fix(tests): provide model name in Codex 401 refresh tests for CI

CI has no config.yaml, so cron/gateway resolve an empty model name.
The Codex Responses validator rejects empty models before the mock
API call is reached. Provide explicit model in job dict and env var.

.github/workflows/deploy-site.yml

@@ -19,6 +19,8 @@ concurrency:
 
 jobs:
   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

.github/workflows/docker-publish.yml

@@ -12,6 +12,8 @@ concurrency:
 
 jobs:
   build-and-push:
+    # Only run on the upstream repository, not on forks
+    if: github.repository == 'NousResearch/hermes-agent'
     runs-on: ubuntu-latest
     timeout-minutes: 30
     steps:

MANIFEST.in

@@ -0,0 +1,4 @@
+graft skills
+graft optional-skills
+global-exclude __pycache__
+global-exclude *.py[cod]

agent/anthropic_adapter.py

@@ -162,6 +162,21 @@ def _is_oauth_token(key: str) -> bool:
     return True
 
 
+def _is_third_party_anthropic_endpoint(base_url: str | None) -> bool:
+    """Return True for non-Anthropic endpoints using the Anthropic Messages API.
+
+    Third-party proxies (Azure AI Foundry, AWS Bedrock, self-hosted) authenticate
+    with their own API keys via x-api-key, not Anthropic OAuth tokens. OAuth
+    detection should be skipped for these endpoints.
+    """
+    if not base_url:
+        return False  # No base_url = direct Anthropic API
+    normalized = base_url.rstrip("/").lower()
+    if "anthropic.com" in normalized:
+        return False  # Direct Anthropic API — OAuth applies
+    return True  # Any other endpoint is a third-party proxy
+
+
 def _requires_bearer_auth(base_url: str | None) -> bool:
     """Return True for Anthropic-compatible providers that require Bearer auth.
 
@@ -205,6 +220,14 @@ def build_anthropic_client(api_key: str, base_url: str = None):
         kwargs["auth_token"] = api_key
         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)}
     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;
@@ -338,7 +361,14 @@ def _refresh_oauth_token(creds: Dict[str, Any]) -> Optional[str]:
 
                 if new_access:
                     new_expires_ms = int(time.time() * 1000) + (expires_in * 1000)
-                    _write_claude_code_credentials(new_access, new_refresh, new_expires_ms)
+                    # Parse scopes from refresh response — Claude Code >=2.1.81
+                    # requires a "scopes" field in the credential store and checks
+                    # for "user:inference" before accepting the token as valid.
+                    scope_str = result.get("scope", "")
+                    scopes = scope_str.split() if scope_str else None
+                    _write_claude_code_credentials(
+                        new_access, new_refresh, new_expires_ms, scopes=scopes,
+                    )
                     logger.debug("Refreshed Claude Code OAuth token via %s", endpoint)
                     return new_access
         except Exception as e:
@@ -347,8 +377,20 @@ def _refresh_oauth_token(creds: Dict[str, Any]) -> Optional[str]:
     return None
 
 
-def _write_claude_code_credentials(access_token: str, refresh_token: str, expires_at_ms: int) -> None:
-    """Write refreshed credentials back to ~/.claude/.credentials.json."""
+def _write_claude_code_credentials(
+    access_token: str,
+    refresh_token: str,
+    expires_at_ms: int,
+    *,
+    scopes: Optional[list] = None,
+) -> None:
+    """Write refreshed credentials back to ~/.claude/.credentials.json.
+
+    The optional *scopes* list (e.g. ``["user:inference", "user:profile", ...]``)
+    is persisted so that Claude Code's own auth check recognises the credential
+    as valid.  Claude Code >=2.1.81 gates on the presence of ``"user:inference"``
+    in the stored scopes before it will use the token.
+    """
     cred_path = Path.home() / ".claude" / ".credentials.json"
     try:
         # Read existing file to preserve other fields
@@ -356,11 +398,19 @@ def _write_claude_code_credentials(access_token: str, refresh_token: str, expire
         if cred_path.exists():
             existing = json.loads(cred_path.read_text(encoding="utf-8"))
 
-        existing["claudeAiOauth"] = {
+        oauth_data: Dict[str, Any] = {
             "accessToken": access_token,
             "refreshToken": refresh_token,
             "expiresAt": expires_at_ms,
         }
+        if scopes is not None:
+            oauth_data["scopes"] = scopes
+        elif "claudeAiOauth" in existing and "scopes" in existing["claudeAiOauth"]:
+            # Preserve previously-stored scopes when the refresh response
+            # does not include a scope field.
+            oauth_data["scopes"] = existing["claudeAiOauth"]["scopes"]
+
+        existing["claudeAiOauth"] = oauth_data
 
         cred_path.parent.mkdir(parents=True, exist_ok=True)
         cred_path.write_text(json.dumps(existing, indent=2), encoding="utf-8")

agent/model_metadata.py

@@ -176,6 +176,7 @@ _URL_TO_PROVIDER: Dict[str, str] = {
     "api.deepseek.com": "deepseek",
     "api.githubcopilot.com": "copilot",
     "models.github.ai": "copilot",
+    "api.fireworks.ai": "fireworks",
 }
 
 

agent/models_dev.py

@@ -43,6 +43,7 @@ PROVIDER_TO_MODELS_DEV: Dict[str, str] = {
     "opencode-zen": "opencode",
     "opencode-go": "opencode-go",
     "kilocode": "kilo",
+    "fireworks": "fireworks-ai",
 }
 
 

cli.py

@@ -3904,6 +3904,8 @@ class HermesCLI:
             self._handle_stop_command()
         elif canonical == "background":
             self._handle_background_command(cmd_original)
+        elif canonical == "btw":
+            self._handle_btw_command(cmd_original)
         elif canonical == "queue":
             # Extract prompt after "/queue " or "/q "
             parts = cmd_original.split(None, 1)
@@ -4190,6 +4192,121 @@ class HermesCLI:
         self._background_tasks[task_id] = thread
         thread.start()
 
+    def _handle_btw_command(self, cmd: str):
+        """Handle /btw <question> — ephemeral side question using session context.
+
+        Snapshots the current conversation history, spawns a no-tools agent in
+        a background thread, and prints the answer without persisting anything
+        to the main session.
+        """
+        parts = cmd.strip().split(maxsplit=1)
+        if len(parts) < 2 or not parts[1].strip():
+            _cprint("  Usage: /btw <question>")
+            _cprint("  Example: /btw what module owns session title sanitization?")
+            _cprint("  Answers using session context. No tools, not persisted.")
+            return
+
+        question = parts[1].strip()
+        task_id = f"btw_{datetime.now().strftime('%H%M%S')}_{uuid.uuid4().hex[:6]}"
+
+        if not self._ensure_runtime_credentials():
+            _cprint("  (>_<) Cannot start /btw: no valid credentials.")
+            return
+
+        turn_route = self._resolve_turn_agent_config(question)
+        history_snapshot = list(self.conversation_history)
+
+        preview = question[:60] + ("..." if len(question) > 60 else "")
+        _cprint(f'  💬 /btw: "{preview}"')
+
+        def run_btw():
+            try:
+                btw_agent = AIAgent(
+                    model=turn_route["model"],
+                    api_key=turn_route["runtime"].get("api_key"),
+                    base_url=turn_route["runtime"].get("base_url"),
+                    provider=turn_route["runtime"].get("provider"),
+                    api_mode=turn_route["runtime"].get("api_mode"),
+                    acp_command=turn_route["runtime"].get("command"),
+                    acp_args=turn_route["runtime"].get("args"),
+                    max_iterations=8,
+                    enabled_toolsets=[],
+                    quiet_mode=True,
+                    verbose_logging=False,
+                    session_id=task_id,
+                    platform="cli",
+                    reasoning_config=self.reasoning_config,
+                    providers_allowed=self._providers_only,
+                    providers_ignored=self._providers_ignore,
+                    providers_order=self._providers_order,
+                    provider_sort=self._provider_sort,
+                    provider_require_parameters=self._provider_require_params,
+                    provider_data_collection=self._provider_data_collection,
+                    fallback_model=self._fallback_model,
+                    session_db=None,
+                    skip_memory=True,
+                    skip_context_files=True,
+                    persist_session=False,
+                )
+
+                btw_prompt = (
+                    "[Ephemeral /btw side question. Answer using the conversation "
+                    "context. No tools available. Be direct and concise.]\n\n"
+                    + question
+                )
+                result = btw_agent.run_conversation(
+                    user_message=btw_prompt,
+                    conversation_history=history_snapshot,
+                    task_id=task_id,
+                    sync_honcho=False,
+                )
+
+                response = (result.get("final_response") or "") if result else ""
+                if not response and result and result.get("error"):
+                    response = f"Error: {result['error']}"
+
+                # TUI refresh before printing
+                if self._app:
+                    self._app.invalidate()
+                    time.sleep(0.05)
+                print()
+
+                if response:
+                    try:
+                        from hermes_cli.skin_engine import get_active_skin
+                        _skin = get_active_skin()
+                        _resp_color = _skin.get_color("response_border", "#4F6D4A")
+                    except Exception:
+                        _resp_color = "#4F6D4A"
+
+                    ChatConsole().print(Panel(
+                        _rich_text_from_ansi(response),
+                        title=f"[{_resp_color} bold]⚕ /btw[/]",
+                        title_align="left",
+                        border_style=_resp_color,
+                        box=rich_box.HORIZONTALS,
+                        padding=(1, 2),
+                    ))
+                else:
+                    _cprint("  💬 /btw: (no response)")
+
+                if self.bell_on_complete:
+                    sys.stdout.write("\a")
+                    sys.stdout.flush()
+
+            except Exception as e:
+                if self._app:
+                    self._app.invalidate()
+                    time.sleep(0.05)
+                print()
+                _cprint(f"  ❌ /btw failed: {e}")
+            finally:
+                if self._app:
+                    self._invalidate(min_interval=0)
+
+        thread = threading.Thread(target=run_btw, daemon=True, name=f"btw-{task_id}")
+        thread.start()
+
     @staticmethod
     def _try_launch_chrome_debug(port: int, system: str) -> bool:
         """Try to launch Chrome/Chromium with remote debugging enabled.
@@ -7447,6 +7564,20 @@ class HermesCLI:
         # Register atexit cleanup so resources are freed even on unexpected exit
         atexit.register(_run_cleanup)
         
+        # Register signal handlers for graceful shutdown on SSH disconnect / SIGTERM
+        def _signal_handler(signum, frame):
+            """Handle SIGHUP/SIGTERM by triggering graceful cleanup."""
+            logger.debug("Received signal %s, triggering graceful shutdown", signum)
+            raise KeyboardInterrupt()
+        
+        try:
+            import signal as _signal
+            _signal.signal(_signal.SIGTERM, _signal_handler)
+            if hasattr(_signal, 'SIGHUP'):
+                _signal.signal(_signal.SIGHUP, _signal_handler)
+        except Exception:
+            pass  # Signal handlers may fail in restricted environments
+        
         # Install a custom asyncio exception handler that suppresses the
         # "Event loop is closed" RuntimeError from httpx transport cleanup.
         # This is defense-in-depth — the primary fix is neuter_async_httpx_del
@@ -7470,7 +7601,7 @@ class HermesCLI:
                 except Exception:
                     pass
                 app.run()
-        except (EOFError, KeyboardInterrupt):
+        except (EOFError, KeyboardInterrupt, BrokenPipeError):
             pass
         finally:
             self._should_exit = True
@@ -7509,6 +7640,23 @@ class HermesCLI:
                     self._session_db.end_session(self.agent.session_id, "cli_close")
                 except (Exception, KeyboardInterrupt) as e:
                     logger.debug("Could not close session in DB: %s", e)
+            # Plugin hook: on_session_end — safety net for interrupted exits.
+            # run_conversation() already fires this per-turn on normal completion,
+            # so only fire here if the agent was mid-turn (_agent_running) when
+            # the exit occurred, meaning run_conversation's hook didn't fire.
+            if self.agent and getattr(self, '_agent_running', False):
+                try:
+                    from hermes_cli.plugins import invoke_hook as _invoke_hook
+                    _invoke_hook(
+                        "on_session_end",
+                        session_id=self.agent.session_id,
+                        completed=False,
+                        interrupted=True,
+                        model=getattr(self.agent, 'model', None),
+                        platform=getattr(self.agent, 'platform', None) or "cli",
+                    )
+                except Exception:
+                    pass
             _run_cleanup()
             self._print_exit_summary()
 

gateway/platforms/api_server.py

@@ -380,6 +380,7 @@ class APIServerAdapter(BasePlatformAdapter):
         ephemeral_system_prompt: Optional[str] = None,
         session_id: Optional[str] = None,
         stream_delta_callback=None,
+        tool_progress_callback=None,
     ) -> Any:
         """
         Create an AIAgent instance using the gateway's runtime config.
@@ -412,6 +413,7 @@ class APIServerAdapter(BasePlatformAdapter):
             session_id=session_id,
             platform="api_server",
             stream_delta_callback=stream_delta_callback,
+            tool_progress_callback=tool_progress_callback,
         )
         return agent
 
@@ -514,6 +516,15 @@ class APIServerAdapter(BasePlatformAdapter):
                 if delta is not None:
                     _stream_q.put(delta)
 
+            def _on_tool_progress(name, preview, args):
+                """Inject tool progress into the SSE stream for Open WebUI."""
+                if name.startswith("_"):
+                    return  # Skip internal events (_thinking)
+                from agent.display import get_tool_emoji
+                emoji = get_tool_emoji(name)
+                label = preview or name
+                _stream_q.put(f"\n`{emoji} {label}`\n")
+
             # Start agent in background.  agent_ref is a mutable container
             # so the SSE writer can interrupt the agent on client disconnect.
             agent_ref = [None]
@@ -523,6 +534,7 @@ class APIServerAdapter(BasePlatformAdapter):
                 ephemeral_system_prompt=system_prompt,
                 session_id=session_id,
                 stream_delta_callback=_on_delta,
+                tool_progress_callback=_on_tool_progress,
                 agent_ref=agent_ref,
             ))
 
@@ -1194,6 +1206,7 @@ class APIServerAdapter(BasePlatformAdapter):
         ephemeral_system_prompt: Optional[str] = None,
         session_id: Optional[str] = None,
         stream_delta_callback=None,
+        tool_progress_callback=None,
         agent_ref: Optional[list] = None,
     ) -> tuple:
         """
@@ -1214,6 +1227,7 @@ class APIServerAdapter(BasePlatformAdapter):
                 ephemeral_system_prompt=ephemeral_system_prompt,
                 session_id=session_id,
                 stream_delta_callback=stream_delta_callback,
+                tool_progress_callback=tool_progress_callback,
             )
             if agent_ref is not None:
                 agent_ref[0] = agent

gateway/platforms/telegram_network.py

@@ -135,6 +135,9 @@ def _normalize_fallback_ips(values: Iterable[str]) -> list[str]:
         if addr.version != 4:
             logger.warning("Ignoring non-IPv4 Telegram fallback IP: %s", raw)
             continue
+        if addr.is_private or addr.is_loopback or addr.is_link_local or addr.is_unspecified:
+            logger.warning("Ignoring private/internal Telegram fallback IP: %s", raw)
+            continue
         normalized.append(str(addr))
     return normalized
 

gateway/run.py

@@ -325,9 +325,9 @@ def _check_unavailable_skill(command_name: str) -> str | None:
                 )
 
         # Check optional skills (shipped with repo but not installed)
-        from hermes_constants import get_hermes_home
+        from hermes_constants import get_hermes_home, get_optional_skills_dir
         repo_root = Path(__file__).resolve().parent.parent
-        optional_dir = repo_root / "optional-skills"
+        optional_dir = get_optional_skills_dir(repo_root / "optional-skills")
         if optional_dir.exists():
             for skill_md in optional_dir.rglob("SKILL.md"):
                 name = skill_md.parent.name.lower().replace("_", "-")
@@ -476,12 +476,7 @@ class GatewayRunner:
         self._honcho_managers: Dict[str, Any] = {}
         self._honcho_configs: Dict[str, Any] = {}
 
-        # Rate-limit compression warning messages sent to users.
-        # Keyed by chat_id — value is the timestamp of the last warning sent.
-        # Prevents the warning from firing on every message when a session
-        # remains above the threshold after compression.
-        self._compression_warn_sent: Dict[str, float] = {}
-        self._compression_warn_cooldown: int = 3600  # seconds (1 hour)
+
 
         # Ensure tirith security scanner is available (downloads if needed)
         try:
@@ -1967,6 +1962,9 @@ class GatewayRunner:
         if canonical == "background":
             return await self._handle_background_command(event)
 
+        if canonical == "btw":
+            return await self._handle_btw_command(event)
+
         if canonical == "voice":
             return await self._handle_voice_command(event)
 
@@ -2354,18 +2352,7 @@ class GatewayRunner:
                         f"{_compress_token_threshold:,}",
                     )
 
-                    _hyg_adapter = self.adapters.get(source.platform)
                     _hyg_meta = {"thread_id": source.thread_id} if source.thread_id else None
-                    if _hyg_adapter:
-                        try:
-                            await _hyg_adapter.send(
-                                source.chat_id,
-                                f"🗜️ Session is large ({_msg_count} messages, "
-                                f"~{_approx_tokens:,} tokens). Auto-compressing...",
-                                metadata=_hyg_meta,
-                            )
-                        except Exception:
-                            pass
 
                     try:
                         from run_agent import AIAgent
@@ -2426,70 +2413,17 @@ class GatewayRunner:
                                     f"{_approx_tokens:,}", f"{_new_tokens:,}",
                                 )
 
