and poll again.
-                            requeue.append(event)
-                    if requeue:
-                        with self._pending_inbound_lock:
-                            self._pending_inbound_events[:0] = requeue
-                    if dispatched:
-                        logger.info(
-                            "[Feishu] Replayed %d queued inbound event(s)",
-                            dispatched,
-                        )
-                    if not requeue:
-                        # Successfully drained; check if more arrived while
-                        # we were dispatching and exit if not.
-                        with self._pending_inbound_lock:
-                            if not self._pending_inbound_events:
-                                return
-                    # More events queued or requeue pending — loop again.
-                    continue
-                if waited >= max_wait_seconds:
-                    with self._pending_inbound_lock:
-                        dropped = len(self._pending_inbound_events)
-                        self._pending_inbound_events.clear()
-                    logger.error(
-                        "[Feishu] Adapter loop unavailable for %.0fs; "
-                        "dropped %d queued inbound event(s)",
-                        max_wait_seconds,
-                        dropped,
-                    )
-                    return
-                time.sleep(poll_interval)
-                waited += poll_interval
-        finally:
-            with self._pending_inbound_lock:
-                self._pending_drain_scheduled = False
-
    async def _handle_message_event_data(self, data: Any) -> None:
        """Shared inbound message handling for websocket and webhook transports."""
        event = getattr(data, "event", None)
@@ -1963,31 +1785,6 @@ class FeishuAdapter(BasePlatformAdapter):
        logger.info("[Feishu] Bot removed from chat: %s", chat_id)
        self._chat_info_cache.pop(chat_id, None)

-    def _on_p2p_chat_entered(self, data: Any) -> None:
-        logger.debug("[Feishu] User entered P2P chat with bot")
-
-    def _on_message_recalled(self, data: Any) -> None:
-        logger.debug("[Feishu] Message recalled by user")
-
-    def _on_drive_comment_event(self, data: Any) -> None:
-        """Handle drive document comment notification (drive.notice.comment_add_v1).
-
-        Delegates to :mod:`gateway.platforms.feishu_comment` for parsing,
-        logging, and reaction.  Scheduling follows the same
-        ``run_coroutine_threadsafe`` pattern used by ``_on_message_event``.
-        """
-        from gateway.platforms.feishu_comment import handle_drive_comment_event
-
-        loop = self._loop
-        if not self._loop_accepts_callbacks(loop):
-            logger.warning("[Feishu] Dropping drive comment event before adapter loop is ready")
-            return
-        future = asyncio.run_coroutine_threadsafe(
-            handle_drive_comment_event(self._client, data, self_open_id=self._bot_open_id),
-            loop,
-        )
-        future.add_done_callback(self._log_background_failure)
-
    def _on_reaction_event(self, event_type: str, data: Any) -> None:
        """Route user reactions on bot messages as synthetic text events."""
        event = getattr(data, "event", None)
@@ -2021,82 +1818,20 @@ class FeishuAdapter(BasePlatformAdapter):
        future.add_done_callback(self._log_background_failure)

    def _on_card_action_trigger(self, data: Any) -> Any:
-        """Handle card-action callback from the Feishu SDK (synchronous).
-
-        For approval actions: parses the event once, returns the resolved card
-        inline (the only reliable way to sync all clients), and schedules a
-        lightweight async method to actually unblock the agent.
-
-        For other card actions: delegates to ``_handle_card_action_event``.
-        """
+        """Schedule Feishu card actions on the adapter loop and acknowledge immediately."""
        loop = self._loop
-        if not self._loop_accepts_callbacks(loop):
+        if loop is None or bool(getattr(loop, "is_closed", lambda: False)()):
            logger.warning("[Feishu] Dropping card action before adapter loop is ready")
-            return P2CardActionTriggerResponse() if P2CardActionTriggerResponse else None
-
-        event = getattr(data, "event", None)
-        action = getattr(event, "action", None)
-        action_value = getattr(action, "value", {}) or {}
-        hermes_action = action_value.get("hermes_action") if isinstance(action_value, dict) else None
-
-        if hermes_action:
-            return self._handle_approval_card_action(event=event, action_value=action_value, loop=loop)
-
-        self._submit_on_loop(loop, self._handle_card_action_event(data))
+        else:
+            future = asyncio.run_coroutine_threadsafe(
+                self._handle_card_action_event(data),
+                loop,
+            )
+            future.add_done_callback(self._log_background_failure)
        if P2CardActionTriggerResponse is None:
            return None
        return P2CardActionTriggerResponse()

-    @staticmethod
-    def _loop_accepts_callbacks(loop: Any) -> bool:
-        """Return True when the adapter loop can accept thread-safe submissions."""
-        return loop is not None and not bool(getattr(loop, "is_closed", lambda: False)())
-
-    def _submit_on_loop(self, loop: Any, coro: Any) -> None:
-        """Schedule background work on the adapter loop with shared failure logging."""
-        future = asyncio.run_coroutine_threadsafe(coro, loop)
-        future.add_done_callback(self._log_background_failure)
-
-    def _handle_approval_card_action(self, *, event: Any, action_value: Dict[str, Any], loop: Any) -> Any:
-        """Schedule approval resolution and build the synchronous callback response."""
-        approval_id = action_value.get("approval_id")
-        if approval_id is None:
-            logger.debug("[Feishu] Card action missing approval_id, ignoring")
-            return P2CardActionTriggerResponse() if P2CardActionTriggerResponse else None
-        choice = _APPROVAL_CHOICE_MAP.get(action_value.get("hermes_action"), "deny")
-
-        operator = getattr(event, "operator", None)
-        open_id = str(getattr(operator, "open_id", "") or "")
-        user_name = self._get_cached_sender_name(open_id) or open_id
-
-        self._submit_on_loop(loop, self._resolve_approval(approval_id, choice, user_name))
-
-        if P2CardActionTriggerResponse is None:
-            return None
-        response = P2CardActionTriggerResponse()
-        if CallBackCard is not None:
-            card = CallBackCard()
-            card.type = "raw"
-            card.data = self._build_resolved_approval_card(choice=choice, user_name=user_name)
-            response.card = card
-        return response
-
-    async def _resolve_approval(self, approval_id: Any, choice: str, user_name: str) -> None:
-        """Pop approval state and unblock the waiting agent thread."""
-        state = self._approval_state.pop(approval_id, None)
-        if not state:
-            logger.debug("[Feishu] Approval %s already resolved or unknown", approval_id)
-            return
-        try:
-            from tools.approval import resolve_gateway_approval
-            count = resolve_gateway_approval(state["session_key"], choice)
-            logger.info(
-                "Feishu button resolved %d approval(s) for session %s (choice=%s, user=%s)",
-                count, state["session_key"], choice, user_name,
-            )
-        except Exception as exc:
-            logger.error("Failed to resolve gateway approval from Feishu button: %s", exc)
-
    async def _handle_reaction_event(self, event_type: str, data: Any) -> None:
        """Fetch the reacted-to message; if it was sent by this bot, emit a synthetic text event."""
        if not self._client:
@@ -2188,6 +1923,51 @@ class FeishuAdapter(BasePlatformAdapter):
        action_tag = str(getattr(action, "tag", "") or "button")
        action_value = getattr(action, "value", {}) or {}