-                                if _hyg_adapter:
-                                    try:
-                                        await _hyg_adapter.send(
-                                            source.chat_id,
-                                            f"🗜️ Compressed: {_msg_count} → "
-                                            f"{_new_count} messages, "
-                                            f"~{_approx_tokens:,} → "
-                                            f"~{_new_tokens:,} tokens",
-                                            metadata=_hyg_meta,
-                                        )
-                                    except Exception:
-                                        pass
-
-                                # Still too large after compression — warn user
-                                # Rate-limited to once per cooldown period per
-                                # chat to avoid spamming on every message.
                                 if _new_tokens >= _warn_token_threshold:
                                     logger.warning(
                                         "Session hygiene: still ~%s tokens after "
-                                        "compression — suggesting /reset",
+                                        "compression",
                                         f"{_new_tokens:,}",
                                     )
-                                    _now = time.time()
-                                    _last_warn = self._compression_warn_sent.get(source.chat_id, 0)
-                                    if _hyg_adapter and _now - _last_warn >= self._compression_warn_cooldown:
-                                        self._compression_warn_sent[source.chat_id] = _now
-                                        try:
-                                            await _hyg_adapter.send(
-                                                source.chat_id,
-                                                "⚠️ Session is still very large "
-                                                "after compression "
-                                                f"(~{_new_tokens:,} tokens). "
-                                                "Consider using /reset to start "
-                                                "fresh if you experience issues.",
-                                                metadata=_hyg_meta,
-                                            )
-                                        except Exception:
-                                            pass
 
                     except Exception as e:
                         logger.warning(
                             "Session hygiene auto-compress failed: %s", e
                         )
-                        # Compression failed and session is dangerously large
-                        if _approx_tokens >= _warn_token_threshold:
-                            _hyg_adapter = self.adapters.get(source.platform)
-                            _hyg_meta = {"thread_id": source.thread_id} if source.thread_id else None
-                            _now = time.time()
-                            _last_warn = self._compression_warn_sent.get(source.chat_id, 0)
-                            if _hyg_adapter and _now - _last_warn >= self._compression_warn_cooldown:
-                                self._compression_warn_sent[source.chat_id] = _now
-                                try:
-                                    await _hyg_adapter.send(
-                                        source.chat_id,
-                                        f"⚠️ Session is very large "
-                                        f"({_msg_count} messages, "
-                                        f"~{_approx_tokens:,} tokens) and "
-                                        "auto-compression failed. Consider "
-                                        "using /compress or /reset to avoid "
-                                        "issues.",
-                                        metadata=_hyg_meta,
-                                    )
-                                except Exception:
-                                    pass
 
         # First-message onboarding -- only on the very first interaction ever
         if not history and not self.session_store.has_any_sessions():
@@ -4107,6 +4041,167 @@ class GatewayRunner:
             except Exception:
                 pass
 
+    async def _handle_btw_command(self, event: MessageEvent) -> str:
+        """Handle /btw <question> — ephemeral side question in the same chat."""
+        question = event.get_command_args().strip()
+        if not question:
+            return (
+                "Usage: /btw <question>\n"
+                "Example: /btw what module owns session title sanitization?\n\n"
+                "Answers using session context. No tools, not persisted."
+            )
+
+        source = event.source
+        session_key = self._session_key_for_source(source)
+
+        # Guard: one /btw at a time per session
+        existing = getattr(self, "_active_btw_tasks", {}).get(session_key)
+        if existing and not existing.done():
+            return "A /btw is already running for this chat. Wait for it to finish."
+
+        if not hasattr(self, "_active_btw_tasks"):
+            self._active_btw_tasks: dict = {}
+
+        import uuid as _uuid
+        task_id = f"btw_{datetime.now().strftime('%H%M%S')}_{_uuid.uuid4().hex[:6]}"
+        _task = asyncio.create_task(self._run_btw_task(question, source, session_key, task_id))
+        self._background_tasks.add(_task)
+        self._active_btw_tasks[session_key] = _task
+
+        def _cleanup(task):
+            self._background_tasks.discard(task)
+            if self._active_btw_tasks.get(session_key) is task:
+                self._active_btw_tasks.pop(session_key, None)
+
+        _task.add_done_callback(_cleanup)
+
+        preview = question[:60] + ("..." if len(question) > 60 else "")
+        return f'💬 /btw: "{preview}"\nReply will appear here shortly.'
+
+    async def _run_btw_task(
+        self, question: str, source, session_key: str, task_id: str,
+    ) -> None:
+        """Execute an ephemeral /btw side question and deliver the answer."""
+        from run_agent import AIAgent
+
+        adapter = self.adapters.get(source.platform)
+        if not adapter:
+            logger.warning("No adapter for platform %s in /btw task %s", source.platform, task_id)
+            return
+
+        _thread_meta = {"thread_id": source.thread_id} if source.thread_id else None
+
+        try:
+            runtime_kwargs = _resolve_runtime_agent_kwargs()
+            if not runtime_kwargs.get("api_key"):
+                await adapter.send(
+                    source.chat_id,
+                    "❌ /btw failed: no provider credentials configured.",
+                    metadata=_thread_meta,
+                )
+                return
+
+            user_config = _load_gateway_config()
+            model = _resolve_gateway_model(user_config)
+            platform_key = _platform_config_key(source.platform)
+            reasoning_config = self._load_reasoning_config()
+            turn_route = self._resolve_turn_agent_config(question, model, runtime_kwargs)
+            pr = self._provider_routing
+
+            # Snapshot history from running agent or stored transcript
+            running_agent = self._running_agents.get(session_key)
+            if running_agent and running_agent is not _AGENT_PENDING_SENTINEL:
+                history_snapshot = list(getattr(running_agent, "_session_messages", []) or [])
+            else:
+                session_entry = self.session_store.get_or_create_session(source)
+                history_snapshot = self.session_store.load_transcript(session_entry.session_id)
+
+            btw_prompt = (
+                "[Ephemeral /btw side question. Answer using the conversation "
+                "context. No tools available. Be direct and concise.]\n\n"
+                + question
+            )
+
+            def run_sync():
+                agent = AIAgent(
+                    model=turn_route["model"],
+                    **turn_route["runtime"],
+                    max_iterations=8,
+                    quiet_mode=True,
+                    verbose_logging=False,
+                    enabled_toolsets=[],
+                    reasoning_config=reasoning_config,
+                    providers_allowed=pr.get("only"),
+                    providers_ignored=pr.get("ignore"),
+                    providers_order=pr.get("order"),
+                    provider_sort=pr.get("sort"),
+                    provider_require_parameters=pr.get("require_parameters", False),
+                    provider_data_collection=pr.get("data_collection"),
+                    session_id=task_id,
+                    platform=platform_key,
+                    session_db=None,
+                    fallback_model=self._fallback_model,
+                    skip_memory=True,
+                    skip_context_files=True,
+                    persist_session=False,
+                )
+                return agent.run_conversation(
+                    user_message=btw_prompt,
+                    conversation_history=history_snapshot,
+                    task_id=task_id,
+                    sync_honcho=False,
+                )
+
+            loop = asyncio.get_event_loop()
+            result = await loop.run_in_executor(None, run_sync)
+
+            response = (result.get("final_response") or "") if result else ""
+            if not response and result and result.get("error"):
+                response = f"Error: {result['error']}"
+            if not response:
+                response = "(No response generated)"
+
+            media_files, response = adapter.extract_media(response)
+            images, text_content = adapter.extract_images(response)
+            preview = question[:60] + ("..." if len(question) > 60 else "")
+            header = f'💬 /btw: "{preview}"\n\n'
+
+            if text_content:
+                await adapter.send(
+                    chat_id=source.chat_id,
+                    content=header + text_content,
+                    metadata=_thread_meta,
+                )
+            elif not images and not media_files:
+                await adapter.send(
+                    chat_id=source.chat_id,
+                    content=header + "(No response generated)",
+                    metadata=_thread_meta,
+                )
+
+            for image_url, alt_text in (images or []):
+                try:
+                    await adapter.send_image(chat_id=source.chat_id, image_url=image_url, caption=alt_text)
+                except Exception:
+                    pass
+
+            for media_path in (media_files or []):
+                try:
+                    await adapter.send_file(chat_id=source.chat_id, file_path=media_path)
+                except Exception:
+                    pass
+
+        except Exception as e:
+            logger.exception("/btw task %s failed", task_id)
+            try:
+                await adapter.send(
+                    chat_id=source.chat_id,
+                    content=f"❌ /btw failed: {e}",
+                    metadata=_thread_meta,
+                )
+            except Exception:
+                pass
+
     async def _handle_reasoning_command(self, event: MessageEvent) -> str:
         """Handle /reasoning command — manage reasoning effort and display toggle.
 
@@ -4686,8 +4781,8 @@ class GatewayRunner:
     async def _handle_update_command(self, event: MessageEvent) -> str:
         """Handle /update command — update Hermes Agent to the latest version.
 
-        Spawns ``hermes update`` in a separate systemd scope so it survives the
-        gateway restart that ``hermes update`` may trigger at the end. Marker
+        Spawns ``hermes update`` in a detached session (via ``setsid``) so it
+        survives the gateway restart that ``hermes update`` may trigger. Marker
         files are written so either the current gateway process or the next one
         can notify the user when the update finishes.
         """
@@ -4695,6 +4790,10 @@ class GatewayRunner:
         import shutil
         import subprocess
         from datetime import datetime
+        from hermes_cli.config import is_managed, format_managed_message
+
+        if is_managed():
+            return f"✗ {format_managed_message('update Hermes Agent')}"
 
         project_root = Path(__file__).parent.parent.resolve()
         git_dir = project_root / '.git'
@@ -4723,28 +4822,28 @@ class GatewayRunner:
         pending_path.write_text(json.dumps(pending))
         exit_code_path.unlink(missing_ok=True)
 
-        # Spawn `hermes update` in a separate cgroup so it survives gateway
-        # restart. systemd-run --user --scope creates a transient scope unit.
+        # Spawn `hermes update` detached so it survives gateway restart.
+        # Use setsid for portable session detach (works under system services
+        # where systemd-run --user fails due to missing D-Bus session).
         hermes_cmd_str = " ".join(shlex.quote(part) for part in hermes_cmd)
         update_cmd = (
             f"{hermes_cmd_str} update > {shlex.quote(str(output_path))} 2>&1; "
             f"status=$?; printf '%s' \"$status\" > {shlex.quote(str(exit_code_path))}"
         )
         try:
-            systemd_run = shutil.which("systemd-run")
-            if systemd_run:
+            setsid_bin = shutil.which("setsid")
+            if setsid_bin:
+                # Preferred: setsid creates a new session, fully detached
                 subprocess.Popen(
-                    [systemd_run, "--user", "--scope",
-                     "--unit=hermes-update", "--",
-                     "bash", "-c", update_cmd],
+                    [setsid_bin, "bash", "-c", update_cmd],
                     stdout=subprocess.DEVNULL,
                     stderr=subprocess.DEVNULL,
                     start_new_session=True,
                 )
             else:
-                # Fallback: best-effort detach with start_new_session
+                # Fallback: start_new_session=True calls os.setsid() in child
                 subprocess.Popen(
-                    ["bash", "-c", f"nohup {update_cmd} &"],
+                    ["bash", "-c", update_cmd],
                     stdout=subprocess.DEVNULL,
                     stderr=subprocess.DEVNULL,
                     start_new_session=True,
@@ -5735,7 +5834,9 @@ class GatewayRunner:
             # If so, update the session store entry so the NEXT message loads
             # the compressed transcript, not the stale pre-compression one.
             agent = agent_holder[0]
+            _session_was_split = False
             if agent and session_key and hasattr(agent, 'session_id') and agent.session_id != session_id:
+                _session_was_split = True
                 logger.info(
                     "Session split detected: %s → %s (compression)",
                     session_id, agent.session_id,
@@ -5747,6 +5848,13 @@ class GatewayRunner:
 
             effective_session_id = getattr(agent, 'session_id', session_id) if agent else session_id
 
+            # When compression created a new session, the messages list was
+            # shortened.  Using the original history offset would produce an
+            # empty new_messages slice, causing the gateway to write only a
+            # user/assistant pair — losing the compressed summary and tail.
+            # Reset to 0 so the gateway writes ALL compressed messages.
+            _effective_history_offset = 0 if _session_was_split else len(agent_history)
+
             # Auto-generate session title after first exchange (non-blocking)
             if final_response and self._session_db:
                 try:
@@ -5768,7 +5876,7 @@ class GatewayRunner:
                 "messages": result_holder[0].get("messages", []) if result_holder[0] else [],
                 "api_calls": result_holder[0].get("api_calls", 0) if result_holder[0] else 0,
                 "tools": tools_holder[0] or [],
-                "history_offset": len(agent_history),
+                "history_offset": _effective_history_offset,
                 "last_prompt_tokens": _last_prompt_toks,
                 "input_tokens": _input_toks,
                 "output_tokens": _output_toks,

hermes_cli/auth.py

@@ -160,7 +160,7 @@ PROVIDER_REGISTRY: Dict[str, ProviderConfig] = {
         id="alibaba",
         name="Alibaba Cloud (DashScope)",
         auth_type="api_key",
-        inference_base_url="https://coding-intl.dashscope.aliyuncs.com/v1",
+        inference_base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
         api_key_env_vars=("DASHSCOPE_API_KEY",),
         base_url_env_var="DASHSCOPE_BASE_URL",
     ),

hermes_cli/banner.py

@@ -432,10 +432,11 @@ def build_welcome_banner(console: Console, model: str, cwd: str,
     try:
         behind = get_update_result(timeout=0.5)
         if behind and behind > 0:
+            from hermes_cli.config import recommended_update_command
             commits_word = "commit" if behind == 1 else "commits"
             right_lines.append(
                 f"[bold yellow]⚠ {behind} {commits_word} behind[/]"
-                f"[dim yellow] — run [bold]hermes update[/bold] to update[/]"
+                f"[dim yellow] — run [bold]{recommended_update_command()}[/bold] to update[/]"
             )
     except Exception:
         pass  # Never break the banner over an update check

hermes_cli/claw.py

@@ -4,14 +4,19 @@ Usage:
     hermes claw migrate              # Interactive migration from ~/.openclaw
     hermes claw migrate --dry-run    # Preview what would be migrated
     hermes claw migrate --preset full --overwrite  # Full migration, overwrite conflicts
+    hermes claw cleanup              # Archive leftover OpenClaw directories
+    hermes claw cleanup --dry-run    # Preview what would be archived
 """
 
 import importlib.util
 import logging
+import shutil
 import sys
+from datetime import datetime
 from pathlib import Path
 
 from hermes_cli.config import get_hermes_home, get_config_path, load_config, save_config
+from hermes_constants import get_optional_skills_dir
 from hermes_cli.setup import (
     Colors,
     color,
@@ -19,6 +24,7 @@ from hermes_cli.setup import (
     print_info,
     print_success,
     print_error,
+    print_warning,
     prompt_yes_no,
 )
 
@@ -27,8 +33,7 @@ logger = logging.getLogger(__name__)
 PROJECT_ROOT = Path(__file__).parent.parent.resolve()
 
 _OPENCLAW_SCRIPT = (
-    PROJECT_ROOT
-    / "optional-skills"
+    get_optional_skills_dir(PROJECT_ROOT / "optional-skills")
     / "migration"
     / "openclaw-migration"
     / "scripts"
@@ -45,6 +50,18 @@ _OPENCLAW_SCRIPT_INSTALLED = (
     / "openclaw_to_hermes.py"
 )
 
+# Known OpenClaw directory names (current + legacy)
+_OPENCLAW_DIR_NAMES = (".openclaw", ".clawdbot", ".moldbot")
+
+# State files commonly found in OpenClaw workspace directories that cause
+# confusion after migration (the agent discovers them and writes to them)
+_WORKSPACE_STATE_GLOBS = (
+    "*/todo.json",
+    "*/sessions/*",
+    "*/memory/*.json",
+    "*/logs/*",
+)
+
 
 def _find_migration_script() -> Path | None:
     """Find the openclaw_to_hermes.py script in known locations."""
@@ -71,19 +88,88 @@ def _load_migration_module(script_path: Path):
     return mod
 
 
+def _find_openclaw_dirs() -> list[Path]:
+    """Find all OpenClaw directories on disk."""
+    found = []
+    for name in _OPENCLAW_DIR_NAMES:
+        candidate = Path.home() / name
+        if candidate.is_dir():
+            found.append(candidate)
+    return found
+
+
+def _scan_workspace_state(source_dir: Path) -> list[tuple[Path, str]]:
+    """Scan an OpenClaw directory for workspace state files that cause confusion.
+
+    Returns a list of (path, description) tuples.
+    """
+    findings: list[tuple[Path, str]] = []
+
+    # Direct state files in the root
+    for name in ("todo.json", "sessions", "logs"):
+        candidate = source_dir / name
+        if candidate.exists():
+            kind = "directory" if candidate.is_dir() else "file"
+            findings.append((candidate, f"Root {kind}: {name}"))
+
+    # State files inside workspace directories
+    for child in sorted(source_dir.iterdir()):
+        if not child.is_dir() or child.name.startswith("."):
+            continue
+        # Check for workspace-like subdirectories
+        for state_name in ("todo.json", "sessions", "logs", "memory"):
+            state_path = child / state_name
+            if state_path.exists():
+                kind = "directory" if state_path.is_dir() else "file"
+                rel = state_path.relative_to(source_dir)
+                findings.append((state_path, f"Workspace {kind}: {rel}"))
+
+    return findings
+
+
+def _archive_directory(source_dir: Path, dry_run: bool = False) -> Path:
+    """Rename an OpenClaw directory to .pre-migration.
+
+    Returns the archive path.
+    """
+    timestamp = datetime.now().strftime("%Y%m%d")
+    archive_name = f"{source_dir.name}.pre-migration"
+    archive_path = source_dir.parent / archive_name
+
+    # If archive already exists, add timestamp
+    if archive_path.exists():
+        archive_name = f"{source_dir.name}.pre-migration-{timestamp}"
+        archive_path = source_dir.parent / archive_name
+
+    # If still exists (multiple runs same day), add counter
+    counter = 2
+    while archive_path.exists():
+        archive_name = f"{source_dir.name}.pre-migration-{timestamp}-{counter}"
+        archive_path = source_dir.parent / archive_name
+        counter += 1
+
+    if not dry_run:
+        source_dir.rename(archive_path)
+
+    return archive_path
+
+
 def claw_command(args):
     """Route hermes claw subcommands."""
     action = getattr(args, "claw_action", None)
 
     if action == "migrate":
         _cmd_migrate(args)
+    elif action in ("cleanup", "clean"):
+        _cmd_cleanup(args)
     else:
-        print("Usage: hermes claw migrate [options]")
+        print("Usage: hermes claw <command> [options]")
         print()
         print("Commands:")
         print("  migrate          Migrate settings from OpenClaw to Hermes")
+        print("  cleanup          Archive leftover OpenClaw directories after migration")
         print()
-        print("Run 'hermes claw migrate --help' for migration options.")
+        print("Run 'hermes claw <command> --help' for options.")
 
 
 def _cmd_migrate(args):
@@ -210,6 +296,168 @@ def _cmd_migrate(args):
     # Print results
     _print_migration_report(report, dry_run)
 
+    # After successful non-dry-run migration, offer to archive the source directory
+    if not dry_run and report.get("summary", {}).get("migrated", 0) > 0:
+        _offer_source_archival(source_dir, getattr(args, "yes", False))
+
+
+def _offer_source_archival(source_dir: Path, auto_yes: bool = False):
+    """After migration, offer to rename the source directory to prevent state fragmentation.
+
+    OpenClaw workspace directories contain state files (todo.json, sessions, etc.)
+    that the agent may discover and write to, causing confusion. Renaming the
+    directory prevents this.
+    """
+    if not source_dir.is_dir():
+        return
+
+    # Scan for state files that could cause problems
+    state_files = _scan_workspace_state(source_dir)
+
+    print()
+    print_header("Post-Migration Cleanup")
+    print_info("The OpenClaw directory still exists and contains workspace state files")
+    print_info("that can confuse the agent (todo lists, sessions, logs).")
+    if state_files:
+        print()
+        print(color("  Found state files:", Colors.YELLOW))
+        # Show up to 10 most relevant findings
+        for path, desc in state_files[:10]:
+            print(f"      {desc}")
+        if len(state_files) > 10:
+            print(f"      ... and {len(state_files) - 10} more")
+    print()
+    print_info(f"Recommend: rename {source_dir.name}/ to {source_dir.name}.pre-migration/")
+    print_info("This prevents the agent from discovering old workspace directories.")
+    print_info("You can always rename it back if needed.")
+    print()
+
+    if auto_yes or prompt_yes_no(f"Archive {source_dir} now?", default=True):
+        try:
+            archive_path = _archive_directory(source_dir)
+            print_success(f"Archived: {source_dir} → {archive_path}")
+            print_info("The original directory has been renamed, not deleted.")
+            print_info(f"To undo: mv {archive_path} {source_dir}")
+        except OSError as e:
+            print_error(f"Could not archive: {e}")
+            print_info(f"You can do it manually: mv {source_dir} {source_dir}.pre-migration")
+    else:
+        print_info("Skipped. You can archive later with: hermes claw cleanup")
+
+
+def _cmd_cleanup(args):
+    """Archive leftover OpenClaw directories after migration.
+
+    Scans for OpenClaw directories that still exist after migration and offers
+    to rename them to .pre-migration to prevent state fragmentation.
+    """
+    dry_run = getattr(args, "dry_run", False)
+    auto_yes = getattr(args, "yes", False)
+    explicit_source = getattr(args, "source", None)
+
+    print()
+    print(
+        color(
+            "┌─────────────────────────────────────────────────────────┐",
+            Colors.MAGENTA,
+        )
+    )
+    print(
+        color(
+            "│          ⚕ Hermes — OpenClaw Cleanup                   │",
+            Colors.MAGENTA,
+        )
+    )
+    print(
+        color(
+            "└─────────────────────────────────────────────────────────┘",
+            Colors.MAGENTA,
+        )
+    )
+
+    # Find OpenClaw directories
+    if explicit_source:
+        dirs_to_check = [Path(explicit_source)]
+    else:
+        dirs_to_check = _find_openclaw_dirs()
+
+    if not dirs_to_check:
+        print()
+        print_success("No OpenClaw directories found. Nothing to clean up.")
+        return
+
+    total_archived = 0
+
+    for source_dir in dirs_to_check:
+        print()
+        print_header(f"Found: {source_dir}")
+
+        # Scan for state files
+        state_files = _scan_workspace_state(source_dir)
+
+        # Show directory stats
+        try:
+            workspace_dirs = [
+                d for d in source_dir.iterdir()
+                if d.is_dir() and not d.name.startswith(".")
+                and any((d / name).exists() for name in ("todo.json", "SOUL.md", "MEMORY.md", "USER.md"))
+            ]
+        except OSError:
+            workspace_dirs = []
+
+        if workspace_dirs:
+            print_info(f"Workspace directories: {len(workspace_dirs)}")
+            for ws in workspace_dirs[:5]:
+                items = []
+                if (ws / "todo.json").exists():
+                    items.append("todo.json")
+                if (ws / "sessions").is_dir():
+                    items.append("sessions/")
+                if (ws / "SOUL.md").exists():
+                    items.append("SOUL.md")
+                if (ws / "MEMORY.md").exists():
+                    items.append("MEMORY.md")
+                detail = ", ".join(items) if items else "empty"
+                print(f"      {ws.name}/  ({detail})")
+            if len(workspace_dirs) > 5:
+                print(f"      ... and {len(workspace_dirs) - 5} more")
+
+        if state_files:
+            print()
+            print(color(f"  {len(state_files)} state file(s) that could cause confusion:", Colors.YELLOW))
+            for path, desc in state_files[:8]:
+                print(f"      {desc}")
+            if len(state_files) > 8:
+                print(f"      ... and {len(state_files) - 8} more")
+
+        print()
+
+        if dry_run:
+            archive_path = _archive_directory(source_dir, dry_run=True)
+            print_info(f"Would archive: {source_dir} → {archive_path}")
+        else:
+            if auto_yes or prompt_yes_no(f"Archive {source_dir}?", default=True):
+                try:
+                    archive_path = _archive_directory(source_dir)
+                    print_success(f"Archived: {source_dir} → {archive_path}")
+                    total_archived += 1
+                except OSError as e:
+                    print_error(f"Could not archive: {e}")
+                    print_info(f"Try manually: mv {source_dir} {source_dir}.pre-migration")
+            else:
+                print_info("Skipped.")
+
+    # Summary
+    print()
+    if dry_run:
+        print_info(f"Dry run complete. {len(dirs_to_check)} directory(ies) would be archived.")
+        print_info("Run without --dry-run to archive them.")
+    elif total_archived:
+        print_success(f"Cleaned up {total_archived} OpenClaw directory(ies).")
+        print_info("Directories were renamed, not deleted. You can undo by renaming them back.")
+    else:
+        print_info("No directories were archived.")
+
 
 def _print_migration_report(report: dict, dry_run: bool):
     """Print a formatted migration report."""

hermes_cli/commands.py

@@ -67,6 +67,8 @@ COMMAND_REGISTRY: list[CommandDef] = [
                gateway_only=True),
     CommandDef("background", "Run a prompt in the background", "Session",
                aliases=("bg",), args_hint="<prompt>"),
+    CommandDef("btw", "Ephemeral side question using session context (no tools, not persisted)", "Session",
+               args_hint="<question>"),
     CommandDef("queue", "Queue a prompt for the next turn (doesn't interrupt)", "Session",
                aliases=("q",), args_hint="<prompt>"),
     CommandDef("status", "Show session info", "Session",

hermes_cli/config.py

@@ -52,26 +52,86 @@ from hermes_cli.default_soul import DEFAULT_SOUL_MD
 # Managed mode (NixOS declarative config)
 # =============================================================================
 
+_MANAGED_TRUE_VALUES = ("true", "1", "yes")
+_MANAGED_SYSTEM_NAMES = {
+    "brew": "Homebrew",
+    "homebrew": "Homebrew",
+    "nix": "NixOS",
+    "nixos": "NixOS",
+}
+
+
+def get_managed_system() -> Optional[str]:
+    """Return the package manager owning this install, if any."""
+    raw = os.getenv("HERMES_MANAGED", "").strip()
+    if raw:
+        normalized = raw.lower()
+        if normalized in _MANAGED_TRUE_VALUES:
+            return "NixOS"
+        return _MANAGED_SYSTEM_NAMES.get(normalized, raw)
+
+    managed_marker = get_hermes_home() / ".managed"
+    if managed_marker.exists():
+        return "NixOS"
+    return None
+
+
 def is_managed() -> bool:
-    """Check if hermes is running in Nix-managed mode.
+    """Check if Hermes is running in package-manager-managed mode.
 
     Two signals: the HERMES_MANAGED env var (set by the systemd service),
     or a .managed marker file in HERMES_HOME (set by the NixOS activation
     script, so interactive shells also see it).
     """
-    if os.getenv("HERMES_MANAGED", "").lower() in ("true", "1", "yes"):
-        return True
-    managed_marker = get_hermes_home() / ".managed"
-    return managed_marker.exists()
+    return get_managed_system() is not None
+
+
+def get_managed_update_command() -> Optional[str]:
+    """Return the preferred upgrade command for a managed install."""
+    managed_system = get_managed_system()
+    if managed_system == "Homebrew":
+        return "brew upgrade hermes-agent"
+    if managed_system == "NixOS":
+        return "sudo nixos-rebuild switch"
+    return None
+
+
+def recommended_update_command() -> str:
+    """Return the best update command for the current installation."""
+    return get_managed_update_command() or "hermes update"
+
+
+def format_managed_message(action: str = "modify this Hermes installation") -> str:
+    """Build a user-facing error for managed installs."""
+    managed_system = get_managed_system() or "a package manager"
+    raw = os.getenv("HERMES_MANAGED", "").strip().lower()
+
+    if managed_system == "NixOS":
+        env_hint = "true" if raw in _MANAGED_TRUE_VALUES else raw or "true"
+        return (
+            f"Cannot {action}: this Hermes installation is managed by NixOS "
+            f"(HERMES_MANAGED={env_hint}).\n"
+            "Edit services.hermes-agent.settings in your configuration.nix and run:\n"
+            "  sudo nixos-rebuild switch"
+        )
+
+    if managed_system == "Homebrew":
+        env_hint = raw or "homebrew"
+        return (
+            f"Cannot {action}: this Hermes installation is managed by Homebrew "
+            f"(HERMES_MANAGED={env_hint}).\n"
+            "Use:\n"
+            "  brew upgrade hermes-agent"
+        )
+
+    return (
+        f"Cannot {action}: this Hermes installation is managed by {managed_system}.\n"
+        "Use your package manager to upgrade or reinstall Hermes."
+    )
 
 def managed_error(action: str = "modify configuration"):
     """Print user-friendly error for managed mode."""
-    print(
-        f"Cannot {action}: configuration is managed by NixOS (HERMES_MANAGED=true).\n"
-        "Edit services.hermes-agent.settings in your configuration.nix and run:\n"
-        "  sudo nixos-rebuild switch",
-        file=sys.stderr,
-    )
+    print(format_managed_message(action), file=sys.stderr)
 
 
 # =============================================================================

hermes_cli/cron.py

@@ -56,7 +56,7 @@ def cron_list(show_all: bool = False):
     print()
 
     for job in jobs:
-        job_id = job.get("id", "?")[:8]
+        job_id = job.get("id", "?")
         name = job.get("name", "(unnamed)")
         schedule = job.get("schedule_display", job.get("schedule", {}).get("value", "?"))
         state = job.get("state", "scheduled" if job.get("enabled", True) else "paused")

hermes_cli/main.py

@@ -2467,10 +2467,14 @@ def cmd_version(args):
     # Show update status (synchronous — acceptable since user asked for version info)
     try:
         from hermes_cli.banner import check_for_updates
+        from hermes_cli.config import recommended_update_command
         behind = check_for_updates()
         if behind and behind > 0:
             commits_word = "commit" if behind == 1 else "commits"
-            print(f"Update available: {behind} {commits_word} behind — run 'hermes update'")
+            print(
+                f"Update available: {behind} {commits_word} behind — "
+                f"run '{recommended_update_command()}'"
+            )
         elif behind == 0:
             print("Up to date")
     except Exception:
@@ -2821,6 +2825,11 @@ def _invalidate_update_cache():
 def cmd_update(args):
     """Update Hermes Agent to the latest version."""
     import shutil
+    from hermes_cli.config import is_managed, managed_error
+
+    if is_managed():
+        managed_error("update Hermes Agent")
+        return
     
     print("⚕ Updating Hermes Agent...")
     print()
@@ -3156,6 +3165,7 @@ def cmd_update(args):
             _gw_service_name = get_service_name()
             existing_pid = get_running_pid()
             has_systemd_service = False
+            has_system_service = False
             has_launchd_service = False
 
             try:
@@ -3168,6 +3178,19 @@ def cmd_update(args):
             except (FileNotFoundError, subprocess.TimeoutExpired):
                 pass
 
+            # Also check for a system-level service (hermes gateway install --system).
+            # This covers gateways running under system systemd where --user
+            # fails due to missing D-Bus session.
+            if not has_systemd_service and is_linux():
+                try:
+                    check = subprocess.run(
+                        ["systemctl", "is-active", _gw_service_name],
+                        capture_output=True, text=True, timeout=5,
+                    )
+                    has_system_service = check.stdout.strip() == "active"
+                except (FileNotFoundError, subprocess.TimeoutExpired):
+                    pass
+
             # Check for macOS launchd service
             if is_macos():
                 try:
@@ -3182,7 +3205,7 @@ def cmd_update(args):
                 except (FileNotFoundError, subprocess.TimeoutExpired):
                     pass
 
-            if existing_pid or has_systemd_service or has_launchd_service:
+            if existing_pid or has_systemd_service or has_system_service or has_launchd_service:
                 print()
 
                 # When a service manager is handling the gateway, let it
@@ -3223,6 +3246,21 @@ def cmd_update(args):
                                 print("    hermes gateway restart")
                             else:
                                 print("  Try manually: hermes gateway restart")
+                elif has_system_service:
+                    # System-level service (hermes gateway install --system).
+                    # No D-Bus session needed — systemctl without --user talks
+                    # directly to the system manager over /run/systemd/private.
+                    print("→ Restarting system gateway service...")
+                    restart = subprocess.run(
+                        ["systemctl", "restart", _gw_service_name],
+                        capture_output=True, text=True, timeout=15,
+                    )
+                    if restart.returncode == 0:
+                        print("✓ Gateway restarted (system service).")
+                    else:
+                        print(f"⚠ Gateway restart failed: {restart.stderr.strip()}")
+                        print("  System services may require root.  Try:")
+                        print(f"    sudo systemctl restart {_gw_service_name}")
                 elif has_launchd_service:
                     # Refresh the plist first (picks up --replace and other
                     # changes from the update we just pulled).
@@ -4703,6 +4741,28 @@ For more help on a command:
         help="Skip confirmation prompts"
     )
 
+    # claw cleanup
+    claw_cleanup = claw_subparsers.add_parser(
+        "cleanup",
+        aliases=["clean"],
+        help="Archive leftover OpenClaw directories after migration",
+        description="Scan for and archive leftover OpenClaw directories to prevent state fragmentation"
+    )
+    claw_cleanup.add_argument(
+        "--source",
+        help="Path to a specific OpenClaw directory to clean up"
+    )
+    claw_cleanup.add_argument(
+        "--dry-run",
+        action="store_true",
+        help="Preview what would be archived without making changes"
+    )
+    claw_cleanup.add_argument(
+        "--yes", "-y",
+        action="store_true",
+        help="Skip confirmation prompts"
+    )
+
     def cmd_claw(args):
         from hermes_cli.claw import claw_command
         claw_command(args)

hermes_cli/models.py

@@ -27,6 +27,7 @@ GITHUB_MODELS_CATALOG_URL = COPILOT_MODELS_URL
 # (model_id, display description shown in menus)
 OPENROUTER_MODELS: list[tuple[str, str]] = [
     ("anthropic/claude-opus-4.6",       "recommended"),
+    ("anthropic/claude-sonnet-4.6",     ""),
     ("anthropic/claude-sonnet-4.5",     ""),
     ("anthropic/claude-haiku-4.5",      ""),
     ("openai/gpt-5.4",                  ""),
@@ -56,6 +57,7 @@ OPENROUTER_MODELS: list[tuple[str, str]] = [
 _PROVIDER_MODELS: dict[str, list[str]] = {
     "nous": [
         "anthropic/claude-opus-4.6",
+        "anthropic/claude-sonnet-4.6",
         "anthropic/claude-sonnet-4.5",
         "anthropic/claude-haiku-4.5",
         "openai/gpt-5.4",

hermes_cli/plugins_cmd.py

@@ -265,10 +265,11 @@ def cmd_install(identifier: str, force: bool = False) -> None:
                 )
                 sys.exit(1)
             if mv_int > _SUPPORTED_MANIFEST_VERSION:
+                from hermes_cli.config import recommended_update_command
                 console.print(
                     f"[red]Error:[/red] Plugin '{plugin_name}' requires manifest_version "
                     f"{mv}, but this installer only supports up to {_SUPPORTED_MANIFEST_VERSION}.\n"
-                    f"Run [bold]hermes update[/bold] to get a newer installer."
+                    f"Run [bold]{recommended_update_command()}[/bold] to get a newer installer."
                 )
                 sys.exit(1)
 

hermes_cli/profiles.py

@@ -241,7 +241,7 @@ def _read_config_model(profile_dir: Path) -> tuple:
         if isinstance(model_cfg, str):
             return model_cfg, None
         if isinstance(model_cfg, dict):
-            return model_cfg.get("model"), model_cfg.get("provider")
+            return model_cfg.get("default") or model_cfg.get("model"), model_cfg.get("provider")
         return None, None
     except Exception:
         return None, None

hermes_cli/setup.py

@@ -18,6 +18,8 @@ import sys
 from pathlib import Path
 from typing import Optional, Dict, Any
 
+from hermes_constants import get_optional_skills_dir
+
 logger = logging.getLogger(__name__)
 
 PROJECT_ROOT = Path(__file__).parent.parent.resolve()
@@ -3121,8 +3123,7 @@ def _skip_configured_section(
 
 
 _OPENCLAW_SCRIPT = (
-    PROJECT_ROOT
-    / "optional-skills"
+    get_optional_skills_dir(PROJECT_ROOT / "optional-skills")
     / "migration"
     / "openclaw-migration"
     / "scripts"

hermes_constants.py

@@ -17,6 +17,20 @@ def get_hermes_home() -> Path:
     return Path(os.getenv("HERMES_HOME", Path.home() / ".hermes"))
 