+        # --- Exec approval button intercept ---
+        hermes_action = action_value.get("hermes_action") if isinstance(action_value, dict) else None
+        if hermes_action:
+            approval_id = action_value.get("approval_id")
+            state = self._approval_state.pop(approval_id, None)
+            if not state:
+                logger.debug("[Feishu] Approval %s already resolved or unknown", approval_id)
+                return
+
+            choice_map = {
+                "approve_once": "once",
+                "approve_session": "session",
+                "approve_always": "always",
+                "deny": "deny",
+            }
+            choice = choice_map.get(hermes_action, "deny")
+
+            label_map = {
+                "once": "Approved once",
+                "session": "Approved for session",
+                "always": "Approved permanently",
+                "deny": "Denied",
+            }
+            label = label_map.get(choice, "Resolved")
+
+            # Resolve sender name for the status card
+            sender_id = SimpleNamespace(open_id=open_id, user_id=None, union_id=None)
+            sender_profile = await self._resolve_sender_profile(sender_id)
+            user_name = sender_profile.get("user_name") or open_id
+
+            # Resolve the approval — unblocks the agent thread
+            try:
+                from tools.approval import resolve_gateway_approval
+                count = resolve_gateway_approval(state["session_key"], choice)
+                logger.info(
+                    "Feishu button resolved %d approval(s) for session %s (choice=%s, user=%s)",
+                    count, state["session_key"], choice, user_name,
+                )
+            except Exception as exc:
+                logger.error("Failed to resolve gateway approval from Feishu button: %s", exc)
+
+            # Update the card to show the decision
+            await self._update_approval_card(state.get("message_id", ""), label, user_name, choice)
+            return
+
        synthetic_text = f"/card {action_tag}"
        if action_value:
            try:
@@ -2613,8 +2393,6 @@ class FeishuAdapter(BasePlatformAdapter):
            self._on_reaction_event(event_type, data)
        elif event_type == "card.action.trigger":
            self._on_card_action_trigger(data)
-        elif event_type == "drive.notice.comment_add_v1":
-            self._on_drive_comment_event(data)
        else:
            logger.debug("[Feishu] Ignoring webhook event type: %s", event_type or "unknown")
        return web.json_response({"code": 0, "msg": "ok"})
@@ -2700,10 +2478,8 @@ class FeishuAdapter(BasePlatformAdapter):
    async def _enqueue_text_event(self, event: MessageEvent) -> None:
        """Debounce rapid Feishu text bursts into a single MessageEvent."""
        key = self._text_batch_key(event)
-        chunk_len = len(event.text or "")
        existing = self._pending_text_batches.get(key)
        if existing is None:
-            event._last_chunk_len = chunk_len  # type: ignore[attr-defined]
            self._pending_text_batches[key] = event
            self._pending_text_batch_counts[key] = 1
            self._schedule_text_batch_flush(key)
@@ -2728,7 +2504,6 @@ class FeishuAdapter(BasePlatformAdapter):
            return

        existing.text = next_text
-        existing._last_chunk_len = chunk_len  # type: ignore[attr-defined]
        existing.timestamp = event.timestamp
        if event.message_id:
            existing.message_id = event.message_id
@@ -2755,22 +2530,10 @@ class FeishuAdapter(BasePlatformAdapter):
        task_map[key] = asyncio.create_task(flush_fn(key))

    async def _flush_text_batch(self, key: str) -> None:
-        """Flush a pending text batch after the quiet period.
-
-        Uses a longer delay when the latest chunk is near Feishu's ~4096-char
-        split point, since a continuation chunk is almost certain.
-        """
+        """Flush a pending text batch after the quiet period."""
        current_task = asyncio.current_task()
        try:
-            # Adaptive delay: if the latest chunk is near the split threshold,
-            # a continuation is almost certain — wait longer.
-            pending = self._pending_text_batches.get(key)
-            last_len = getattr(pending, "_last_chunk_len", 0) if pending else 0
-            if last_len >= self._SPLIT_THRESHOLD:
-                delay = self._text_batch_split_delay_seconds
-            else:
-                delay = self._text_batch_delay_seconds
-            await asyncio.sleep(delay)
+            await asyncio.sleep(self._text_batch_delay_seconds)
            await self._flush_text_batch_now(key)
        finally:
            if self._pending_text_batch_tasks.get(key) is current_task:
@@ -2875,6 +2638,12 @@ class FeishuAdapter(BasePlatformAdapter):
            return self._resolve_media_message_type(media_types[0] if media_types else "", default=MessageType.DOCUMENT)
        return MessageType.TEXT

+    def _normalize_inbound_text(self, text: str) -> str:
+        """Strip Feishu mention placeholders from inbound text."""
+        text = _MENTION_RE.sub(" ", text or "")
+        text = _MULTISPACE_RE.sub(" ", text)
+        return text.strip()
+
    async def _maybe_extract_text_document(self, cached_path: str, media_type: str) -> str:
        if not cached_path or not media_type.startswith("text/"):
            return ""
@@ -3092,19 +2861,6 @@ class FeishuAdapter(BasePlatformAdapter):
            "user_id_alt": union_id,
        }

-    def _get_cached_sender_name(self, sender_id: Optional[str]) -> Optional[str]:
-        """Return a cached sender name only while its TTL is still valid."""
-        if not sender_id:
-            return None
-        cached = self._sender_name_cache.get(sender_id)
-        if cached is None:
-            return None
-        name, expire_at = cached
-        if time.time() < expire_at:
-            return name
-        self._sender_name_cache.pop(sender_id, None)
-        return None
-
    async def _resolve_sender_name_from_api(self, sender_id: Optional[str]) -> Optional[str]:
        """Fetch the sender's display name from the Feishu contact API with a 10-minute cache.

@@ -3117,9 +2873,11 @@ class FeishuAdapter(BasePlatformAdapter):
        if not trimmed:
            return None
        now = time.time()
-        cached_name = self._get_cached_sender_name(trimmed)
-        if cached_name is not None:
-            return cached_name
+        cached = self._sender_name_cache.get(trimmed)
+        if cached is not None:
+            name, expire_at = cached
+            if now < expire_at:
+                return name
        try:
            from lark_oapi.api.contact.v3 import GetUserRequest  # lazy import
            if trimmed.startswith("ou_"):
@@ -3829,328 +3587,3 @@ class FeishuAdapter(BasePlatformAdapter):
            return _FEISHU_FILE_UPLOAD_TYPE, "file"

        return _FEISHU_FILE_UPLOAD_TYPE, "file"
-
-
-# =============================================================================
-# QR scan-to-create onboarding
-#
-# Device-code flow: user scans a QR code with Feishu/Lark mobile app and the
-# platform creates a fully configured bot application automatically.
-# Called by `hermes gateway setup` via _setup_feishu() in hermes_cli/gateway.py.
-# =============================================================================
-
-
-def _accounts_base_url(domain: str) -> str:
-    return _ONBOARD_ACCOUNTS_URLS.get(domain, _ONBOARD_ACCOUNTS_URLS["feishu"])
-
-
-def _onboard_open_base_url(domain: str) -> str:
-    return _ONBOARD_OPEN_URLS.get(domain, _ONBOARD_OPEN_URLS["feishu"])
-
-
-def _post_registration(base_url: str, body: Dict[str, str]) -> dict:
-    """POST form-encoded data to the registration endpoint, return parsed JSON.
-
-    The registration endpoint returns JSON even on 4xx (e.g. poll returns
-    authorization_pending as a 400). We always parse the body regardless of
-    HTTP status.
-    """
-    url = f"{base_url}{_REGISTRATION_PATH}"
-    data = urlencode(body).encode("utf-8")
-    req = Request(url, data=data, headers={"Content-Type": "application/x-www-form-urlencoded"})
-    try:
-        with urlopen(req, timeout=_ONBOARD_REQUEST_TIMEOUT_S) as resp:
-            return json.loads(resp.read().decode("utf-8"))
-    except HTTPError as exc:
-        body_bytes = exc.read()
-        if body_bytes:
-            try:
-                return json.loads(body_bytes.decode("utf-8"))
-            except (ValueError, json.JSONDecodeError):
-                raise exc from None
-        raise
-
-
-def _init_registration(domain: str = "feishu") -> None:
-    """Verify the environment supports client_secret auth.
-
-    Raises RuntimeError if not supported.
-    """
-    base_url = _accounts_base_url(domain)
-    res = _post_registration(base_url, {"action": "init"})
-    methods = res.get("supported_auth_methods") or []
-    if "client_secret" not in methods:
-        raise RuntimeError(
-            f"Feishu / Lark registration environment does not support client_secret auth. "
-            f"Supported: {methods}"
-        )
-
-
-def _begin_registration(domain: str = "feishu") -> dict:
-    """Start the device-code flow. Returns device_code, qr_url, user_code, interval, expire_in."""
-    base_url = _accounts_base_url(domain)
-    res = _post_registration(base_url, {
-        "action": "begin",
-        "archetype": "PersonalAgent",
-        "auth_method": "client_secret",
-        "request_user_info": "open_id",
-    })
-    device_code = res.get("device_code")
-    if not device_code:
-        raise RuntimeError("Feishu / Lark registration did not return a device_code")
-    qr_url = res.get("verification_uri_complete", "")
-    if "?" in qr_url:
-        qr_url += "&from=hermes&tp=hermes"
-    else:
-        qr_url += "?from=hermes&tp=hermes"
-    return {
-        "device_code": device_code,
-        "qr_url": qr_url,
-        "user_code": res.get("user_code", ""),
-        "interval": res.get("interval") or 5,
-        "expire_in": res.get("expire_in") or 600,
-    }
-
-
-def _poll_registration(
-    *,
-    device_code: str,
-    interval: int,
-    expire_in: int,
-    domain: str = "feishu",
-) -> Optional[dict]:
-    """Poll until the user scans the QR code, or timeout/denial.
-
-    Returns dict with app_id, app_secret, domain, open_id on success.
-    Returns None on failure.
-    """
-    deadline = time.time() + expire_in
-    current_domain = domain
-    domain_switched = False
-    poll_count = 0
-
-    while time.time() < deadline:
-        base_url = _accounts_base_url(current_domain)
-        try:
-            res = _post_registration(base_url, {
-                "action": "poll",
-                "device_code": device_code,
-                "tp": "ob_app",
-            })
-        except (URLError, OSError, json.JSONDecodeError):
-            time.sleep(interval)
-            continue
-
-        poll_count += 1
-        if poll_count == 1:
-            print("  Fetching configuration results...", end="", flush=True)
-        elif poll_count % 6 == 0:
-            print(".", end="", flush=True)
-
-        # Domain auto-detection
-        user_info = res.get("user_info") or {}
-        tenant_brand = user_info.get("tenant_brand")
-        if tenant_brand == "lark" and not domain_switched:
-            current_domain = "lark"
-            domain_switched = True
-            # Fall through — server may return credentials in this same response.
-
-        # Success
-        if res.get("client_id") and res.get("client_secret"):
-            if poll_count > 0:
-                print()  # newline after "Fetching configuration results..." dots
-            return {
-                "app_id": res["client_id"],
-                "app_secret": res["client_secret"],
-                "domain": current_domain,
-                "open_id": user_info.get("open_id"),
-            }
-
-        # Terminal errors
-        error = res.get("error", "")
-        if error in ("access_denied", "expired_token"):
-            if poll_count > 0:
-                print()
-            logger.warning("[Feishu onboard] Registration %s", error)
-            return None
-
-        # authorization_pending or unknown — keep polling
-        time.sleep(interval)
-
-    if poll_count > 0:
-        print()
-    logger.warning("[Feishu onboard] Poll timed out after %ds", expire_in)
-    return None
-
-
-try:
-    import qrcode as _qrcode_mod
-except (ImportError, TypeError):
-    _qrcode_mod = None  # type: ignore[assignment]
-
-
-def _render_qr(url: str) -> bool:
-    """Try to render a QR code in the terminal. Returns True if successful."""
-    if _qrcode_mod is None:
-        return False
-    try:
-        qr = _qrcode_mod.QRCode()
-        qr.add_data(url)
-        qr.make(fit=True)
-        qr.print_ascii(invert=True)
-        return True
-    except Exception:
-        return False
-
-
-def probe_bot(app_id: str, app_secret: str, domain: str) -> Optional[dict]:
-    """Verify bot connectivity via /open-apis/bot/v3/info.
-
-    Uses lark_oapi SDK when available, falls back to raw HTTP otherwise.
-    Returns {"bot_name": ..., "bot_open_id": ...} on success, None on failure.
-    """
-    if FEISHU_AVAILABLE:
-        return _probe_bot_sdk(app_id, app_secret, domain)
-    return _probe_bot_http(app_id, app_secret, domain)
-
-
-def _build_onboard_client(app_id: str, app_secret: str, domain: str) -> Any:
-    """Build a lark Client for the given credentials and domain."""
-    sdk_domain = LARK_DOMAIN if domain == "lark" else FEISHU_DOMAIN
-    return (
-        lark.Client.builder()
-        .app_id(app_id)
-        .app_secret(app_secret)
-        .domain(sdk_domain)
-        .log_level(lark.LogLevel.WARNING)
-        .build()
-    )
-
-
-def _parse_bot_response(data: dict) -> Optional[dict]:
-    """Extract bot_name and bot_open_id from a /bot/v3/info response."""
-    if data.get("code") != 0:
-        return None
-    bot = data.get("bot") or data.get("data", {}).get("bot") or {}
-    return {
-        "bot_name": bot.get("bot_name"),
-        "bot_open_id": bot.get("open_id"),
-    }
-
-
-def _probe_bot_sdk(app_id: str, app_secret: str, domain: str) -> Optional[dict]:
-    """Probe bot info using lark_oapi SDK."""
-    try:
-        client = _build_onboard_client(app_id, app_secret, domain)
-        resp = client.request(
-            method="GET",
-            url="/open-apis/bot/v3/info",
-            body=None,
-            raw_response=True,
-        )
-        return _parse_bot_response(json.loads(resp.content))
-    except Exception as exc:
-        logger.debug("[Feishu onboard] SDK probe failed: %s", exc)
-        return None
-
-
-def _probe_bot_http(app_id: str, app_secret: str, domain: str) -> Optional[dict]:
-    """Fallback probe using raw HTTP (when lark_oapi is not installed)."""
-    base_url = _onboard_open_base_url(domain)
-    try:
-        token_data = json.dumps({"app_id": app_id, "app_secret": app_secret}).encode("utf-8")
-        token_req = Request(
-            f"{base_url}/open-apis/auth/v3/tenant_access_token/internal",
-            data=token_data,
-            headers={"Content-Type": "application/json"},
-        )
-        with urlopen(token_req, timeout=_ONBOARD_REQUEST_TIMEOUT_S) as resp:
-            token_res = json.loads(resp.read().decode("utf-8"))
-
-        access_token = token_res.get("tenant_access_token")
-        if not access_token:
-            return None
-
-        bot_req = Request(
-            f"{base_url}/open-apis/bot/v3/info",
-            headers={
-                "Authorization": f"Bearer {access_token}",
-                "Content-Type": "application/json",
-            },