 
+def get_optional_skills_dir(default: Path | None = None) -> Path:
+    """Return the optional-skills directory, honoring package-manager wrappers.
+
+    Packaged installs may ship ``optional-skills`` outside the Python package
+    tree and expose it via ``HERMES_OPTIONAL_SKILLS``.
+    """
+    override = os.getenv("HERMES_OPTIONAL_SKILLS", "").strip()
+    if override:
+        return Path(override)
+    if default is not None:
+        return default
+    return get_hermes_home() / "optional-skills"
+
+
 def get_hermes_dir(new_subpath: str, old_name: str) -> Path:
     """Resolve a Hermes subdirectory with backward compatibility.
 

model_tools.py

@@ -252,7 +252,7 @@ def get_tool_definitions(
     # Determine which tool names the caller wants
     tools_to_include: set = set()
 
-    if enabled_toolsets:
+    if enabled_toolsets is not None:
         for toolset_name in enabled_toolsets:
             if validate_toolset(toolset_name):
                 resolved = resolve_toolset(toolset_name)

optional-skills/migration/openclaw-migration/scripts/openclaw_to_hermes.py

@@ -2455,9 +2455,24 @@ class Migrator:
             notes.append("")
 
         notes.extend([
+            "## IMPORTANT: Archive the OpenClaw Directory",
+            "",
+            "After migration, your OpenClaw directory still exists on disk with workspace",
+            "state files (todo.json, sessions, logs). If the Hermes agent discovers these",
+            "directories, it may read/write to them instead of the Hermes state, causing",
+            "confusion (e.g., cron jobs reading a different todo list than interactive sessions).",
+            "",
+            "**Strongly recommended:** Run `hermes claw cleanup` to rename the OpenClaw",
+            "directory to `.openclaw.pre-migration`. This prevents the agent from finding it.",
+            "The directory is renamed, not deleted — you can undo this at any time.",
+            "",
+            "If you skip this step and notice the agent getting confused about workspaces",
+            "or todo lists, run `hermes claw cleanup` to fix it.",
+            "",
             "## Hermes-Specific Setup",
             "",
             "After migration, you may want to:",
+            "- Run `hermes claw cleanup` to archive the OpenClaw directory (prevents state confusion)",
             "- Run `hermes setup` to configure any remaining settings",
             "- Run `hermes mcp list` to verify MCP servers were imported correctly",
             "- Run `hermes cron` to recreate scheduled tasks (see archive/cron-config.json)",

packaging/homebrew/README.md

@@ -0,0 +1,14 @@
+Homebrew packaging notes for Hermes Agent.
+
+Use `packaging/homebrew/hermes-agent.rb` as a tap or `homebrew-core` starting point.
+
+Key choices:
+- Stable builds should target the semver-named sdist asset attached to each GitHub release, not the CalVer tag tarball.
+- `faster-whisper` now lives in the `voice` extra, which keeps wheel-only transitive dependencies out of the base Homebrew formula.
+- The wrapper exports `HERMES_BUNDLED_SKILLS`, `HERMES_OPTIONAL_SKILLS`, and `HERMES_MANAGED=homebrew` so packaged installs keep runtime assets and defer upgrades to Homebrew.
+
+Typical update flow:
+1. Bump the formula `url`, `version`, and `sha256`.
+2. Refresh Python resources with `brew update-python-resources --print-only hermes-agent`.
+3. Keep `ignore_packages: %w[certifi cryptography pydantic]`.
+4. Verify `brew audit --new --strict hermes-agent` and `brew test hermes-agent`.

packaging/homebrew/hermes-agent.rb

@@ -0,0 +1,48 @@
+class HermesAgent < Formula
+  include Language::Python::Virtualenv
+
+  desc "Self-improving AI agent that creates skills from experience"
+  homepage "https://hermes-agent.nousresearch.com"
+  # Stable source should point at the semver-named sdist asset attached by
+  # scripts/release.py, not the CalVer tag tarball.
+  url "https://github.com/NousResearch/hermes-agent/releases/download/v2026.3.30/hermes_agent-0.6.0.tar.gz"
+  sha256 "<replace-with-release-asset-sha256>"
+  license "MIT"
+
+  depends_on "certifi" => :no_linkage
+  depends_on "cryptography" => :no_linkage
+  depends_on "libyaml"
+  depends_on "python@3.14"
+
+  pypi_packages ignore_packages: %w[certifi cryptography pydantic]
+
+  # Refresh resource stanzas after bumping the source url/version:
+  #   brew update-python-resources --print-only hermes-agent
+
+  def install
+    venv = virtualenv_create(libexec, "python3.14")
+    venv.pip_install resources
+    venv.pip_install buildpath
+
+    pkgshare.install "skills", "optional-skills"
+
+    %w[hermes hermes-agent hermes-acp].each do |exe|
+      next unless (libexec/"bin"/exe).exist?
+
+      (bin/exe).write_env_script(
+        libexec/"bin"/exe,
+        HERMES_BUNDLED_SKILLS: pkgshare/"skills",
+        HERMES_OPTIONAL_SKILLS: pkgshare/"optional-skills",
+        HERMES_MANAGED: "homebrew"
+      )
+    end
+  end
+
+  test do
+    assert_match "Hermes Agent v#{version}", shell_output("#{bin}/hermes version")
+
+    managed = shell_output("#{bin}/hermes update 2>&1")
+    assert_match "managed by Homebrew", managed
+    assert_match "brew upgrade hermes-agent", managed
+  end
+end

pyproject.toml

@@ -32,7 +32,6 @@ dependencies = [
   "fal-client>=0.13.1,<1",
   # Text-to-speech (Edge TTS is free, no API key needed)
   "edge-tts>=7.2.7,<8",
-  "faster-whisper>=1.0.0,<2",
   # Skills Hub (GitHub App JWT auth — optional, only needed for bot identity)
   "PyJWT[crypto]>=2.12.0,<3",  # CVE-2026-32597
 ]
@@ -47,7 +46,13 @@ slack = ["slack-bolt>=1.18.0,<2", "slack-sdk>=3.27.0,<4"]
 matrix = ["matrix-nio[e2e]>=0.24.0,<1"]
 cli = ["simple-term-menu>=1.0,<2"]
 tts-premium = ["elevenlabs>=1.0,<2"]
-voice = ["sounddevice>=0.4.6,<1", "numpy>=1.24.0,<3"]
+voice = [
+  # Local STT pulls in wheel-only transitive deps (ctranslate2, onnxruntime),
+  # so keep it out of the base install for source-build packagers like Homebrew.
+  "faster-whisper>=1.0.0,<2",
+  "sounddevice>=0.4.6,<1",
+  "numpy>=1.24.0,<3",
+]
 pty = [
   "ptyprocess>=0.7.0,<1; sys_platform != 'win32'",
   "pywinpty>=2.0.0,<3; sys_platform == 'win32'",

run_agent.py

@@ -508,6 +508,7 @@ class AIAgent:
         checkpoints_enabled: bool = False,
         checkpoint_max_snapshots: int = 50,
         pass_session_id: bool = False,
+        persist_session: bool = True,
     ):
         """
         Initialize the AI Agent.
@@ -573,6 +574,7 @@ class AIAgent:
         self.background_review_callback = None  # Optional sync callback for gateway delivery
         self.skip_context_files = skip_context_files
         self.pass_session_id = pass_session_id
+        self.persist_session = persist_session
         self.log_prefix_chars = log_prefix_chars
         self.log_prefix = f"{log_prefix} " if log_prefix else ""
         # Store effective base URL for feature detection (prompt caching, reasoning, etc.)
@@ -1700,7 +1702,10 @@ class AIAgent:
         """Save session state to both JSON log and SQLite on any exit path.
 
         Ensures conversations are never lost, even on errors or early returns.
+        Skipped when ``persist_session=False`` (ephemeral helper flows).
         """
+        if not self.persist_session:
+            return
         self._apply_persist_user_message_override(messages)
         self._session_messages = messages
         self._save_session_log(messages)
@@ -6250,6 +6255,12 @@ class AIAgent:
                     )
                     if len(messages) >= _orig_len:
                         break  # Cannot compress further
+                    # Compression created a new session — clear the history
+                    # reference so _flush_messages_to_session_db writes ALL
+                    # compressed messages to the new session's SQLite, not
+                    # skipping them because conversation_history is still the
+                    # pre-compression length.
+                    conversation_history = None
                     # Re-estimate after compression
                     _preflight_tokens = estimate_request_tokens_rough(
                         messages,
@@ -7050,6 +7061,7 @@ class AIAgent:
                         compression_attempts += 1
                         if compression_attempts > max_compression_attempts:
                             self._vprint(f"{self.log_prefix}❌ Max compression attempts ({max_compression_attempts}) reached for payload-too-large error.", force=True)
+                            self._vprint(f"{self.log_prefix}   💡 Try /new to start a fresh conversation, or /compress to retry compression.", force=True)
                             logging.error(f"{self.log_prefix}413 compression failed after {max_compression_attempts} attempts.")
                             self._persist_session(messages, conversation_history)
                             return {
@@ -7074,6 +7086,7 @@ class AIAgent:
                             break
                         else:
                             self._vprint(f"{self.log_prefix}❌ Payload too large and cannot compress further.", force=True)
+                            self._vprint(f"{self.log_prefix}   💡 Try /new to start a fresh conversation, or /compress to retry compression.", force=True)
                             logging.error(f"{self.log_prefix}413 payload too large. Cannot compress further.")
                             self._persist_session(messages, conversation_history)
                             return {
@@ -7150,6 +7163,7 @@ class AIAgent:
                         compression_attempts += 1
                         if compression_attempts > max_compression_attempts:
                             self._vprint(f"{self.log_prefix}❌ Max compression attempts ({max_compression_attempts}) reached.", force=True)
+                            self._vprint(f"{self.log_prefix}   💡 Try /new to start a fresh conversation, or /compress to retry compression.", force=True)
                             logging.error(f"{self.log_prefix}Context compression failed after {max_compression_attempts} attempts.")
                             self._persist_session(messages, conversation_history)
                             return {
@@ -7176,7 +7190,7 @@ class AIAgent:
                         else:
                             # Can't compress further and already at minimum tier
                             self._vprint(f"{self.log_prefix}❌ Context length exceeded and cannot compress further.", force=True)
-                            self._vprint(f"{self.log_prefix}   💡 The conversation has accumulated too much content.", force=True)
+                            self._vprint(f"{self.log_prefix}   💡 The conversation has accumulated too much content. Try /new to start fresh, or /compress to manually trigger compression.", force=True)
                             logging.error(f"{self.log_prefix}Context length exceeded: {approx_tokens:,} tokens. Cannot compress further.")
                             self._persist_session(messages, conversation_history)
                             return {
@@ -7765,6 +7779,10 @@ class AIAgent:
                             approx_tokens=self.context_compressor.last_prompt_tokens,
                             task_id=effective_task_id,
                         )
+                        # Compression created a new session — clear history so
+                        # _flush_messages_to_session_db writes compressed messages
+                        # to the new session (see preflight compression comment).
+                        conversation_history = None
                     
                     # Save session log incrementally (so progress is visible even if interrupted)
                     self._session_messages = messages

scripts/release.py

@@ -24,6 +24,7 @@ import argparse
 import json
 import os
 import re
+import shutil
 import subprocess
 import sys
 from collections import defaultdict
@@ -128,6 +129,16 @@ def git(*args, cwd=None):
     return result.stdout.strip()
 
 
+def git_result(*args, cwd=None):
+    """Run a git command and return the full CompletedProcess."""
+    return subprocess.run(
+        ["git"] + list(args),
+        capture_output=True,
+        text=True,
+        cwd=cwd or str(REPO_ROOT),
+    )
+
+
 def get_last_tag():
     """Get the most recent CalVer tag."""
     tags = git("tag", "--list", "v20*", "--sort=-v:refname")
@@ -136,6 +147,18 @@ def get_last_tag():
     return None
 
 
+def next_available_tag(base_tag: str) -> tuple[str, str]:
+    """Return a tag/calver pair, suffixing same-day releases when needed."""
+    if not git("tag", "--list", base_tag):
+        return base_tag, base_tag.removeprefix("v")
+
+    suffix = 2
+    while git("tag", "--list", f"{base_tag}.{suffix}"):
+        suffix += 1
+    tag_name = f"{base_tag}.{suffix}"
+    return tag_name, tag_name.removeprefix("v")
+
+
 def get_current_version():
     """Read current semver from __init__.py."""
     content = VERSION_FILE.read_text()
@@ -192,6 +215,41 @@ def update_version_files(semver: str, calver_date: str):
     PYPROJECT_FILE.write_text(pyproject)
 
 
+def build_release_artifacts(semver: str) -> list[Path]:
+    """Build sdist/wheel artifacts for the current release.
+
+    Returns the artifact paths when the local environment has ``python -m build``
+    available. If build tooling is missing or the build fails, returns an empty
+    list and lets the release proceed without attached Python artifacts.
+    """
+    dist_dir = REPO_ROOT / "dist"
+    shutil.rmtree(dist_dir, ignore_errors=True)
+
+    result = subprocess.run(
+        [sys.executable, "-m", "build", "--sdist", "--wheel"],
+        cwd=str(REPO_ROOT),
+        capture_output=True,
+        text=True,
+    )
+    if result.returncode != 0:
+        print("  ⚠ Could not build Python release artifacts.")
+        stderr = result.stderr.strip()
+        stdout = result.stdout.strip()
+        if stderr:
+            print(f"    {stderr.splitlines()[-1]}")
+        elif stdout:
+            print(f"    {stdout.splitlines()[-1]}")
+        print("    Install the 'build' package to attach semver-named sdist/wheel assets.")
+        return []
+
+    artifacts = sorted(p for p in dist_dir.iterdir() if p.is_file())
+    matching = [p for p in artifacts if semver in p.name]
+    if not matching:
+        print("  ⚠ Built artifacts did not match the expected release version.")
+        return []
+    return matching
+
+
 def resolve_author(name: str, email: str) -> str:
     """Resolve a git author to a GitHub @mention."""
     # Try email lookup first
@@ -424,18 +482,10 @@ def main():
         now = datetime.now()
         calver_date = f"{now.year}.{now.month}.{now.day}"
 
-    tag_name = f"v{calver_date}"
-
-    # Check for existing tag with same date
-    existing = git("tag", "--list", tag_name)
-    if existing and not args.publish:
-        # Append a suffix for same-day releases
-        suffix = 2
-        while git("tag", "--list", f"{tag_name}.{suffix}"):
-            suffix += 1
-        tag_name = f"{tag_name}.{suffix}"
-        calver_date = f"{calver_date}.{suffix}"
-        print(f"Note: Tag {tag_name[:-2]} already exists, using {tag_name}")
+    base_tag = f"v{calver_date}"
+    tag_name, calver_date = next_available_tag(base_tag)
+    if tag_name != base_tag:
+        print(f"Note: Tag {base_tag} already exists, using {tag_name}")
 
     # Determine semver
     current_version = get_current_version()
@@ -494,41 +544,83 @@ def main():
             print(f"  ✓ Updated version files to v{new_version} ({calver_date})")
 
             # Commit version bump
-            git("add", str(VERSION_FILE), str(PYPROJECT_FILE))
-            git("commit", "-m", f"chore: bump version to v{new_version} ({calver_date})")
+            add_result = git_result("add", str(VERSION_FILE), str(PYPROJECT_FILE))
+            if add_result.returncode != 0:
+                print(f"  ✗ Failed to stage version files: {add_result.stderr.strip()}")
+                return
+
+            commit_result = git_result(
+                "commit", "-m", f"chore: bump version to v{new_version} ({calver_date})"
+            )
+            if commit_result.returncode != 0:
+                print(f"  ✗ Failed to commit version bump: {commit_result.stderr.strip()}")
+                return
             print(f"  ✓ Committed version bump")
 
         # Create annotated tag
-        git("tag", "-a", tag_name, "-m",
-            f"Hermes Agent v{new_version} ({calver_date})\n\nWeekly release")
+        tag_result = git_result(
+            "tag", "-a", tag_name, "-m",
+            f"Hermes Agent v{new_version} ({calver_date})\n\nWeekly release"
+        )
+        if tag_result.returncode != 0:
+            print(f"  ✗ Failed to create tag {tag_name}: {tag_result.stderr.strip()}")
+            return
         print(f"  ✓ Created tag {tag_name}")
 
         # Push
-        push_result = git("push", "origin", "HEAD", "--tags")
-        print(f"  ✓ Pushed to origin")
+        push_result = git_result("push", "origin", "HEAD", "--tags")
+        if push_result.returncode == 0:
+            print(f"  ✓ Pushed to origin")
+        else:
+            print(f"  ✗ Failed to push to origin: {push_result.stderr.strip()}")
+            print("    Continue manually after fixing access:")
+            print("    git push origin HEAD --tags")
+
+        # Build semver-named Python artifacts so downstream packagers
+        # (e.g. Homebrew) can target them without relying on CalVer tag names.
+        artifacts = build_release_artifacts(new_version)
+        if artifacts:
+            print("  ✓ Built release artifacts:")
+            for artifact in artifacts:
+                print(f"    - {artifact.relative_to(REPO_ROOT)}")
 
         # Create GitHub release
         changelog_file = REPO_ROOT / ".release_notes.md"
         changelog_file.write_text(changelog)
 
-        result = subprocess.run(
-            ["gh", "release", "create", tag_name,
-             "--title", f"Hermes Agent v{new_version} ({calver_date})",
-             "--notes-file", str(changelog_file)],
-            capture_output=True, text=True,
-            cwd=str(REPO_ROOT),
-        )
+        gh_cmd = [
+            "gh", "release", "create", tag_name,
+            "--title", f"Hermes Agent v{new_version} ({calver_date})",
+            "--notes-file", str(changelog_file),
+        ]
+        gh_cmd.extend(str(path) for path in artifacts)
 