-        )
-        with urlopen(bot_req, timeout=_ONBOARD_REQUEST_TIMEOUT_S) as resp:
-            bot_res = json.loads(resp.read().decode("utf-8"))
-
-        return _parse_bot_response(bot_res)
-    except (URLError, OSError, KeyError, json.JSONDecodeError) as exc:
-        logger.debug("[Feishu onboard] HTTP probe failed: %s", exc)
-        return None
-
-
-def qr_register(
-    *,
-    initial_domain: str = "feishu",
-    timeout_seconds: int = 600,
-) -> Optional[dict]:
-    """Run the Feishu / Lark scan-to-create QR registration flow.
-
-    Returns on success::
-
-        {
-            "app_id": str,
-            "app_secret": str,
-            "domain": "feishu" | "lark",
-            "open_id": str | None,
-            "bot_name": str | None,
-            "bot_open_id": str | None,
-        }
-
-    Returns None on expected failures (network, auth denied, timeout).
-    Unexpected errors (bugs, protocol regressions) propagate to the caller.
-    """
-    try:
-        return _qr_register_inner(initial_domain=initial_domain, timeout_seconds=timeout_seconds)
-    except (RuntimeError, URLError, OSError, json.JSONDecodeError) as exc:
-        logger.warning("[Feishu onboard] Registration failed: %s", exc)
-        return None
-
-
-def _qr_register_inner(
-    *,
-    initial_domain: str,
-    timeout_seconds: int,
-) -> Optional[dict]:
-    """Run init → begin → poll → probe. Raises on network/protocol errors."""
-    print("  Connecting to Feishu / Lark...", end="", flush=True)
-    _init_registration(initial_domain)
-    begin = _begin_registration(initial_domain)
-    print(" done.")
-
-    print()
-    qr_url = begin["qr_url"]
-    if _render_qr(qr_url):
-        print(f"\n  Scan the QR code above, or open this URL directly:\n  {qr_url}")
-    else:
-        print(f"  Open this URL in Feishu / Lark on your phone:\n\n  {qr_url}\n")
-        print("  Tip: pip install qrcode  to display a scannable QR code here next time")
-    print()
-
-    result = _poll_registration(
-        device_code=begin["device_code"],
-        interval=begin["interval"],
-        expire_in=min(begin["expire_in"], timeout_seconds),
-        domain=initial_domain,
-    )
-    if not result:
-        return None
-
-    # Probe bot — best-effort, don't fail the registration
-    bot_info = probe_bot(result["app_id"], result["app_secret"], result["domain"])
-    if bot_info:
-        result["bot_name"] = bot_info.get("bot_name")
-        result["bot_open_id"] = bot_info.get("bot_open_id")
-    else:
-        result["bot_name"] = None
-        result["bot_open_id"] = None
-
-    return result
@@ -1,429 +0,0 @@
-"""
-Feishu document comment access-control rules.
-
-3-tier rule resolution: exact doc > wildcard "*" > top-level > code defaults.
-Each field (enabled/policy/allow_from) falls back independently.
-Config: ~/.hermes/feishu_comment_rules.json (mtime-cached, hot-reload).
-Pairing store: ~/.hermes/feishu_comment_pairing.json.
-"""
-
-from __future__ import annotations
-
-import json
-import logging
-import time
-from dataclasses import dataclass, field
-from pathlib import Path
-from typing import Any, Dict, Optional
-
-from hermes_constants import get_hermes_home
-
-logger = logging.getLogger(__name__)
-
-# ---------------------------------------------------------------------------
-# Paths
-# ---------------------------------------------------------------------------
-#
-# Uses the canonical ``get_hermes_home()`` helper (HERMES_HOME-aware and
-# profile-safe). Resolved at import time; this module is lazy-imported by
-# the Feishu comment event handler, which runs long after profile overrides
-# have been applied, so freezing paths here is safe.
-
-RULES_FILE = get_hermes_home() / "feishu_comment_rules.json"
-PAIRING_FILE = get_hermes_home() / "feishu_comment_pairing.json"
-
-# ---------------------------------------------------------------------------
-# Data models
-# ---------------------------------------------------------------------------
-
-_VALID_POLICIES = ("allowlist", "pairing")
-
-
-@dataclass(frozen=True)
-class CommentDocumentRule:
-    """Per-document rule.  ``None`` means 'inherit from lower tier'."""
-    enabled: Optional[bool] = None
-    policy: Optional[str] = None
-    allow_from: Optional[frozenset] = None
-
-
-@dataclass(frozen=True)
-class CommentsConfig:
-    """Top-level comment access config."""
-    enabled: bool = True
-    policy: str = "pairing"
-    allow_from: frozenset = field(default_factory=frozenset)
-    documents: Dict[str, CommentDocumentRule] = field(default_factory=dict)
-
-
-@dataclass(frozen=True)
-class ResolvedCommentRule:
-    """Fully resolved rule after field-by-field fallback."""
-    enabled: bool
-    policy: str
-    allow_from: frozenset
-    match_source: str  # e.g. "exact:docx:xxx" | "wildcard" | "top" | "default"
-
-
-# ---------------------------------------------------------------------------
-# Mtime-cached file loading
-# ---------------------------------------------------------------------------
-
-class _MtimeCache:
-    """Generic mtime-based file cache.  ``stat()`` per access, re-read only on change."""
-
-    def __init__(self, path: Path):
-        self._path = path
-        self._mtime: float = 0.0
-        self._data: Optional[dict] = None
-
-    def load(self) -> dict:
-        try:
-            st = self._path.stat()
-            mtime = st.st_mtime
-        except FileNotFoundError:
-            self._mtime = 0.0
-            self._data = {}
-            return {}
-
-        if mtime == self._mtime and self._data is not None:
-            return self._data
-
-        try:
-            with open(self._path, "r", encoding="utf-8") as f:
-                data = json.load(f)
-            if not isinstance(data, dict):
-                data = {}
-        except (json.JSONDecodeError, OSError):
-            logger.warning("[Feishu-Rules] Failed to read %s, using empty config", self._path)
-            data = {}
-
-        self._mtime = mtime
-        self._data = data
-        return data
-
-
-_rules_cache = _MtimeCache(RULES_FILE)
-_pairing_cache = _MtimeCache(PAIRING_FILE)
-
-
-# ---------------------------------------------------------------------------
-# Config parsing
-# ---------------------------------------------------------------------------
-
-def _parse_frozenset(raw: Any) -> Optional[frozenset]:
-    """Parse a list of strings into a frozenset; return None if key absent."""
-    if raw is None:
-        return None
-    if isinstance(raw, (list, tuple)):
-        return frozenset(str(u).strip() for u in raw if str(u).strip())
-    return None
-
-
-def _parse_document_rule(raw: dict) -> CommentDocumentRule:
-    enabled = raw.get("enabled")
-    if enabled is not None:
-        enabled = bool(enabled)
-    policy = raw.get("policy")
-    if policy is not None:
-        policy = str(policy).strip().lower()
-        if policy not in _VALID_POLICIES:
-            policy = None
-    allow_from = _parse_frozenset(raw.get("allow_from"))
-    return CommentDocumentRule(enabled=enabled, policy=policy, allow_from=allow_from)
-
-
-def load_config() -> CommentsConfig:
-    """Load comment rules from disk (mtime-cached)."""
-    raw = _rules_cache.load()
-    if not raw:
-        return CommentsConfig()
-
-    documents: Dict[str, CommentDocumentRule] = {}
-    raw_docs = raw.get("documents", {})
-    if isinstance(raw_docs, dict):
-        for key, rule_raw in raw_docs.items():
-            if isinstance(rule_raw, dict):
-                documents[str(key)] = _parse_document_rule(rule_raw)
-
-    policy = str(raw.get("policy", "pairing")).strip().lower()
-    if policy not in _VALID_POLICIES:
-        policy = "pairing"
-
-    return CommentsConfig(
-        enabled=raw.get("enabled", True),
-        policy=policy,
-        allow_from=_parse_frozenset(raw.get("allow_from")) or frozenset(),
-        documents=documents,
-    )
-
-
-# ---------------------------------------------------------------------------
-# Rule resolution  (§8.4 field-by-field fallback)
-# ---------------------------------------------------------------------------
-
-def has_wiki_keys(cfg: CommentsConfig) -> bool:
-    """Check if any document rule key starts with 'wiki:'."""
-    return any(k.startswith("wiki:") for k in cfg.documents)
-
-
-def resolve_rule(
-    cfg: CommentsConfig,
-    file_type: str,
-    file_token: str,
-    wiki_token: str = "",
-) -> ResolvedCommentRule:
-    """Resolve effective rule: exact doc → wiki key → wildcard → top-level → defaults."""
-    exact_key = f"{file_type}:{file_token}"
-
-    exact = cfg.documents.get(exact_key)
-    exact_src = f"exact:{exact_key}"
-    if exact is None and wiki_token:
-        wiki_key = f"wiki:{wiki_token}"
-        exact = cfg.documents.get(wiki_key)
-        exact_src = f"exact:{wiki_key}"
-
-    wildcard = cfg.documents.get("*")
-
-    layers = []
-    if exact is not None:
-        layers.append((exact, exact_src))
-    if wildcard is not None:
-        layers.append((wildcard, "wildcard"))
-
-    def _pick(field_name: str):
-        for layer, source in layers:
-            val = getattr(layer, field_name)
-            if val is not None:
-                return val, source
-        return getattr(cfg, field_name), "top"
-
-    enabled, en_src = _pick("enabled")
-    policy, pol_src = _pick("policy")
-    allow_from, _ = _pick("allow_from")
-
-    # match_source = highest-priority tier that contributed any field
-    priority_order = {"exact": 0, "wildcard": 1, "top": 2}
-    best_src = min(
-        [en_src, pol_src],
-        key=lambda s: priority_order.get(s.split(":")[0], 3),
-    )
-
-    return ResolvedCommentRule(
-        enabled=enabled,
-        policy=policy,
-        allow_from=allow_from,
-        match_source=best_src,
-    )
-
-
-# ---------------------------------------------------------------------------
-# Pairing store
-# ---------------------------------------------------------------------------
-
-def _load_pairing_approved() -> set:
-    """Return set of approved user open_ids (mtime-cached)."""
-    data = _pairing_cache.load()
-    approved = data.get("approved", {})
-    if isinstance(approved, dict):
-        return set(approved.keys())
-    if isinstance(approved, list):
-        return set(str(u) for u in approved if u)
-    return set()
-
-
-def _save_pairing(data: dict) -> None:
-    PAIRING_FILE.parent.mkdir(parents=True, exist_ok=True)
-    tmp = PAIRING_FILE.with_suffix(".tmp")
-    with open(tmp, "w", encoding="utf-8") as f:
-        json.dump(data, f, indent=2, ensure_ascii=False)
-    tmp.replace(PAIRING_FILE)
-    # Invalidate cache so next load picks up change
-    _pairing_cache._mtime = 0.0
-    _pairing_cache._data = None
-
-
-def pairing_add(user_open_id: str) -> bool:
-    """Add a user to the pairing-approved list. Returns True if newly added."""
-    data = _pairing_cache.load()
-    approved = data.get("approved", {})
-    if not isinstance(approved, dict):
-        approved = {}
-    if user_open_id in approved:
-        return False
-    approved[user_open_id] = {"approved_at": time.time()}
-    data["approved"] = approved
-    _save_pairing(data)
-    return True
-
-
-def pairing_remove(user_open_id: str) -> bool:
-    """Remove a user from the pairing-approved list. Returns True if removed."""
-    data = _pairing_cache.load()
-    approved = data.get("approved", {})
-    if not isinstance(approved, dict):
-        return False
-    if user_open_id not in approved:
-        return False
-    del approved[user_open_id]
-    data["approved"] = approved
-    _save_pairing(data)
-    return True
-
-
-def pairing_list() -> Dict[str, Any]:
-    """Return the approved dict  {user_open_id: {approved_at: ...}}."""
-    data = _pairing_cache.load()
-    approved = data.get("approved", {})
-    return dict(approved) if isinstance(approved, dict) else {}
-
-
-# ---------------------------------------------------------------------------
-# Access check  (public API for feishu_comment.py)
-# ---------------------------------------------------------------------------
-
-def is_user_allowed(rule: ResolvedCommentRule, user_open_id: str) -> bool:
-    """Check if user passes the resolved rule's policy gate."""
-    if user_open_id in rule.allow_from:
-        return True
-    if rule.policy == "pairing":
-        return user_open_id in _load_pairing_approved()
-    return False
-
-
-# ---------------------------------------------------------------------------
-# CLI
-# ---------------------------------------------------------------------------
-
-def _print_status() -> None:
-    cfg = load_config()
-    print(f"Rules file: {RULES_FILE}")
-    print(f"  exists: {RULES_FILE.exists()}")
-    print(f"Pairing file: {PAIRING_FILE}")
-    print(f"  exists: {PAIRING_FILE.exists()}")
-    print()
-    print(f"Top-level:")
-    print(f"  enabled:    {cfg.enabled}")
-    print(f"  policy:     {cfg.policy}")
-    print(f"  allow_from: {sorted(cfg.allow_from) if cfg.allow_from else '[]'}")
-    print()
-    if cfg.documents:
-        print(f"Document rules ({len(cfg.documents)}):")
-        for key, rule in sorted(cfg.documents.items()):
-            parts = []
-            if rule.enabled is not None:
-                parts.append(f"enabled={rule.enabled}")
-            if rule.policy is not None:
-                parts.append(f"policy={rule.policy}")
-            if rule.allow_from is not None:
-                parts.append(f"allow_from={sorted(rule.allow_from)}")
-            print(f"  [{key}] {', '.join(parts) if parts else '(empty — inherits all)'}")
-    else:
-        print("Document rules: (none)")
-    print()
-    approved = pairing_list()
-    print(f"Pairing approved ({len(approved)}):")
-    for uid, meta in sorted(approved.items()):
-        ts = meta.get("approved_at", 0)
-        print(f"  {uid}  (approved_at={ts})")
-
-
-def _do_check(doc_key: str, user_open_id: str) -> None:
-    cfg = load_config()
-    parts = doc_key.split(":", 1)
-    if len(parts) != 2:
-        print(f"Error: doc_key must be 'fileType:fileToken', got '{doc_key}'")
-        return
-    file_type, file_token = parts
-    rule = resolve_rule(cfg, file_type, file_token)
-    allowed = is_user_allowed(rule, user_open_id)
-    print(f"Document:     {doc_key}")
-    print(f"User:         {user_open_id}")
-    print(f"Resolved rule:")
-    print(f"  enabled:      {rule.enabled}")
-    print(f"  policy:       {rule.policy}")
-    print(f"  allow_from:   {sorted(rule.allow_from) if rule.allow_from else '[]'}")
-    print(f"  match_source: {rule.match_source}")
-    print(f"Result:       {'ALLOWED' if allowed else 'DENIED'}")
-
-
-def _main() -> int:
-    import sys
-
-    try:
-        from hermes_cli.env_loader import load_hermes_dotenv
-        load_hermes_dotenv()
-    except Exception:
-        pass
-
-    usage = (
-        "Usage: python -m gateway.platforms.feishu_comment_rules <command> [args]\n"
-        "\n"
-        "Commands:\n"
-        "  status                              Show rules config and pairing state\n"
-        "  check <fileType:token> <user>        Simulate access check\n"
-        "  pairing add <user_open_id>           Add user to pairing-approved list\n"
-        "  pairing remove <user_open_id>        Remove user from pairing-approved list\n"
-        "  pairing list                         List pairing-approved users\n"
-        "\n"
-        f"Rules config file: {RULES_FILE}\n"
-        "  Edit this JSON file directly to configure policies and document rules.\n"
-        "  Changes take effect on the next comment event (no restart needed).\n"
-    )
-
-    args = sys.argv[1:]
-    if not args:
-        print(usage)
-        return 1
-
-    cmd = args[0]
-
-    if cmd == "status":
-        _print_status()
-
-    elif cmd == "check":
-        if len(args) < 3:
-            print("Usage: check <fileType:fileToken> <user_open_id>")
-            return 1
-        _do_check(args[1], args[2])
-
-    elif cmd == "pairing":
-        if len(args) < 2:
-            print("Usage: pairing <add|remove|list> [args]")
-            return 1
-        sub = args[1]
-        if sub == "add":
-            if len(args) < 3:
-                print("Usage: pairing add <user_open_id>")
-                return 1
-            if pairing_add(args[2]):
-                print(f"Added: {args[2]}")
-            else:
-                print(f"Already approved: {args[2]}")
-        elif sub == "remove":
-            if len(args) < 3:
-                print("Usage: pairing remove <user_open_id>")
-                return 1
-            if pairing_remove(args[2]):
-                print(f"Removed: {args[2]}")
-            else:
-                print(f"Not in approved list: {args[2]}")
-        elif sub == "list":
-            approved = pairing_list()
-            if not approved:
-                print("(no approved users)")
-            for uid, meta in sorted(approved.items()):
-                print(f"  {uid}  approved_at={meta.get('approved_at', '?')}")
-        else:
-            print(f"Unknown pairing subcommand: {sub}")
-            return 1
-    else:
-        print(f"Unknown command: {cmd}\n")
-        print(usage)
-        return 1
-    return 0
-
-
-if __name__ == "__main__":
-    import sys
-    sys.exit(_main())
@@ -1,264 +0,0 @@
-"""Shared helper classes for gateway platform adapters.
-
-Extracts common patterns that were duplicated across 5-7 adapters:
-message deduplication, text batch aggregation, markdown stripping,
-and thread participation tracking.
-"""
-
-import asyncio
-import json
-import logging