-        changelog_file.unlink(missing_ok=True)
-
-        if result.returncode == 0:
-            print(f"  ✓ GitHub release created: {result.stdout.strip()}")
+        gh_bin = shutil.which("gh")
+        if gh_bin:
+            result = subprocess.run(
+                gh_cmd,
+                capture_output=True, text=True,
+                cwd=str(REPO_ROOT),
+            )
         else:
-            print(f"  ✗ GitHub release failed: {result.stderr}")
-            print(f"    Tag was created. Create the release manually:")
-            print(f"    gh release create {tag_name} --title 'Hermes Agent v{new_version} ({calver_date})'")
+            result = None
 
-        print(f"\n  🎉 Release v{new_version} ({tag_name}) published!")
+        if result and result.returncode == 0:
+            changelog_file.unlink(missing_ok=True)
+            print(f"  ✓ GitHub release created: {result.stdout.strip()}")
+            print(f"\n  🎉 Release v{new_version} ({tag_name}) published!")
+        else:
+            if result is None:
+                print("  ✗ GitHub release skipped: `gh` CLI not found.")
+            else:
+                print(f"  ✗ GitHub release failed: {result.stderr.strip()}")
+            print(f"    Release notes kept at: {changelog_file}")
+            print(f"    Tag was created locally. Create the release manually:")
+            print(
+                f"    gh release create {tag_name} --title 'Hermes Agent v{new_version} ({calver_date})' "
+                f"--notes-file .release_notes.md {' '.join(str(path) for path in artifacts)}"
+            )
+            print(f"\n  ✓ Release artifacts prepared for manual publish: v{new_version} ({tag_name})")
     else:
         print(f"\n{'='*60}")
         print(f"  Dry run complete. To publish, add --publish")

tests/gateway/test_api_server.py

@@ -427,6 +427,81 @@ class TestChatCompletionsEndpoint:
                 assert "Thinking" in body
                 assert " about it..." in body
 
+    @pytest.mark.asyncio
+    async def test_stream_includes_tool_progress(self, adapter):
+        """tool_progress_callback fires → progress appears in the SSE stream."""
+        import asyncio
+
+        app = _create_app(adapter)
+        async with TestClient(TestServer(app)) as cli:
+            async def _mock_run_agent(**kwargs):
+                cb = kwargs.get("stream_delta_callback")
+                tp_cb = kwargs.get("tool_progress_callback")
+                # Simulate tool progress before streaming content
+                if tp_cb:
+                    tp_cb("terminal", "ls -la", {"command": "ls -la"})
+                if cb:
+                    await asyncio.sleep(0.05)
+                    cb("Here are the files.")
+                return (
+                    {"final_response": "Here are the files.", "messages": [], "api_calls": 1},
+                    {"input_tokens": 10, "output_tokens": 5, "total_tokens": 15},
+                )
+
+            with patch.object(adapter, "_run_agent", side_effect=_mock_run_agent):
+                resp = await cli.post(
+                    "/v1/chat/completions",
+                    json={
+                        "model": "test",
+                        "messages": [{"role": "user", "content": "list files"}],
+                        "stream": True,
+                    },
+                )
+                assert resp.status == 200
+                body = await resp.text()
+                assert "[DONE]" in body
+                # Tool progress message must appear in the stream
+                assert "ls -la" in body
+                # Final content must also be present
+                assert "Here are the files." in body
+
+    @pytest.mark.asyncio
+    async def test_stream_tool_progress_skips_internal_events(self, adapter):
+        """Internal events (name starting with _) are not streamed."""
+        import asyncio
+
+        app = _create_app(adapter)
+        async with TestClient(TestServer(app)) as cli:
+            async def _mock_run_agent(**kwargs):
+                cb = kwargs.get("stream_delta_callback")
+                tp_cb = kwargs.get("tool_progress_callback")
+                if tp_cb:
+                    tp_cb("_thinking", "some internal state", {})
+                    tp_cb("web_search", "Python docs", {"query": "Python docs"})
+                if cb:
+                    await asyncio.sleep(0.05)
+                    cb("Found it.")
+                return (
+                    {"final_response": "Found it.", "messages": [], "api_calls": 1},
+                    {"input_tokens": 10, "output_tokens": 5, "total_tokens": 15},
+                )
+
+            with patch.object(adapter, "_run_agent", side_effect=_mock_run_agent):
+                resp = await cli.post(
+                    "/v1/chat/completions",
+                    json={
+                        "model": "test",
+                        "messages": [{"role": "user", "content": "search"}],
+                        "stream": True,
+                    },
+                )
+                assert resp.status == 200
+                body = await resp.text()
+                # Internal _thinking event should NOT appear
+                assert "some internal state" not in body
+                # Real tool progress should appear
+                assert "Python docs" in body
+
     @pytest.mark.asyncio
     async def test_no_user_message_returns_400(self, adapter):
         app = _create_app(adapter)

tests/gateway/test_session_hygiene.py

@@ -212,47 +212,7 @@ class TestSessionHygieneWarnThreshold:
         assert post_compress_tokens < warn_threshold
 
 
-class TestCompressionWarnRateLimit:
-    """Compression warning messages must be rate-limited per chat_id."""
 
-    def _make_runner(self):
-        from unittest.mock import MagicMock, patch
-        with patch("gateway.run.load_gateway_config"), \
-             patch("gateway.run.SessionStore"), \
-             patch("gateway.run.DeliveryRouter"):
-            from gateway.run import GatewayRunner
-            runner = GatewayRunner.__new__(GatewayRunner)
-            runner._compression_warn_sent = {}
-            runner._compression_warn_cooldown = 3600
-            return runner
-
-    def test_first_warn_is_sent(self):
-        runner = self._make_runner()
-        now = 1_000_000.0
-        last = runner._compression_warn_sent.get("chat:1", 0)
-        assert now - last >= runner._compression_warn_cooldown
-
-    def test_second_warn_suppressed_within_cooldown(self):
-        runner = self._make_runner()
-        now = 1_000_000.0
-        runner._compression_warn_sent["chat:1"] = now - 60  # 1 minute ago
-        last = runner._compression_warn_sent.get("chat:1", 0)
-        assert now - last < runner._compression_warn_cooldown
-
-    def test_warn_allowed_after_cooldown(self):
-        runner = self._make_runner()
-        now = 1_000_000.0
-        runner._compression_warn_sent["chat:1"] = now - 3601  # just past cooldown
-        last = runner._compression_warn_sent.get("chat:1", 0)
-        assert now - last >= runner._compression_warn_cooldown
-
-    def test_rate_limit_is_per_chat(self):
-        """Rate-limiting one chat must not suppress warnings for another."""
-        runner = self._make_runner()
-        now = 1_000_000.0
-        runner._compression_warn_sent["chat:1"] = now - 60  # suppressed
-        last_other = runner._compression_warn_sent.get("chat:2", 0)
-        assert now - last_other >= runner._compression_warn_cooldown
 
 
 class TestEstimatedTokenThreshold:
@@ -421,10 +381,6 @@ async def test_session_hygiene_messages_stay_in_originating_topic(monkeypatch, t
     result = await runner._handle_message(event)
 
     assert result == "ok"
-    assert len(adapter.sent) == 2
-    assert adapter.sent[0]["chat_id"] == "-1001"
-    assert "Session is large" in adapter.sent[0]["content"]
-    assert adapter.sent[0]["metadata"] == {"thread_id": "17585"}
-    assert adapter.sent[1]["chat_id"] == "-1001"
-    assert "Compressed:" in adapter.sent[1]["content"]
-    assert adapter.sent[1]["metadata"] == {"thread_id": "17585"}
+    # Compression warnings are no longer sent to users — compression
+    # happens silently with server-side logging only.
+    assert len(adapter.sent) == 0

tests/gateway/test_update_command.py

@@ -45,6 +45,17 @@ def _make_runner():
 class TestHandleUpdateCommand:
     """Tests for GatewayRunner._handle_update_command."""
 
+    @pytest.mark.asyncio
+    async def test_managed_install_returns_package_manager_guidance(self, monkeypatch):
+        runner = _make_runner()
+        event = _make_event()
+        monkeypatch.setenv("HERMES_MANAGED", "homebrew")
+
+        result = await runner._handle_update_command(event)
+
+        assert "managed by Homebrew" in result
+        assert "brew upgrade hermes-agent" in result
+
     @pytest.mark.asyncio
     async def test_no_git_directory(self, tmp_path):
         """Returns an error when .git does not exist."""
@@ -191,7 +202,7 @@ class TestHandleUpdateCommand:
 
         with patch("gateway.run._hermes_home", hermes_home), \
              patch("gateway.run.__file__", fake_file), \
-             patch("shutil.which", side_effect=lambda x: "/usr/bin/hermes" if x == "hermes" else "/usr/bin/systemd-run"), \
+             patch("shutil.which", side_effect=lambda x: "/usr/bin/hermes" if x == "hermes" else "/usr/bin/setsid"), \
              patch("subprocess.Popen"):
             result = await runner._handle_update_command(event)
 
@@ -204,8 +215,8 @@ class TestHandleUpdateCommand:
         assert not (hermes_home / ".update_exit_code").exists()
 
     @pytest.mark.asyncio
-    async def test_spawns_systemd_run(self, tmp_path):
-        """Uses systemd-run when available."""
+    async def test_spawns_setsid(self, tmp_path):
+        """Uses setsid when available."""
         runner = _make_runner()
         event = _make_event()
 
@@ -225,16 +236,16 @@ class TestHandleUpdateCommand:
              patch("subprocess.Popen", mock_popen):
             result = await runner._handle_update_command(event)
 
-        # Verify systemd-run was used
+        # Verify setsid was used
         call_args = mock_popen.call_args[0][0]
-        assert call_args[0] == "/usr/bin/systemd-run"
-        assert "--scope" in call_args
+        assert call_args[0] == "/usr/bin/setsid"
+        assert call_args[1] == "bash"
         assert ".update_exit_code" in call_args[-1]
         assert "Starting Hermes update" in result
 
     @pytest.mark.asyncio
-    async def test_fallback_nohup_when_no_systemd_run(self, tmp_path):
-        """Falls back to nohup when systemd-run is not available."""
+    async def test_fallback_when_no_setsid(self, tmp_path):
+        """Falls back to start_new_session=True when setsid is not available."""
         runner = _make_runner()
         event = _make_event()
 
@@ -249,24 +260,27 @@ class TestHandleUpdateCommand:
 
         mock_popen = MagicMock()
 
-        def which_no_systemd(x):
+        def which_no_setsid(x):
             if x == "hermes":
                 return "/usr/bin/hermes"
-            if x == "systemd-run":
+            if x == "setsid":
                 return None
             return None
 
         with patch("gateway.run._hermes_home", hermes_home), \
              patch("gateway.run.__file__", fake_file), \
-             patch("shutil.which", side_effect=which_no_systemd), \
+             patch("shutil.which", side_effect=which_no_setsid), \
              patch("subprocess.Popen", mock_popen):
             result = await runner._handle_update_command(event)
 
-        # Verify bash -c nohup fallback was used
+        # Verify plain bash -c fallback (no nohup, no setsid)
         call_args = mock_popen.call_args[0][0]
         assert call_args[0] == "bash"
-        assert "nohup" in call_args[2]
+        assert "nohup" not in call_args[2]
         assert ".update_exit_code" in call_args[2]
+        # start_new_session=True should be in kwargs
+        call_kwargs = mock_popen.call_args[1]
+        assert call_kwargs.get("start_new_session") is True
         assert "Starting Hermes update" in result
 
     @pytest.mark.asyncio

tests/hermes_cli/test_claw.py

@@ -40,6 +40,119 @@ class TestFindMigrationScript:
             assert claw_mod._find_migration_script() is None
 
 
+# ---------------------------------------------------------------------------
+# _find_openclaw_dirs
+# ---------------------------------------------------------------------------
+
+
+class TestFindOpenclawDirs:
+    """Test discovery of OpenClaw directories."""
+
+    def test_finds_openclaw_dir(self, tmp_path):
+        openclaw = tmp_path / ".openclaw"
+        openclaw.mkdir()
+        with patch("pathlib.Path.home", return_value=tmp_path):
+            found = claw_mod._find_openclaw_dirs()
+        assert openclaw in found
+
+    def test_finds_legacy_dirs(self, tmp_path):
+        clawdbot = tmp_path / ".clawdbot"
+        clawdbot.mkdir()
+        moldbot = tmp_path / ".moldbot"
+        moldbot.mkdir()
+        with patch("pathlib.Path.home", return_value=tmp_path):
+            found = claw_mod._find_openclaw_dirs()
+        assert len(found) == 2
+        assert clawdbot in found
+        assert moldbot in found
+
+    def test_returns_empty_when_none_exist(self, tmp_path):
+        with patch("pathlib.Path.home", return_value=tmp_path):
+            found = claw_mod._find_openclaw_dirs()
+        assert found == []
+
+
+# ---------------------------------------------------------------------------
+# _scan_workspace_state
+# ---------------------------------------------------------------------------
+
+
+class TestScanWorkspaceState:
+    """Test scanning for workspace state files."""
+
+    def test_finds_root_state_files(self, tmp_path):
+        (tmp_path / "todo.json").write_text("{}")
+        (tmp_path / "sessions").mkdir()
+        findings = claw_mod._scan_workspace_state(tmp_path)
+        descs = [desc for _, desc in findings]
+        assert any("todo.json" in d for d in descs)
+        assert any("sessions" in d for d in descs)
+
+    def test_finds_workspace_state_files(self, tmp_path):
+        ws = tmp_path / "workspace"
+        ws.mkdir()
+        (ws / "todo.json").write_text("{}")
+        (ws / "sessions").mkdir()
+        findings = claw_mod._scan_workspace_state(tmp_path)
+        descs = [desc for _, desc in findings]
+        assert any("workspace/todo.json" in d for d in descs)
+        assert any("workspace/sessions" in d for d in descs)
+
+    def test_ignores_hidden_dirs(self, tmp_path):
+        scan_dir = tmp_path / "scan_target"
+        scan_dir.mkdir()
+        hidden = scan_dir / ".git"
+        hidden.mkdir()
+        (hidden / "todo.json").write_text("{}")
+        findings = claw_mod._scan_workspace_state(scan_dir)
+        assert len(findings) == 0
+
+    def test_empty_dir_returns_empty(self, tmp_path):
+        scan_dir = tmp_path / "scan_target"
+        scan_dir.mkdir()
+        findings = claw_mod._scan_workspace_state(scan_dir)
+        assert findings == []
+
+
+# ---------------------------------------------------------------------------
+# _archive_directory
+# ---------------------------------------------------------------------------
+
+
+class TestArchiveDirectory:
+    """Test directory archival (rename)."""
+
+    def test_renames_to_pre_migration(self, tmp_path):
+        source = tmp_path / ".openclaw"
+        source.mkdir()
+        (source / "test.txt").write_text("data")
+
+        archive_path = claw_mod._archive_directory(source)
+        assert archive_path == tmp_path / ".openclaw.pre-migration"
+        assert archive_path.is_dir()
+        assert not source.exists()
+        assert (archive_path / "test.txt").read_text() == "data"
+
+    def test_adds_timestamp_when_archive_exists(self, tmp_path):
+        source = tmp_path / ".openclaw"
+        source.mkdir()
+        # Pre-existing archive
+        (tmp_path / ".openclaw.pre-migration").mkdir()
+
+        archive_path = claw_mod._archive_directory(source)
+        assert ".pre-migration-" in archive_path.name
+        assert archive_path.is_dir()
+        assert not source.exists()
+
+    def test_dry_run_does_not_rename(self, tmp_path):
+        source = tmp_path / ".openclaw"
+        source.mkdir()
+
+        archive_path = claw_mod._archive_directory(source, dry_run=True)
+        assert archive_path == tmp_path / ".openclaw.pre-migration"
+        assert source.is_dir()  # Still exists
+
+
 # ---------------------------------------------------------------------------
 # claw_command routing
 # ---------------------------------------------------------------------------
@@ -56,11 +169,24 @@ class TestClawCommand:
             claw_mod.claw_command(args)
         mock.assert_called_once_with(args)
 
+    def test_routes_to_cleanup(self):
+        args = Namespace(claw_action="cleanup", source=None, dry_run=False, yes=False)
+        with patch.object(claw_mod, "_cmd_cleanup") as mock:
+            claw_mod.claw_command(args)
+        mock.assert_called_once_with(args)
+
+    def test_routes_clean_alias(self):
+        args = Namespace(claw_action="clean", source=None, dry_run=False, yes=False)
+        with patch.object(claw_mod, "_cmd_cleanup") as mock:
+            claw_mod.claw_command(args)
+        mock.assert_called_once_with(args)
+
     def test_shows_help_for_no_action(self, capsys):
         args = Namespace(claw_action=None)
         claw_mod.claw_command(args)
         captured = capsys.readouterr()
         assert "migrate" in captured.out
+        assert "cleanup" in captured.out
 
 
 # ---------------------------------------------------------------------------
@@ -168,6 +294,7 @@ class TestCmdMigrate:
             patch.object(claw_mod, "_load_migration_module", return_value=fake_mod),
             patch.object(claw_mod, "get_config_path", return_value=config_path),
             patch.object(claw_mod, "prompt_yes_no", return_value=True),
+            patch.object(claw_mod, "_offer_source_archival"),
         ):
             claw_mod._cmd_migrate(args)
 
@@ -175,6 +302,75 @@ class TestCmdMigrate:
         assert "Migration Results" in captured.out
         assert "Migration complete!" in captured.out
 