-import re
-import time
-from pathlib import Path
-from typing import TYPE_CHECKING, Dict, Optional
-
-if TYPE_CHECKING:
-    from gateway.platforms.base import BasePlatformAdapter, MessageEvent
-
-logger = logging.getLogger(__name__)
-
-
-# ─── Message Deduplication ────────────────────────────────────────────────────
-
-
-class MessageDeduplicator:
-    """TTL-based message deduplication cache.
-
-    Replaces the identical ``_seen_messages`` / ``_is_duplicate()`` pattern
-    previously duplicated in discord, slack, dingtalk, wecom, weixin,
-    mattermost, and feishu adapters.
-
-    Usage::
-
-        self._dedup = MessageDeduplicator()
-
-        # In message handler:
-        if self._dedup.is_duplicate(msg_id):
-            return
-    """
-
-    def __init__(self, max_size: int = 2000, ttl_seconds: float = 300):
-        self._seen: Dict[str, float] = {}
-        self._max_size = max_size
-        self._ttl = ttl_seconds
-
-    def is_duplicate(self, msg_id: str) -> bool:
-        """Return True if *msg_id* was already seen within the TTL window."""
-        if not msg_id:
-            return False
-        now = time.time()
-        if msg_id in self._seen:
-            if now - self._seen[msg_id] < self._ttl:
-                return True
-            # Entry has expired — remove it and treat as new
-            del self._seen[msg_id]
-        self._seen[msg_id] = now
-        if len(self._seen) > self._max_size:
-            cutoff = now - self._ttl
-            self._seen = {k: v for k, v in self._seen.items() if v > cutoff}
-        return False
-
-    def clear(self):
-        """Clear all tracked messages."""
-        self._seen.clear()
-
-
-# ─── Text Batch Aggregation ──────────────────────────────────────────────────
-
-
-class TextBatchAggregator:
-    """Aggregates rapid-fire text events into single messages.
-
-    Replaces the ``_enqueue_text_event`` / ``_flush_text_batch`` pattern
-    previously duplicated in telegram, discord, matrix, wecom, and feishu.
-
-    Usage::
-
-        self._text_batcher = TextBatchAggregator(
-            handler=self._message_handler,
-            batch_delay=0.6,
-            split_threshold=1900,
-        )
-
-        # In message dispatch:
-        if msg_type == MessageType.TEXT and self._text_batcher.is_enabled():
-            self._text_batcher.enqueue(event, session_key)
-            return
-    """
-
-    def __init__(
-        self,
-        handler,
-        *,
-        batch_delay: float = 0.6,
-        split_delay: float = 2.0,
-        split_threshold: int = 4000,
-    ):
-        self._handler = handler
-        self._batch_delay = batch_delay
-        self._split_delay = split_delay
-        self._split_threshold = split_threshold
-        self._pending: Dict[str, "MessageEvent"] = {}
-        self._pending_tasks: Dict[str, asyncio.Task] = {}
-
-    def is_enabled(self) -> bool:
-        """Return True if batching is active (delay > 0)."""
-        return self._batch_delay > 0
-
-    def enqueue(self, event: "MessageEvent", key: str) -> None:
-        """Add *event* to the pending batch for *key*."""
-        chunk_len = len(event.text or "")
-        existing = self._pending.get(key)
-        if not existing:
-            event._last_chunk_len = chunk_len  # type: ignore[attr-defined]
-            self._pending[key] = event
-        else:
-            existing.text = f"{existing.text}\n{event.text}"
-            existing._last_chunk_len = chunk_len  # type: ignore[attr-defined]
-
-        # Cancel prior flush timer, start a new one
-        prior = self._pending_tasks.get(key)
-        if prior and not prior.done():
-            prior.cancel()
-        self._pending_tasks[key] = asyncio.create_task(self._flush(key))
-
-    async def _flush(self, key: str) -> None:
-        """Wait then dispatch the batched event for *key*."""
-        current_task = self._pending_tasks.get(key)
-        pending = self._pending.get(key)
-        last_len = getattr(pending, "_last_chunk_len", 0) if pending else 0
-
-        # Use longer delay when the last chunk looks like a split message
-        delay = self._split_delay if last_len >= self._split_threshold else self._batch_delay
-        await asyncio.sleep(delay)
-
-        event = self._pending.pop(key, None)
-        if event:
-            try:
-                await self._handler(event)
-            except Exception:
-                logger.exception("[TextBatchAggregator] Error dispatching batched event for %s", key)
-
-        if self._pending_tasks.get(key) is current_task:
-            self._pending_tasks.pop(key, None)
-
-    def cancel_all(self) -> None:
-        """Cancel all pending flush tasks."""
-        for task in self._pending_tasks.values():
-            if not task.done():
-                task.cancel()
-        self._pending_tasks.clear()
-        self._pending.clear()
-
-
-# ─── Markdown Stripping ──────────────────────────────────────────────────────
-
-# Pre-compiled regexes for performance
-_RE_BOLD = re.compile(r"\*\*(.+?)\*\*", re.DOTALL)
-_RE_ITALIC_STAR = re.compile(r"\*(.+?)\*", re.DOTALL)
-_RE_BOLD_UNDER = re.compile(r"__(.+?)__", re.DOTALL)
-_RE_ITALIC_UNDER = re.compile(r"_(.+?)_", re.DOTALL)
-_RE_CODE_BLOCK = re.compile(r"```[a-zA-Z0-9_+-]*\n?")
-_RE_INLINE_CODE = re.compile(r"`(.+?)`")
-_RE_HEADING = re.compile(r"^#{1,6}\s+", re.MULTILINE)
-_RE_LINK = re.compile(r"\[([^\]]+)\]\([^\)]+\)")
-_RE_MULTI_NEWLINE = re.compile(r"\n{3,}")
-
-
-def strip_markdown(text: str) -> str:
-    """Strip markdown formatting for plain-text platforms (SMS, iMessage, etc.).
-
-    Replaces the identical ``_strip_markdown()`` functions previously
-    duplicated in sms.py, bluebubbles.py, and feishu.py.
-    """
-    text = _RE_BOLD.sub(r"\1", text)
-    text = _RE_ITALIC_STAR.sub(r"\1", text)
-    text = _RE_BOLD_UNDER.sub(r"\1", text)
-    text = _RE_ITALIC_UNDER.sub(r"\1", text)
-    text = _RE_CODE_BLOCK.sub("", text)
-    text = _RE_INLINE_CODE.sub(r"\1", text)
-    text = _RE_HEADING.sub("", text)
-    text = _RE_LINK.sub(r"\1", text)
-    text = _RE_MULTI_NEWLINE.sub("\n\n", text)
-    return text.strip()
-
-
-# ─── Thread Participation Tracking ───────────────────────────────────────────
-
-
-class ThreadParticipationTracker:
-    """Persistent tracking of threads the bot has participated in.
-
-    Replaces the identical ``_load/_save_participated_threads`` +
-    ``_mark_thread_participated`` pattern previously duplicated in
-    discord.py and matrix.py.
-
-    Usage::
-
-        self._threads = ThreadParticipationTracker("discord")
-
-        # Check membership:
-        if thread_id in self._threads:
-            ...
-
-        # Mark participation:
-        self._threads.mark(thread_id)
-    """
-
-    _MAX_TRACKED = 500
-
-    def __init__(self, platform_name: str, max_tracked: int = 500):
-        self._platform = platform_name
-        self._max_tracked = max_tracked
-        self._threads: set = self._load()
-
-    def _state_path(self) -> Path:
-        from hermes_constants import get_hermes_home
-        return get_hermes_home() / f"{self._platform}_threads.json"
-
-    def _load(self) -> set:
-        path = self._state_path()
-        if path.exists():
-            try:
-                return set(json.loads(path.read_text(encoding="utf-8")))
-            except Exception:
-                pass
-        return set()
-
-    def _save(self) -> None:
-        path = self._state_path()
-        path.parent.mkdir(parents=True, exist_ok=True)
-        thread_list = list(self._threads)
-        if len(thread_list) > self._max_tracked:
-            thread_list = thread_list[-self._max_tracked:]
-            self._threads = set(thread_list)
-        path.write_text(json.dumps(thread_list), encoding="utf-8")
-
-    def mark(self, thread_id: str) -> None:
-        """Mark *thread_id* as participated and persist."""
-        if thread_id not in self._threads:
-            self._threads.add(thread_id)
-            self._save()
-
-    def __contains__(self, thread_id: str) -> bool:
-        return thread_id in self._threads
-
-    def clear(self) -> None:
-        self._threads.clear()
-
-
-# ─── Phone Number Redaction ──────────────────────────────────────────────────
-
-
-def redact_phone(phone: str) -> str:
-    """Redact a phone number for logging, preserving country code and last 4.
-
-    Replaces the identical ``_redact_phone()`` functions in signal.py,
-    sms.py, and bluebubbles.py.
-    """
-    if not phone:
-        return "<none>"
-    if len(phone) <= 8:
-        return phone[:2] + "****" + phone[-2:] if len(phone) > 4 else "****"
-    return phone[:4] + "****" + phone[-4:]
@@ -18,11 +18,11 @@ import json
 import logging
 import os
 import re
+import time
 from pathlib import Path
 from typing import Any, Dict, List, Optional

 from gateway.config import Platform, PlatformConfig
-from gateway.platforms.helpers import MessageDeduplicator
 from gateway.platforms.base import (
    BasePlatformAdapter,
    MessageEvent,
@@ -96,8 +96,10 @@ class MattermostAdapter(BasePlatformAdapter):
            or os.getenv("MATTERMOST_REPLY_MODE", "off")
        ).lower()

-        # Dedup cache (prevent reprocessing)
-        self._dedup = MessageDeduplicator()
+        # Dedup cache: post_id → timestamp (prevent reprocessing)
+        self._seen_posts: Dict[str, float] = {}
+        self._SEEN_MAX = 2000
+        self._SEEN_TTL = 300  # 5 minutes

    # ------------------------------------------------------------------
    # HTTP helpers
@@ -602,8 +604,10 @@ class MattermostAdapter(BasePlatformAdapter):
        post_id = post.get("id", "")

        # Dedup.
-        if self._dedup.is_duplicate(post_id):
+        self._prune_seen()
+        if post_id in self._seen_posts:
            return
+        self._seen_posts[post_id] = time.time()

        # Build message event.
        channel_id = post.get("channel_id", "")
@@ -718,12 +722,6 @@ class MattermostAdapter(BasePlatformAdapter):
            thread_id=thread_id,
        )

-        # Per-channel ephemeral prompt
-        from gateway.platforms.base import resolve_channel_prompt
-        _channel_prompt = resolve_channel_prompt(
-            self.config.extra, channel_id, None,
-        )
-
        msg_event = MessageEvent(
            text=message_text,
            message_type=msg_type,
@@ -732,9 +730,17 @@ class MattermostAdapter(BasePlatformAdapter):
            message_id=post_id,
            media_urls=media_urls if media_urls else None,
            media_types=media_types if media_types else None,
-            channel_prompt=_channel_prompt,
        )

        await self.handle_message(msg_event)