+    def test_execute_offers_archival_on_success(self, tmp_path, capsys):
+        """After successful migration, _offer_source_archival should be called."""
+        openclaw_dir = tmp_path / ".openclaw"
+        openclaw_dir.mkdir()
+
+        fake_mod = ModuleType("openclaw_to_hermes")
+        fake_mod.resolve_selected_options = MagicMock(return_value={"soul"})
+        fake_migrator = MagicMock()
+        fake_migrator.migrate.return_value = {
+            "summary": {"migrated": 3, "skipped": 0, "conflict": 0, "error": 0},
+            "items": [
+                {"kind": "soul", "status": "migrated", "destination": str(tmp_path / "SOUL.md")},
+            ],
+        }
+        fake_mod.Migrator = MagicMock(return_value=fake_migrator)
+
+        args = Namespace(
+            source=str(openclaw_dir),
+            dry_run=False, preset="full", overwrite=False,
+            migrate_secrets=False, workspace_target=None,
+            skill_conflict="skip", yes=True,
+        )
+
+        with (
+            patch.object(claw_mod, "_find_migration_script", return_value=tmp_path / "s.py"),
+            patch.object(claw_mod, "_load_migration_module", return_value=fake_mod),
+            patch.object(claw_mod, "get_config_path", return_value=tmp_path / "config.yaml"),
+            patch.object(claw_mod, "save_config"),
+            patch.object(claw_mod, "load_config", return_value={}),
+            patch.object(claw_mod, "_offer_source_archival") as mock_archival,
+        ):
+            claw_mod._cmd_migrate(args)
+
+        mock_archival.assert_called_once_with(openclaw_dir, True)
+
+    def test_dry_run_skips_archival(self, tmp_path, capsys):
+        """Dry run should not offer archival."""
+        openclaw_dir = tmp_path / ".openclaw"
+        openclaw_dir.mkdir()
+
+        fake_mod = ModuleType("openclaw_to_hermes")
+        fake_mod.resolve_selected_options = MagicMock(return_value=set())
+        fake_migrator = MagicMock()
+        fake_migrator.migrate.return_value = {
+            "summary": {"migrated": 2, "skipped": 0, "conflict": 0, "error": 0},
+            "items": [],
+            "preset": "full",
+        }
+        fake_mod.Migrator = MagicMock(return_value=fake_migrator)
+
+        args = Namespace(
+            source=str(openclaw_dir),
+            dry_run=True, preset="full", overwrite=False,
+            migrate_secrets=False, workspace_target=None,
+            skill_conflict="skip", yes=False,
+        )
+
+        with (
+            patch.object(claw_mod, "_find_migration_script", return_value=tmp_path / "s.py"),
+            patch.object(claw_mod, "_load_migration_module", return_value=fake_mod),
+            patch.object(claw_mod, "get_config_path", return_value=tmp_path / "config.yaml"),
+            patch.object(claw_mod, "save_config"),
+            patch.object(claw_mod, "load_config", return_value={}),
+            patch.object(claw_mod, "_offer_source_archival") as mock_archival,
+        ):
+            claw_mod._cmd_migrate(args)
+
+        mock_archival.assert_not_called()
+
     def test_execute_cancelled_by_user(self, tmp_path, capsys):
         openclaw_dir = tmp_path / ".openclaw"
         openclaw_dir.mkdir()
@@ -290,6 +486,172 @@ class TestCmdMigrate:
         assert call_kwargs["migrate_secrets"] is True
 
 
+# ---------------------------------------------------------------------------
+# _offer_source_archival
+# ---------------------------------------------------------------------------
+
+
+class TestOfferSourceArchival:
+    """Test the post-migration archival offer."""
+
+    def test_archives_with_auto_yes(self, tmp_path, capsys):
+        source = tmp_path / ".openclaw"
+        source.mkdir()
+        (source / "workspace").mkdir()
+        (source / "workspace" / "todo.json").write_text("{}")
+
+        claw_mod._offer_source_archival(source, auto_yes=True)
+
+        captured = capsys.readouterr()
+        assert "Archived" in captured.out
+        assert not source.exists()
+        assert (tmp_path / ".openclaw.pre-migration").is_dir()
+
+    def test_skips_when_user_declines(self, tmp_path, capsys):
+        source = tmp_path / ".openclaw"
+        source.mkdir()
+
+        with patch.object(claw_mod, "prompt_yes_no", return_value=False):
+            claw_mod._offer_source_archival(source, auto_yes=False)
+
+        captured = capsys.readouterr()
+        assert "Skipped" in captured.out
+        assert source.is_dir()  # Still exists
+
+    def test_noop_when_source_missing(self, tmp_path, capsys):
+        claw_mod._offer_source_archival(tmp_path / "nonexistent", auto_yes=True)
+        captured = capsys.readouterr()
+        assert captured.out == ""  # No output
+
+    def test_shows_state_files(self, tmp_path, capsys):
+        source = tmp_path / ".openclaw"
+        source.mkdir()
+        ws = source / "workspace"
+        ws.mkdir()
+        (ws / "todo.json").write_text("{}")
+
+        with patch.object(claw_mod, "prompt_yes_no", return_value=False):
+            claw_mod._offer_source_archival(source, auto_yes=False)
+
+        captured = capsys.readouterr()
+        assert "todo.json" in captured.out
+
+    def test_handles_archive_error(self, tmp_path, capsys):
+        source = tmp_path / ".openclaw"
+        source.mkdir()
+
+        with patch.object(claw_mod, "_archive_directory", side_effect=OSError("permission denied")):
+            claw_mod._offer_source_archival(source, auto_yes=True)
+
+        captured = capsys.readouterr()
+        assert "Could not archive" in captured.out
+
+
+# ---------------------------------------------------------------------------
+# _cmd_cleanup
+# ---------------------------------------------------------------------------
+
+
+class TestCmdCleanup:
+    """Test the cleanup command handler."""
+
+    def test_no_dirs_found(self, tmp_path, capsys):
+        args = Namespace(source=None, dry_run=False, yes=False)
+        with patch.object(claw_mod, "_find_openclaw_dirs", return_value=[]):
+            claw_mod._cmd_cleanup(args)
+        captured = capsys.readouterr()
+        assert "No OpenClaw directories found" in captured.out
+
+    def test_dry_run_lists_dirs(self, tmp_path, capsys):
+        openclaw = tmp_path / ".openclaw"
+        openclaw.mkdir()
+        ws = openclaw / "workspace"
+        ws.mkdir()
+        (ws / "todo.json").write_text("{}")
+
+        args = Namespace(source=None, dry_run=True, yes=False)
+        with patch.object(claw_mod, "_find_openclaw_dirs", return_value=[openclaw]):
+            claw_mod._cmd_cleanup(args)
+
+        captured = capsys.readouterr()
+        assert "Would archive" in captured.out
+        assert openclaw.is_dir()  # Not actually archived
+
+    def test_archives_with_yes(self, tmp_path, capsys):
+        openclaw = tmp_path / ".openclaw"
+        openclaw.mkdir()
+        (openclaw / "workspace").mkdir()
+        (openclaw / "workspace" / "todo.json").write_text("{}")
+
+        args = Namespace(source=None, dry_run=False, yes=True)
+        with patch.object(claw_mod, "_find_openclaw_dirs", return_value=[openclaw]):
+            claw_mod._cmd_cleanup(args)
+
+        captured = capsys.readouterr()
+        assert "Archived" in captured.out
+        assert "Cleaned up 1" in captured.out
+        assert not openclaw.exists()
+        assert (tmp_path / ".openclaw.pre-migration").is_dir()
+
+    def test_skips_when_user_declines(self, tmp_path, capsys):
+        openclaw = tmp_path / ".openclaw"
+        openclaw.mkdir()
+
+        args = Namespace(source=None, dry_run=False, yes=False)
+        with (
+            patch.object(claw_mod, "_find_openclaw_dirs", return_value=[openclaw]),
+            patch.object(claw_mod, "prompt_yes_no", return_value=False),
+        ):
+            claw_mod._cmd_cleanup(args)
+
+        captured = capsys.readouterr()
+        assert "Skipped" in captured.out
+        assert openclaw.is_dir()
+
+    def test_explicit_source(self, tmp_path, capsys):
+        custom_dir = tmp_path / "my-openclaw"
+        custom_dir.mkdir()
+        (custom_dir / "todo.json").write_text("{}")
+
+        args = Namespace(source=str(custom_dir), dry_run=False, yes=True)
+        claw_mod._cmd_cleanup(args)
+
+        captured = capsys.readouterr()
+        assert "Archived" in captured.out
+        assert not custom_dir.exists()
+
+    def test_shows_workspace_details(self, tmp_path, capsys):
+        openclaw = tmp_path / ".openclaw"
+        openclaw.mkdir()
+        ws = openclaw / "workspace"
+        ws.mkdir()
+        (ws / "todo.json").write_text("{}")
+        (ws / "SOUL.md").write_text("# Soul")
+
+        args = Namespace(source=None, dry_run=True, yes=False)
+        with patch.object(claw_mod, "_find_openclaw_dirs", return_value=[openclaw]):
+            claw_mod._cmd_cleanup(args)
+
+        captured = capsys.readouterr()
+        assert "workspace/" in captured.out
+        assert "todo.json" in captured.out
+
+    def test_handles_multiple_dirs(self, tmp_path, capsys):
+        openclaw = tmp_path / ".openclaw"
+        openclaw.mkdir()
+        clawdbot = tmp_path / ".clawdbot"
+        clawdbot.mkdir()
+
+        args = Namespace(source=None, dry_run=False, yes=True)
+        with patch.object(claw_mod, "_find_openclaw_dirs", return_value=[openclaw, clawdbot]):
+            claw_mod._cmd_cleanup(args)
+
+        captured = capsys.readouterr()
+        assert "Cleaned up 2" in captured.out
+        assert not openclaw.exists()
+        assert not clawdbot.exists()
+
+
 # ---------------------------------------------------------------------------
 # _print_migration_report
 # ---------------------------------------------------------------------------

tests/hermes_cli/test_managed_installs.py

@@ -0,0 +1,54 @@
+from types import SimpleNamespace
+from unittest.mock import patch
+
+from hermes_cli.config import (
+    format_managed_message,
+    get_managed_system,
+    recommended_update_command,
+)
+from hermes_cli.main import cmd_update
+from tools.skills_hub import OptionalSkillSource
+
+
+def test_get_managed_system_homebrew(monkeypatch):
+    monkeypatch.setenv("HERMES_MANAGED", "homebrew")
+
+    assert get_managed_system() == "Homebrew"
+    assert recommended_update_command() == "brew upgrade hermes-agent"
+
+
+def test_format_managed_message_homebrew(monkeypatch):
+    monkeypatch.setenv("HERMES_MANAGED", "homebrew")
+
+    message = format_managed_message("update Hermes Agent")
+
+    assert "managed by Homebrew" in message
+    assert "brew upgrade hermes-agent" in message
+
+
+def test_recommended_update_command_defaults_to_hermes_update(monkeypatch):
+    monkeypatch.delenv("HERMES_MANAGED", raising=False)
+
+    assert recommended_update_command() == "hermes update"
+
+
+def test_cmd_update_blocks_managed_homebrew(monkeypatch, capsys):
+    monkeypatch.setenv("HERMES_MANAGED", "homebrew")
+
+    with patch("hermes_cli.main.subprocess.run") as mock_run:
+        cmd_update(SimpleNamespace())
+
+    assert not mock_run.called
+    captured = capsys.readouterr()
+    assert "managed by Homebrew" in captured.err
+    assert "brew upgrade hermes-agent" in captured.err
+
+
+def test_optional_skill_source_honors_env_override(monkeypatch, tmp_path):
+    optional_dir = tmp_path / "optional-skills"
+    optional_dir.mkdir()
+    monkeypatch.setenv("HERMES_OPTIONAL_SKILLS", str(optional_dir))
+
+    source = OptionalSkillSource()
+
+    assert source._optional_dir == optional_dir

tests/hermes_cli/test_update_gateway_restart.py

@@ -25,6 +25,8 @@ def _make_run_side_effect(
     verify_ok=True,
     commit_count="3",
     systemd_active=False,
+    system_service_active=False,
+    system_restart_rc=0,
     launchctl_loaded=False,
 ):
     """Build a subprocess.run side_effect that simulates git + service commands."""
@@ -45,14 +47,23 @@ def _make_run_side_effect(
         if "rev-list" in joined:
             return subprocess.CompletedProcess(cmd, 0, stdout=f"{commit_count}\n", stderr="")
 
-        # systemctl --user is-active
+        # systemctl is-active — distinguish --user from system scope
         if "systemctl" in joined and "is-active" in joined:
-            if systemd_active:
-                return subprocess.CompletedProcess(cmd, 0, stdout="active\n", stderr="")
-            return subprocess.CompletedProcess(cmd, 3, stdout="inactive\n", stderr="")
+            if "--user" in joined:
+                if systemd_active:
+                    return subprocess.CompletedProcess(cmd, 0, stdout="active\n", stderr="")
+                return subprocess.CompletedProcess(cmd, 3, stdout="inactive\n", stderr="")
+            else:
+                # System-level check (no --user)
+                if system_service_active:
+                    return subprocess.CompletedProcess(cmd, 0, stdout="active\n", stderr="")
+                return subprocess.CompletedProcess(cmd, 3, stdout="inactive\n", stderr="")
 
-        # systemctl --user restart
+        # systemctl restart — distinguish --user from system scope
         if "systemctl" in joined and "restart" in joined:
+            if "--user" not in joined and system_service_active:
+                stderr = "" if system_restart_rc == 0 else "Failed to restart: Permission denied"
+                return subprocess.CompletedProcess(cmd, system_restart_rc, stdout="", stderr=stderr)
             return subprocess.CompletedProcess(cmd, 0, stdout="", stderr="")
 
         # launchctl list ai.hermes.gateway
@@ -393,3 +404,91 @@ class TestCmdUpdateLaunchdRestart:
         assert "Stopped gateway" not in captured
         assert "Gateway restarted" not in captured
         assert "Gateway restarted via launchd" not in captured
+
+
+# ---------------------------------------------------------------------------
+# cmd_update — system-level systemd service detection
+# ---------------------------------------------------------------------------
+
+
+class TestCmdUpdateSystemService:
+    """cmd_update detects system-level gateway services where --user fails."""
+
+    @patch("shutil.which", return_value=None)
+    @patch("subprocess.run")
+    def test_update_detects_system_service_and_restarts(
+        self, mock_run, _mock_which, mock_args, capsys, monkeypatch,
+    ):
+        """When user systemd is inactive but a system service exists, restart via system scope."""
+        monkeypatch.setattr(gateway_cli, "is_macos", lambda: False)
+        monkeypatch.setattr(gateway_cli, "is_linux", lambda: True)
+
+        mock_run.side_effect = _make_run_side_effect(
+            commit_count="3",
+            systemd_active=False,
+            system_service_active=True,
+        )
+
+        with patch("gateway.status.get_running_pid", return_value=12345), \
+             patch("gateway.status.remove_pid_file"):
+            cmd_update(mock_args)
+
+        captured = capsys.readouterr().out
+        assert "system gateway service" in captured.lower()
+        assert "Gateway restarted (system service)" in captured
+        # Verify systemctl restart (no --user) was called
+        restart_calls = [
+            c for c in mock_run.call_args_list
+            if "restart" in " ".join(str(a) for a in c.args[0])
+            and "systemctl" in " ".join(str(a) for a in c.args[0])
+            and "--user" not in " ".join(str(a) for a in c.args[0])
+        ]
+        assert len(restart_calls) == 1
+
+    @patch("shutil.which", return_value=None)
+    @patch("subprocess.run")
+    def test_update_system_service_restart_failure_shows_sudo_hint(
+        self, mock_run, _mock_which, mock_args, capsys, monkeypatch,
+    ):
+        """When system service restart fails (e.g. no root), show sudo hint."""
+        monkeypatch.setattr(gateway_cli, "is_macos", lambda: False)
+        monkeypatch.setattr(gateway_cli, "is_linux", lambda: True)
+
+        mock_run.side_effect = _make_run_side_effect(
+            commit_count="3",
+            systemd_active=False,
+            system_service_active=True,
+            system_restart_rc=1,
+        )
+
+        with patch("gateway.status.get_running_pid", return_value=12345), \
+             patch("gateway.status.remove_pid_file"):
+            cmd_update(mock_args)
+
+        captured = capsys.readouterr().out
+        assert "sudo systemctl restart" in captured
+
+    @patch("shutil.which", return_value=None)
+    @patch("subprocess.run")
+    def test_user_service_takes_priority_over_system(
+        self, mock_run, _mock_which, mock_args, capsys, monkeypatch,
+    ):
+        """When both user and system services are active, user wins."""
+        monkeypatch.setattr(gateway_cli, "is_macos", lambda: False)
+        monkeypatch.setattr(gateway_cli, "is_linux", lambda: True)
+
+        mock_run.side_effect = _make_run_side_effect(
+            commit_count="3",
+            systemd_active=True,
+            system_service_active=True,
+        )
+
+        with patch("gateway.status.get_running_pid", return_value=12345), \
+             patch("gateway.status.remove_pid_file"), \
+             patch("os.kill"):
+            cmd_update(mock_args)
+
+        captured = capsys.readouterr().out
+        # Should restart via user service, not system
+        assert "Gateway restarted." in captured
+        assert "(system service)" not in captured

tests/test_cli_provider_resolution.py

@@ -424,6 +424,7 @@ def test_cmd_model_falls_back_to_auto_on_invalid_provider(monkeypatch, capsys):
 
     monkeypatch.setattr("hermes_cli.auth.resolve_provider", _resolve_provider)
     monkeypatch.setattr(hermes_main, "_prompt_provider_choice", lambda choices: len(choices) - 1)
+    monkeypatch.setattr("sys.stdin", type("FakeTTY", (), {"isatty": lambda self: True})())
 
     hermes_main.cmd_model(SimpleNamespace())
     output = capsys.readouterr().out

tests/test_codex_execution_paths.py

@@ -112,7 +112,7 @@ def test_cron_run_job_codex_path_handles_internal_401_refresh(monkeypatch):
     _Codex401ThenSuccessAgent.last_init = {}
 
     success, output, final_response, error = cron_scheduler.run_job(
-        {"id": "job-1", "name": "Codex Refresh Test", "prompt": "ping"}
+        {"id": "job-1", "name": "Codex Refresh Test", "prompt": "ping", "model": "gpt-5.3-codex"}
     )
 
     assert success is True
@@ -139,6 +139,7 @@ def test_gateway_run_agent_codex_path_handles_internal_401_refresh(monkeypatch):
         },
     )
     monkeypatch.setenv("HERMES_TOOL_PROGRESS", "false")
+    monkeypatch.setenv("HERMES_MODEL", "gpt-5.3-codex")
 