-
+    def _prune_seen(self) -> None:
+        """Remove expired entries from the dedup cache."""
+        if len(self._seen_posts) < self._SEEN_MAX:
+            return
+        now = time.time()
+        self._seen_posts = {
+            pid: ts
+            for pid, ts in self._seen_posts.items()
+            if now - ts < self._SEEN_TTL
+        }
@@ -1,57 +0,0 @@
-"""
-QQBot platform package.
-
-Re-exports the main adapter symbols from ``adapter.py`` (the original
-``qqbot.py``) so that **all existing import paths remain unchanged**::
-
-    from gateway.platforms.qqbot import QQAdapter          # works
-    from gateway.platforms.qqbot import check_qq_requirements  # works
-
-New modules:
-    - ``constants`` — shared constants (API URLs, timeouts, message types)
-    - ``utils`` — User-Agent builder, config helpers
-    - ``crypto`` — AES-256-GCM key generation and decryption
-    - ``onboard`` — QR-code scan-to-configure flow
-"""
-
-# -- Adapter (original qqbot.py) ------------------------------------------
-from .adapter import (  # noqa: F401
-    QQAdapter,
-    QQCloseError,
-    check_qq_requirements,
-    _coerce_list,
-    _ssrf_redirect_guard,
-)
-
-# -- Onboard (QR-code scan-to-configure) -----------------------------------
-from .onboard import (  # noqa: F401
-    BindStatus,
-    create_bind_task,
-    poll_bind_result,
-    build_connect_url,
-)
-from .crypto import decrypt_secret, generate_bind_key  # noqa: F401
-
-# -- Utils -----------------------------------------------------------------
-from .utils import build_user_agent, get_api_headers, coerce_list  # noqa: F401
-
-__all__ = [
-    # adapter
-    "QQAdapter",
-    "QQCloseError",
-    "check_qq_requirements",
-    "_coerce_list",
-    "_ssrf_redirect_guard",
-    # onboard
-    "BindStatus",
-    "create_bind_task",
-    "poll_bind_result",
-    "build_connect_url",
-    # crypto
-    "decrypt_secret",
-    "generate_bind_key",
-    # utils
-    "build_user_agent",
-    "get_api_headers",
-    "coerce_list",
-]
@@ -1,74 +0,0 @@
-"""QQBot package-level constants shared across adapter, onboard, and other modules."""
-
-from __future__ import annotations
-
-import os
-
-# ---------------------------------------------------------------------------
-# QQBot adapter version — bump on functional changes to the adapter package.
-# ---------------------------------------------------------------------------
-
-QQBOT_VERSION = "1.1.0"
-
-# ---------------------------------------------------------------------------
-# API endpoints
-# ---------------------------------------------------------------------------
-
-# The portal domain is configurable via QQ_API_HOST for corporate proxies
-# or test environments.  Default: q.qq.com (production).
-PORTAL_HOST = os.getenv("QQ_PORTAL_HOST", "q.qq.com")
-
-API_BASE = "https://api.sgroup.qq.com"
-TOKEN_URL = "https://bots.qq.com/app/getAppAccessToken"
-GATEWAY_URL_PATH = "/gateway"
-
-# QR-code onboard endpoints (on the portal host)
-ONBOARD_CREATE_PATH = "/lite/create_bind_task"
-ONBOARD_POLL_PATH = "/lite/poll_bind_result"
-QR_URL_TEMPLATE = (
-    "https://q.qq.com/qqbot/openclaw/connect.html"
-    "?task_id={task_id}&_wv=2&source=hermes"
-)
-
-# ---------------------------------------------------------------------------
-# Timeouts & retry
-# ---------------------------------------------------------------------------
-
-DEFAULT_API_TIMEOUT = 30.0
-FILE_UPLOAD_TIMEOUT = 120.0
-CONNECT_TIMEOUT_SECONDS = 20.0
-
-RECONNECT_BACKOFF = [2, 5, 10, 30, 60]
-MAX_RECONNECT_ATTEMPTS = 100
-RATE_LIMIT_DELAY = 60  # seconds
-QUICK_DISCONNECT_THRESHOLD = 5.0  # seconds
-MAX_QUICK_DISCONNECT_COUNT = 3
-
-ONBOARD_POLL_INTERVAL = 2.0  # seconds between poll_bind_result calls
-ONBOARD_API_TIMEOUT = 10.0
-
-# ---------------------------------------------------------------------------
-# Message limits
-# ---------------------------------------------------------------------------
-
-MAX_MESSAGE_LENGTH = 4000
-DEDUP_WINDOW_SECONDS = 300
-DEDUP_MAX_SIZE = 1000
-
-# ---------------------------------------------------------------------------
-# QQ Bot message types
-# ---------------------------------------------------------------------------
-
-MSG_TYPE_TEXT = 0
-MSG_TYPE_MARKDOWN = 2
-MSG_TYPE_MEDIA = 7
-MSG_TYPE_INPUT_NOTIFY = 6
-
-# ---------------------------------------------------------------------------
-# QQ Bot file media types
-# ---------------------------------------------------------------------------
-
-MEDIA_TYPE_IMAGE = 1
-MEDIA_TYPE_VIDEO = 2
-MEDIA_TYPE_VOICE = 3
-MEDIA_TYPE_FILE = 4
@@ -1,45 +0,0 @@
-"""AES-256-GCM utilities for QQBot scan-to-configure credential decryption."""
-
-from __future__ import annotations
-
-import base64
-import os
-
-
-def generate_bind_key() -> str:
-    """Generate a 256-bit random AES key and return it as base64.
-
-    The key is passed to ``create_bind_task`` so the server can encrypt
-    the bot's *client_secret* before returning it.  Only this CLI holds
-    the key, ensuring the secret never travels in plaintext.
-    """
-    return base64.b64encode(os.urandom(32)).decode()
-
-
-def decrypt_secret(encrypted_base64: str, key_base64: str) -> str:
-    """Decrypt a base64-encoded AES-256-GCM ciphertext.
-
-    Ciphertext layout (after base64-decoding)::
-
-        IV (12 bytes) ‖ ciphertext (N bytes) ‖ AuthTag (16 bytes)
-
-    Args:
-        encrypted_base64: The ``bot_encrypt_secret`` value from
-            ``poll_bind_result``.
-        key_base64: The base64 AES key generated by
-            :func:`generate_bind_key`.
-
-    Returns:
-        The decrypted *client_secret* as a UTF-8 string.
-    """
-    from cryptography.hazmat.primitives.ciphers.aead import AESGCM
-
-    key = base64.b64decode(key_base64)
-    raw = base64.b64decode(encrypted_base64)
-
-    iv = raw[:12]
-    ciphertext_with_tag = raw[12:]  # AESGCM expects ciphertext + tag concatenated
-
-    aesgcm = AESGCM(key)
-    plaintext = aesgcm.decrypt(iv, ciphertext_with_tag, None)
-    return plaintext.decode("utf-8")
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
alt-glitch	f7d57c0108	fix: pass terminal_lifetime through to Modal sandbox timeout The lifetime_seconds config was computed and stored as an env var but never reached Modal's Sandbox.create(timeout=...), causing sandboxes to always die after the hardcoded 1h default regardless of configuration.	2026-04-13 11:57:39 -07:00
alt-glitch	d77783d198	fix: use hermes agent system prompt and nudges	2026-04-08 18:34:01 -07:00
alt-glitch	4af69097f2	add top_p + user nudges for incorrect format	2026-04-08 18:34:01 -07:00
alt-glitch	59471b79e5	update trajectory writing	2026-04-08 18:33:02 -07:00
alt-glitch	0e459f2b7b	Update default.yaml	2026-04-08 18:33:02 -07:00
alt-glitch	3befb9389f	wip: run tb2 and fix modal instantiation	2026-04-08 18:33:02 -07:00