     _Codex401ThenSuccessAgent.refresh_attempts = 0
     _Codex401ThenSuccessAgent.last_init = {}

tests/test_compression_persistence.py

@@ -0,0 +1,202 @@
+"""Tests for context compression persistence in the gateway.
+
+Verifies that when context compression fires during run_conversation(),
+the compressed messages are properly persisted to both SQLite (via the
+agent) and JSONL (via the gateway).
+
+Bug scenario (pre-fix):
+  1. Gateway loads 200-message history, passes to agent
+  2. Agent's run_conversation() compresses to ~30 messages mid-run
+  3. _compress_context() resets _last_flushed_db_idx = 0
+  4. On exit, _flush_messages_to_session_db() calculates:
+     flush_from = max(len(conversation_history=200), _last_flushed_db_idx=0) = 200
+  5. messages[200:] is empty (only ~30 messages after compression)
+  6. Nothing written to new session's SQLite — compressed context lost
+  7. Gateway's history_offset was still 200, producing empty new_messages
+  8. Fallback wrote only user/assistant pair — summary lost
+"""
+
+import os
+import tempfile
+from pathlib import Path
+from unittest.mock import MagicMock, patch
+
+import pytest
+
+
+# ---------------------------------------------------------------------------
+# Part 1: Agent-side — _flush_messages_to_session_db after compression
+# ---------------------------------------------------------------------------
+
+class TestFlushAfterCompression:
+    """Verify that compressed messages are flushed to the new session's SQLite
+    even when conversation_history (from the original session) is longer than
+    the compressed messages list."""
+
+    def _make_agent(self, session_db):
+        with patch.dict(os.environ, {"OPENROUTER_API_KEY": "test-key"}):
+            from run_agent import AIAgent
+            agent = AIAgent(
+                model="test/model",
+                quiet_mode=True,
+                session_db=session_db,
+                session_id="original-session",
+                skip_context_files=True,
+                skip_memory=True,
+            )
+        return agent
+
+    def test_flush_after_compression_with_long_history(self):
+        """The actual bug: conversation_history longer than compressed messages.
+
+        Before the fix, flush_from = max(len(conversation_history), 0) = 200,
+        but messages only has ~30 entries, so messages[200:] is empty.
+        After the fix, conversation_history is cleared to None after compression,
+        so flush_from = max(0, 0) = 0, and ALL compressed messages are written.
+        """
+        from hermes_state import SessionDB
+
+        with tempfile.TemporaryDirectory() as tmpdir:
+            db_path = Path(tmpdir) / "test.db"
+            db = SessionDB(db_path=db_path)
+
+            agent = self._make_agent(db)
+
+            # Simulate the original long history (200 messages)
+            original_history = [
+                {"role": "user" if i % 2 == 0 else "assistant",
+                 "content": f"message {i}"}
+                for i in range(200)
+            ]
+
+            # First, flush original messages to the original session
+            agent._flush_messages_to_session_db(original_history, [])
+            original_rows = db.get_messages("original-session")
+            assert len(original_rows) == 200
+
+            # Now simulate compression: new session, reset idx, shorter messages
+            agent.session_id = "compressed-session"
+            db.create_session(session_id="compressed-session", source="test")
+            agent._last_flushed_db_idx = 0
+
+            # The compressed messages (summary + tail + new turn)
+            compressed_messages = [
+                {"role": "user", "content": "[CONTEXT COMPACTION] Summary of work..."},
+                {"role": "user", "content": "What should we do next?"},
+                {"role": "assistant", "content": "Let me check..."},
+                {"role": "user", "content": "new question"},
+                {"role": "assistant", "content": "new answer"},
+            ]
+
+            # THE BUG: passing the original history as conversation_history
+            # causes flush_from = max(200, 0) = 200, skipping everything.
+            # After the fix, conversation_history should be None.
+            agent._flush_messages_to_session_db(compressed_messages, None)
+
+            new_rows = db.get_messages("compressed-session")
+            assert len(new_rows) == 5, (
+                f"Expected 5 compressed messages in new session, got {len(new_rows)}. "
+                f"Compression persistence bug: messages not written to SQLite."
+            )
+
+    def test_flush_with_stale_history_loses_messages(self):
+        """Demonstrates the bug condition: stale conversation_history causes data loss."""
+        from hermes_state import SessionDB
+
+        with tempfile.TemporaryDirectory() as tmpdir:
+            db_path = Path(tmpdir) / "test.db"
+            db = SessionDB(db_path=db_path)
+
+            agent = self._make_agent(db)
+
+            # Simulate compression reset
+            agent.session_id = "new-session"
+            db.create_session(session_id="new-session", source="test")
+            agent._last_flushed_db_idx = 0
+
+            compressed = [
+                {"role": "user", "content": "summary"},
+                {"role": "assistant", "content": "continuing..."},
+            ]
+
+            # Bug: passing a conversation_history longer than compressed messages
+            stale_history = [{"role": "user", "content": f"msg{i}"} for i in range(100)]
+            agent._flush_messages_to_session_db(compressed, stale_history)
+
+            rows = db.get_messages("new-session")
+            # With the stale history, flush_from = max(100, 0) = 100
+            # But compressed only has 2 entries → messages[100:] = empty
+            assert len(rows) == 0, (
+                "Expected 0 messages with stale conversation_history "
+                "(this test verifies the bug condition exists)"
+            )
+
+
+# ---------------------------------------------------------------------------
+# Part 2: Gateway-side — history_offset after session split
+# ---------------------------------------------------------------------------
+
+class TestGatewayHistoryOffsetAfterSplit:
+    """Verify that when the agent creates a new session during compression,
+    the gateway uses history_offset=0 so all compressed messages are written
+    to the JSONL transcript."""
+
+    def test_history_offset_zero_on_session_split(self):
+        """When agent.session_id differs from the original, history_offset must be 0."""
+        # This tests the logic in gateway/run.py run_sync():
+        # _session_was_split = agent.session_id != session_id
+        # _effective_history_offset = 0 if _session_was_split else len(agent_history)
+
+        original_session_id = "session-abc"
+        agent_session_id = "session-compressed-xyz"  # Different = compression happened
+        agent_history_len = 200
+
+        # Simulate the gateway's offset calculation (post-fix)
+        _session_was_split = (agent_session_id != original_session_id)
+        _effective_history_offset = 0 if _session_was_split else agent_history_len
+
+        assert _session_was_split is True
+        assert _effective_history_offset == 0
+
+    def test_history_offset_preserved_without_split(self):
+        """When no compression happened, history_offset is the original length."""
+        session_id = "session-abc"
+        agent_session_id = "session-abc"  # Same = no compression
+        agent_history_len = 200
+
+        _session_was_split = (agent_session_id != session_id)
+        _effective_history_offset = 0 if _session_was_split else agent_history_len
+
+        assert _session_was_split is False
+        assert _effective_history_offset == 200
+
+    def test_new_messages_extraction_after_split(self):
+        """After compression with offset=0, new_messages should be ALL agent messages."""
+        # Simulates the gateway's new_messages calculation
+        agent_messages = [
+            {"role": "user", "content": "[CONTEXT COMPACTION] Summary..."},
+            {"role": "user", "content": "recent question"},
+            {"role": "assistant", "content": "recent answer"},
+            {"role": "user", "content": "new question"},
+            {"role": "assistant", "content": "new answer"},
+        ]
+        history_offset = 0  # After fix: 0 on session split
+
+        new_messages = agent_messages[history_offset:] if len(agent_messages) > history_offset else []
+        assert len(new_messages) == 5, (
+            f"Expected all 5 messages with offset=0, got {len(new_messages)}"
+        )
+
+    def test_new_messages_empty_with_stale_offset(self):
+        """Demonstrates the bug: stale offset produces empty new_messages."""
+        agent_messages = [
+            {"role": "user", "content": "summary"},
+            {"role": "assistant", "content": "answer"},
+        ]
+        # Bug: offset is the pre-compression history length
+        history_offset = 200
+
+        new_messages = agent_messages[history_offset:] if len(agent_messages) > history_offset else []
+        assert len(new_messages) == 0, (
+            "Expected 0 messages with stale offset=200 (demonstrates the bug)"
+        )

tests/test_packaging_metadata.py

@@ -0,0 +1,22 @@
+from pathlib import Path
+import tomllib
+
+
+REPO_ROOT = Path(__file__).resolve().parents[1]
+
+
+def test_faster_whisper_is_not_a_base_dependency():
+    data = tomllib.loads((REPO_ROOT / "pyproject.toml").read_text(encoding="utf-8"))
+    deps = data["project"]["dependencies"]
+
+    assert not any(dep.startswith("faster-whisper") for dep in deps)
+
+    voice_extra = data["project"]["optional-dependencies"]["voice"]
+    assert any(dep.startswith("faster-whisper") for dep in voice_extra)
+
+
+def test_manifest_includes_bundled_skills():
+    manifest = (REPO_ROOT / "MANIFEST.in").read_text(encoding="utf-8")
+
+    assert "graft skills" in manifest
+    assert "graft optional-skills" in manifest

tests/test_runtime_provider_resolution.py

@@ -545,7 +545,7 @@ def test_alibaba_default_coding_intl_endpoint_uses_chat_completions(monkeypatch)
 
     assert resolved["provider"] == "alibaba"
     assert resolved["api_mode"] == "chat_completions"
-    assert resolved["base_url"] == "https://coding-intl.dashscope.aliyuncs.com/v1"
+    assert resolved["base_url"] == "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
 
 
 def test_alibaba_anthropic_endpoint_override_uses_anthropic_messages(monkeypatch):

tests/test_trajectory_compressor.py

@@ -405,12 +405,13 @@ class TestGenerateSummary:
     @pytest.mark.asyncio
     async def test_generate_summary_async_handles_none_content(self):
         tc = _make_compressor()
-        tc.async_client = MagicMock()
-        tc.async_client.chat.completions.create = AsyncMock(
+        mock_client = MagicMock()
+        mock_client.chat.completions.create = AsyncMock(
             return_value=SimpleNamespace(
                 choices=[SimpleNamespace(message=SimpleNamespace(content=None))]
             )
         )
+        tc._get_async_client = MagicMock(return_value=mock_client)
         metrics = TrajectoryMetrics()
 
         summary = await tc._generate_summary_async("Turn content", metrics)

tests/tools/test_browser_camofox.py

@@ -235,8 +235,13 @@ class TestCamofoxGetImages:
         mock_post.return_value = _mock_response(json_data={"tabId": "tab10", "url": "https://x.com"})
         camofox_navigate("https://x.com", task_id="t10")
 
+        # camofox_get_images parses images from the accessibility tree snapshot
+        snapshot_text = (
+            '- img "Logo"\n'
+            '  /url: https://x.com/img.png\n'
+        )
         mock_get.return_value = _mock_response(json_data={
-            "images": [{"src": "https://x.com/img.png", "alt": "Logo"}],
+            "snapshot": snapshot_text,
         })
         result = json.loads(camofox_get_images(task_id="t10"))
         assert result["success"] is True

tools/skills_hub.py

@@ -2115,7 +2115,11 @@ class OptionalSkillSource(SkillSource):
     """
 
     def __init__(self):
-        self._optional_dir = Path(__file__).parent.parent / "optional-skills"
+        from hermes_constants import get_optional_skills_dir
+
+        self._optional_dir = get_optional_skills_dir(
+            Path(__file__).parent.parent / "optional-skills"
+        )
 
     def source_id(self) -> str:
         return "official"

uv.lock

@@ -1133,24 +1133,6 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/97/a8/c070e1340636acb38d4e6a7e45c46d168a462b48b9b3257e14ca0e5af79b/environs-14.6.0-py3-none-any.whl", hash = "sha256:f8fb3d6c6a55872b0c6db077a28f5a8c7b8984b7c32029613d44cef95cfc0812", size = 17205, upload-time = "2026-02-20T04:02:07.299Z" },
 ]
 
-[[package]]
-name = "exa-py"
-version = "2.10.2"
-source = { registry = "https://pypi.org/simple" }
-dependencies = [
-    { name = "httpcore" },
-    { name = "httpx" },
-    { name = "openai" },
-    { name = "pydantic" },
-    { name = "python-dotenv" },
-    { name = "requests" },
-    { name = "typing-extensions" },
-]
-sdist = { url = "https://files.pythonhosted.org/packages/fe/4f/f06a6f277d668f143e330fe503b0027cc5fed753b22c3e161f8cbbccdf65/exa_py-2.10.2.tar.gz", hash = "sha256:f781f30b199f1102333384728adae64bb15a6bbcabfa97e91fd705f90acffc45", size = 53792, upload-time = "2026-03-26T20:29:35.764Z" }
-wheels = [
-    { url = "https://files.pythonhosted.org/packages/e2/bc/7a34e904a415040ba626948d0b0a36a08cd073f12b13342578a68331be3c/exa_py-2.10.2-py3-none-any.whl", hash = "sha256:ecb2a7581f4b7a8aeb6b434acce1bbc40f92ed1d4126b2aa6029913acd904a47", size = 72248, upload-time = "2026-03-26T20:29:37.306Z" },
-]
-
 [[package]]
 name = "execnet"
 version = "2.1.2"
@@ -1528,7 +1510,6 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/f3/47/16400cb42d18d7a6bb46f0626852c1718612e35dcb0dffa16bbaffdf5dd2/greenlet-3.3.2-cp311-cp311-macosx_11_0_universal2.whl", hash = "sha256:c56692189a7d1c7606cb794be0a8381470d95c57ce5be03fb3d0ef57c7853b86", size = 278890, upload-time = "2026-02-20T20:19:39.263Z" },
     { url = "https://files.pythonhosted.org/packages/a3/90/42762b77a5b6aa96cd8c0e80612663d39211e8ae8a6cd47c7f1249a66262/greenlet-3.3.2-cp311-cp311-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:1ebd458fa8285960f382841da585e02201b53a5ec2bac6b156fc623b5ce4499f", size = 581120, upload-time = "2026-02-20T20:47:30.161Z" },
     { url = "https://files.pythonhosted.org/packages/bf/6f/f3d64f4fa0a9c7b5c5b3c810ff1df614540d5aa7d519261b53fba55d4df9/greenlet-3.3.2-cp311-cp311-manylinux_2_24_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:a443358b33c4ec7b05b79a7c8b466f5d275025e750298be7340f8fc63dff2a55", size = 594363, upload-time = "2026-02-20T20:55:56.965Z" },
-    { url = "https://files.pythonhosted.org/packages/9c/8b/1430a04657735a3f23116c2e0d5eb10220928846e4537a938a41b350bed6/greenlet-3.3.2-cp311-cp311-manylinux_2_24_s390x.manylinux_2_28_s390x.whl", hash = "sha256:4375a58e49522698d3e70cc0b801c19433021b5c37686f7ce9c65b0d5c8677d2", size = 605046, upload-time = "2026-02-20T21:02:45.234Z" },
     { url = "https://files.pythonhosted.org/packages/72/83/3e06a52aca8128bdd4dcd67e932b809e76a96ab8c232a8b025b2850264c5/greenlet-3.3.2-cp311-cp311-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:8e2cd90d413acbf5e77ae41e5d3c9b3ac1d011a756d7284d7f3f2b806bbd6358", size = 594156, upload-time = "2026-02-20T20:20:59.955Z" },
     { url = "https://files.pythonhosted.org/packages/70/79/0de5e62b873e08fe3cef7dbe84e5c4bc0e8ed0c7ff131bccb8405cd107c8/greenlet-3.3.2-cp311-cp311-musllinux_1_2_aarch64.whl", hash = "sha256:442b6057453c8cb29b4fb36a2ac689382fc71112273726e2423f7f17dc73bf99", size = 1554649, upload-time = "2026-02-20T20:49:32.293Z" },
     { url = "https://files.pythonhosted.org/packages/5a/00/32d30dee8389dc36d42170a9c66217757289e2afb0de59a3565260f38373/greenlet-3.3.2-cp311-cp311-musllinux_1_2_x86_64.whl", hash = "sha256:45abe8eb6339518180d5a7fa47fa01945414d7cca5ecb745346fc6a87d2750be", size = 1619472, upload-time = "2026-02-20T20:21:07.966Z" },
@@ -1537,7 +1518,6 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/ea/ab/1608e5a7578e62113506740b88066bf09888322a311cff602105e619bd87/greenlet-3.3.2-cp312-cp312-macosx_11_0_universal2.whl", hash = "sha256:ac8d61d4343b799d1e526db579833d72f23759c71e07181c2d2944e429eb09cd", size = 280358, upload-time = "2026-02-20T20:17:43.971Z" },
     { url = "https://files.pythonhosted.org/packages/a5/23/0eae412a4ade4e6623ff7626e38998cb9b11e9ff1ebacaa021e4e108ec15/greenlet-3.3.2-cp312-cp312-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:3ceec72030dae6ac0c8ed7591b96b70410a8be370b6a477b1dbc072856ad02bd", size = 601217, upload-time = "2026-02-20T20:47:31.462Z" },
     { url = "https://files.pythonhosted.org/packages/f8/16/5b1678a9c07098ecb9ab2dd159fafaf12e963293e61ee8d10ecb55273e5e/greenlet-3.3.2-cp312-cp312-manylinux_2_24_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:a2a5be83a45ce6188c045bcc44b0ee037d6a518978de9a5d97438548b953a1ac", size = 611792, upload-time = "2026-02-20T20:55:58.423Z" },
-    { url = "https://files.pythonhosted.org/packages/5c/c5/cc09412a29e43406eba18d61c70baa936e299bc27e074e2be3806ed29098/greenlet-3.3.2-cp312-cp312-manylinux_2_24_s390x.manylinux_2_28_s390x.whl", hash = "sha256:ae9e21c84035c490506c17002f5c8ab25f980205c3e61ddb3a2a2a2e6c411fcb", size = 626250, upload-time = "2026-02-20T21:02:46.596Z" },
     { url = "https://files.pythonhosted.org/packages/50/1f/5155f55bd71cabd03765a4aac9ac446be129895271f73872c36ebd4b04b6/greenlet-3.3.2-cp312-cp312-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:43e99d1749147ac21dde49b99c9abffcbc1e2d55c67501465ef0930d6e78e070", size = 613875, upload-time = "2026-02-20T20:21:01.102Z" },
     { url = "https://files.pythonhosted.org/packages/fc/dd/845f249c3fcd69e32df80cdab059b4be8b766ef5830a3d0aa9d6cad55beb/greenlet-3.3.2-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:4c956a19350e2c37f2c48b336a3afb4bff120b36076d9d7fb68cb44e05d95b79", size = 1571467, upload-time = "2026-02-20T20:49:33.495Z" },
     { url = "https://files.pythonhosted.org/packages/2a/50/2649fe21fcc2b56659a452868e695634722a6655ba245d9f77f5656010bf/greenlet-3.3.2-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:6c6f8ba97d17a1e7d664151284cb3315fc5f8353e75221ed4324f84eb162b395", size = 1640001, upload-time = "2026-02-20T20:21:09.154Z" },
@@ -1546,7 +1526,6 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/ac/48/f8b875fa7dea7dd9b33245e37f065af59df6a25af2f9561efa8d822fde51/greenlet-3.3.2-cp313-cp313-macosx_11_0_universal2.whl", hash = "sha256:aa6ac98bdfd716a749b84d4034486863fd81c3abde9aa3cf8eff9127981a4ae4", size = 279120, upload-time = "2026-02-20T20:19:01.9Z" },
     { url = "https://files.pythonhosted.org/packages/49/8d/9771d03e7a8b1ee456511961e1b97a6d77ae1dea4a34a5b98eee706689d3/greenlet-3.3.2-cp313-cp313-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:ab0c7e7901a00bc0a7284907273dc165b32e0d109a6713babd04471327ff7986", size = 603238, upload-time = "2026-02-20T20:47:32.873Z" },
     { url = "https://files.pythonhosted.org/packages/59/0e/4223c2bbb63cd5c97f28ffb2a8aee71bdfb30b323c35d409450f51b91e3e/greenlet-3.3.2-cp313-cp313-manylinux_2_24_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:d248d8c23c67d2291ffd47af766e2a3aa9fa1c6703155c099feb11f526c63a92", size = 614219, upload-time = "2026-02-20T20:55:59.817Z" },
-    { url = "https://files.pythonhosted.org/packages/94/2b/4d012a69759ac9d77210b8bfb128bc621125f5b20fc398bce3940d036b1c/greenlet-3.3.2-cp313-cp313-manylinux_2_24_s390x.manylinux_2_28_s390x.whl", hash = "sha256:ccd21bb86944ca9be6d967cf7691e658e43417782bce90b5d2faeda0ff78a7dd", size = 628268, upload-time = "2026-02-20T21:02:48.024Z" },
     { url = "https://files.pythonhosted.org/packages/7a/34/259b28ea7a2a0c904b11cd36c79b8cef8019b26ee5dbe24e73b469dea347/greenlet-3.3.2-cp313-cp313-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:b6997d360a4e6a4e936c0f9625b1c20416b8a0ea18a8e19cabbefc712e7397ab", size = 616774, upload-time = "2026-02-20T20:21:02.454Z" },
     { url = "https://files.pythonhosted.org/packages/0a/03/996c2d1689d486a6e199cb0f1cf9e4aa940c500e01bdf201299d7d61fa69/greenlet-3.3.2-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:64970c33a50551c7c50491671265d8954046cb6e8e2999aacdd60e439b70418a", size = 1571277, upload-time = "2026-02-20T20:49:34.795Z" },
     { url = "https://files.pythonhosted.org/packages/d9/c4/2570fc07f34a39f2caf0bf9f24b0a1a0a47bc2e8e465b2c2424821389dfc/greenlet-3.3.2-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:1a9172f5bf6bd88e6ba5a84e0a68afeac9dc7b6b412b245dd64f52d83c81e55b", size = 1640455, upload-time = "2026-02-20T20:21:10.261Z" },
@@ -1555,7 +1534,6 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/3f/ae/8bffcbd373b57a5992cd077cbe8858fff39110480a9d50697091faea6f39/greenlet-3.3.2-cp314-cp314-macosx_11_0_universal2.whl", hash = "sha256:8d1658d7291f9859beed69a776c10822a0a799bc4bfe1bd4272bb60e62507dab", size = 279650, upload-time = "2026-02-20T20:18:00.783Z" },
     { url = "https://files.pythonhosted.org/packages/d1/c0/45f93f348fa49abf32ac8439938726c480bd96b2a3c6f4d949ec0124b69f/greenlet-3.3.2-cp314-cp314-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:18cb1b7337bca281915b3c5d5ae19f4e76d35e1df80f4ad3c1a7be91fadf1082", size = 650295, upload-time = "2026-02-20T20:47:34.036Z" },
     { url = "https://files.pythonhosted.org/packages/b3/de/dd7589b3f2b8372069ab3e4763ea5329940fc7ad9dcd3e272a37516d7c9b/greenlet-3.3.2-cp314-cp314-manylinux_2_24_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:c2e47408e8ce1c6f1ceea0dffcdf6ebb85cc09e55c7af407c99f1112016e45e9", size = 662163, upload-time = "2026-02-20T20:56:01.295Z" },
-    { url = "https://files.pythonhosted.org/packages/cd/ac/85804f74f1ccea31ba518dcc8ee6f14c79f73fe36fa1beba38930806df09/greenlet-3.3.2-cp314-cp314-manylinux_2_24_s390x.manylinux_2_28_s390x.whl", hash = "sha256:e3cb43ce200f59483eb82949bf1835a99cf43d7571e900d7c8d5c62cdf25d2f9", size = 675371, upload-time = "2026-02-20T21:02:49.664Z" },
     { url = "https://files.pythonhosted.org/packages/d2/d8/09bfa816572a4d83bccd6750df1926f79158b1c36c5f73786e26dbe4ee38/greenlet-3.3.2-cp314-cp314-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:63d10328839d1973e5ba35e98cccbca71b232b14051fd957b6f8b6e8e80d0506", size = 664160, upload-time = "2026-02-20T20:21:04.015Z" },
     { url = "https://files.pythonhosted.org/packages/48/cf/56832f0c8255d27f6c35d41b5ec91168d74ec721d85f01a12131eec6b93c/greenlet-3.3.2-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:8e4ab3cfb02993c8cc248ea73d7dae6cec0253e9afa311c9b37e603ca9fad2ce", size = 1619181, upload-time = "2026-02-20T20:49:36.052Z" },
     { url = "https://files.pythonhosted.org/packages/0a/23/b90b60a4aabb4cec0796e55f25ffbfb579a907c3898cd2905c8918acaa16/greenlet-3.3.2-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:94ad81f0fd3c0c0681a018a976e5c2bd2ca2d9d94895f23e7bb1af4e8af4e2d5", size = 1687713, upload-time = "2026-02-20T20:21:11.684Z" },
@@ -1564,7 +1542,6 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/98/6d/8f2ef704e614bcf58ed43cfb8d87afa1c285e98194ab2cfad351bf04f81e/greenlet-3.3.2-cp314-cp314t-macosx_11_0_universal2.whl", hash = "sha256:e26e72bec7ab387ac80caa7496e0f908ff954f31065b0ffc1f8ecb1338b11b54", size = 286617, upload-time = "2026-02-20T20:19:29.856Z" },
     { url = "https://files.pythonhosted.org/packages/5e/0d/93894161d307c6ea237a43988f27eba0947b360b99ac5239ad3fe09f0b47/greenlet-3.3.2-cp314-cp314t-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:8b466dff7a4ffda6ca975979bab80bdadde979e29fc947ac3be4451428d8b0e4", size = 655189, upload-time = "2026-02-20T20:47:35.742Z" },
     { url = "https://files.pythonhosted.org/packages/f5/2c/d2d506ebd8abcb57386ec4f7ba20f4030cbe56eae541bc6fd6ef399c0b41/greenlet-3.3.2-cp314-cp314t-manylinux_2_24_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:b8bddc5b73c9720bea487b3bffdb1840fe4e3656fba3bd40aa1489e9f37877ff", size = 658225, upload-time = "2026-02-20T20:56:02.527Z" },
-    { url = "https://files.pythonhosted.org/packages/d1/67/8197b7e7e602150938049d8e7f30de1660cfb87e4c8ee349b42b67bdb2e1/greenlet-3.3.2-cp314-cp314t-manylinux_2_24_s390x.manylinux_2_28_s390x.whl", hash = "sha256:59b3e2c40f6706b05a9cd299c836c6aa2378cabe25d021acd80f13abf81181cf", size = 666581, upload-time = "2026-02-20T21:02:51.526Z" },
     { url = "https://files.pythonhosted.org/packages/8e/30/3a09155fbf728673a1dea713572d2d31159f824a37c22da82127056c44e4/greenlet-3.3.2-cp314-cp314t-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:b26b0f4428b871a751968285a1ac9648944cea09807177ac639b030bddebcea4", size = 657907, upload-time = "2026-02-20T20:21:05.259Z" },
     { url = "https://files.pythonhosted.org/packages/f3/fd/d05a4b7acd0154ed758797f0a43b4c0962a843bedfe980115e842c5b2d08/greenlet-3.3.2-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:1fb39a11ee2e4d94be9a76671482be9398560955c9e568550de0224e41104727", size = 1618857, upload-time = "2026-02-20T20:49:37.309Z" },
     { url = "https://files.pythonhosted.org/packages/6f/e1/50ee92a5db521de8f35075b5eff060dd43d39ebd46c2181a2042f7070385/greenlet-3.3.2-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:20154044d9085151bc309e7689d6f7ba10027f8f5a8c0676ad398b951913d89e", size = 1680010, upload-time = "2026-02-20T20:21:13.427Z" },
@@ -1623,12 +1600,11 @@ wheels = [
 
 [[package]]
 name = "hermes-agent"
-version = "0.6.0"
+version = "0.5.0"
 source = { editable = "." }
 dependencies = [
     { name = "anthropic" },
     { name = "edge-tts" },
-    { name = "exa-py" },
     { name = "fal-client" },
     { name = "faster-whisper" },
     { name = "fire" },
@@ -1660,8 +1636,6 @@ all = [
     { name = "discord-py", extra = ["voice"] },
     { name = "elevenlabs" },
     { name = "honcho-ai" },
-    { name = "lark-oapi" },
-    { name = "matrix-nio", extra = ["e2e"] },
     { name = "mcp" },
     { name = "modal" },
     { name = "numpy" },
@@ -1694,9 +1668,6 @@ dev = [
 dingtalk = [
     { name = "dingtalk-stream" },
 ]
-feishu = [
-    { name = "lark-oapi" },
-]
 homeassistant = [
     { name = "aiohttp" },
 ]
@@ -1762,7 +1733,6 @@ requires-dist = [
     { name = "discord-py", extras = ["voice"], marker = "extra == 'messaging'", specifier = ">=2.7.1,<3" },
     { name = "edge-tts", specifier = ">=7.2.7,<8" },
     { name = "elevenlabs", marker = "extra == 'tts-premium'", specifier = ">=1.0,<2" },
-    { name = "exa-py", specifier = ">=2.9.0,<3" },
     { name = "fal-client", specifier = ">=0.13.1,<1" },
     { name = "fastapi", marker = "extra == 'rl'", specifier = ">=0.104.0,<1" },
     { name = "faster-whisper", specifier = ">=1.0.0,<2" },
@@ -1774,10 +1744,8 @@ requires-dist = [
     { name = "hermes-agent", extras = ["daytona"], marker = "extra == 'all'" },
     { name = "hermes-agent", extras = ["dev"], marker = "extra == 'all'" },
     { name = "hermes-agent", extras = ["dingtalk"], marker = "extra == 'all'" },
-    { name = "hermes-agent", extras = ["feishu"], marker = "extra == 'all'" },
     { name = "hermes-agent", extras = ["homeassistant"], marker = "extra == 'all'" },
     { name = "hermes-agent", extras = ["honcho"], marker = "extra == 'all'" },
-    { name = "hermes-agent", extras = ["matrix"], marker = "extra == 'all'" },
     { name = "hermes-agent", extras = ["mcp"], marker = "extra == 'all'" },
     { name = "hermes-agent", extras = ["messaging"], marker = "extra == 'all'" },
     { name = "hermes-agent", extras = ["modal"], marker = "extra == 'all'" },
@@ -1789,7 +1757,6 @@ requires-dist = [
     { name = "honcho-ai", marker = "extra == 'honcho'", specifier = ">=2.0.1,<3" },
     { name = "httpx", specifier = ">=0.28.1,<1" },
     { name = "jinja2", specifier = ">=3.1.5,<4" },
-    { name = "lark-oapi", marker = "extra == 'feishu'", specifier = ">=1.5.3,<2" },
     { name = "matrix-nio", extras = ["e2e"], marker = "extra == 'matrix'", specifier = ">=0.24.0,<1" },
     { name = "mcp", marker = "extra == 'dev'", specifier = ">=1.2.0,<2" },
     { name = "mcp", marker = "extra == 'mcp'", specifier = ">=1.2.0,<2" },
@@ -1822,7 +1789,7 @@ requires-dist = [
     { name = "wandb", marker = "extra == 'rl'", specifier = ">=0.15.0,<1" },
     { name = "yc-bench", marker = "python_full_version >= '3.12' and extra == 'yc-bench'", git = "https://github.com/collinear-ai/yc-bench.git" },
 ]
-provides-extras = ["modal", "daytona", "dev", "messaging", "cron", "slack", "matrix", "cli", "tts-premium", "voice", "pty", "honcho", "mcp", "homeassistant", "sms", "acp", "dingtalk", "feishu", "rl", "yc-bench", "all"]
+provides-extras = ["modal", "daytona", "dev", "messaging", "cron", "slack", "matrix", "cli", "tts-premium", "voice", "pty", "honcho", "mcp", "homeassistant", "sms", "acp", "dingtalk", "rl", "yc-bench", "all"]
 
 [[package]]
 name = "hf-transfer"
@@ -2300,21 +2267,6 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/0a/dd/8050c947d435c8d4bc94e3252f4d8bb8a76cfb424f043a8680be637a57f1/kiwisolver-1.5.0-pp311-pypy311_pp73-win_amd64.whl", hash = "sha256:59cd8683f575d96df5bb48f6add94afc055012c29e28124fcae2b63661b9efb1", size = 73558, upload-time = "2026-03-09T13:15:52.112Z" },
 ]
 
-[[package]]
-name = "lark-oapi"
-version = "1.5.3"
-source = { registry = "https://pypi.org/simple" }
-dependencies = [
-    { name = "httpx" },
-    { name = "pycryptodome" },
-    { name = "requests" },
-    { name = "requests-toolbelt" },
-    { name = "websockets" },
-]
-wheels = [
-    { url = "https://files.pythonhosted.org/packages/bf/ff/2ece5d735ebfa2af600a53176f2636ae47af2bf934e08effab64f0d1e047/lark_oapi-1.5.3-py3-none-any.whl", hash = "sha256:fda6b32bb38d21b6bdaae94979c600b94c7c521e985adade63a54e4b3e20cc36", size = 6993016, upload-time = "2026-01-27T08:21:49.307Z" },
-]
-
 [[package]]
 name = "latex2sympy2-extended"
 version = "1.11.0"
@@ -4170,18 +4122,6 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/56/5d/c814546c2333ceea4ba42262d8c4d55763003e767fa169adc693bd524478/requests-2.33.0-py3-none-any.whl", hash = "sha256:3324635456fa185245e24865e810cecec7b4caf933d7eb133dcde67d48cee69b", size = 65017, upload-time = "2026-03-25T15:10:40.382Z" },
 ]
 
-[[package]]
-name = "requests-toolbelt"
-version = "1.0.0"
-source = { registry = "https://pypi.org/simple" }
-dependencies = [
-    { name = "requests" },
-]
-sdist = { url = "https://files.pythonhosted.org/packages/f3/61/d7545dafb7ac2230c70d38d31cbfe4cc64f7144dc41f6e4e4b78ecd9f5bb/requests-toolbelt-1.0.0.tar.gz", hash = "sha256:7681a0a3d047012b5bdc0ee37d7f8f07ebe76ab08caeccfc3921ce23c88d5bc6", size = 206888, upload-time = "2023-05-01T04:11:33.229Z" }
-wheels = [
-    { url = "https://files.pythonhosted.org/packages/3f/51/d4db610ef29373b879047326cbf6fa98b6c1969d6f6dc423279de2b1be2c/requests_toolbelt-1.0.0-py2.py3-none-any.whl", hash = "sha256:cccfdd665f0a24fcf4726e690f65639d272bb0637b9b92dfd91a5568ccf6bd06", size = 54481, upload-time = "2023-05-01T04:11:28.427Z" },
-]
-
 [[package]]
 name = "rich"
 version = "14.3.3"

website/docs/developer-guide/context-compression-and-caching.md

@@ -1,72 +1,321 @@
----
-sidebar_position: 6
-title: "Context Compression & Prompt Caching"
-description: "How Hermes compresses long conversations and applies provider-side prompt caching"
----
+# Context Compression and Caching
 
-# Context Compression & Prompt Caching
+Hermes Agent uses a dual compression system and Anthropic prompt caching to
+manage context window usage efficiently across long conversations.
 
-Hermes manages long conversations with two complementary mechanisms:
+Source files: `agent/context_compressor.py`, `agent/prompt_caching.py`,
+`gateway/run.py` (session hygiene), `run_agent.py` (lines 1146-1204)
 
-- prompt caching
-- context compression
 
-Primary files:
+## Dual Compression System
 
-- `agent/prompt_caching.py`
-- `agent/context_compressor.py`
-- `run_agent.py`
+Hermes has two separate compression layers that operate independently:
 
-## Prompt caching
+```
+                     ┌──────────────────────────┐
+  Incoming message   │   Gateway Session Hygiene │  Fires at 85% of context
+  ─────────────────► │   (pre-agent, rough est.) │  Safety net for large sessions
+                     └─────────────┬────────────┘
+                                   │
+                                   ▼
+                     ┌──────────────────────────┐
+                     │   Agent ContextCompressor │  Fires at 50% of context (default)
+                     │   (in-loop, real tokens)  │  Normal context management
+                     └──────────────────────────┘
+```
 
-For Anthropic/native and Claude-via-OpenRouter flows, Hermes applies Anthropic-style cache markers.
+### 1. Gateway Session Hygiene (85% threshold)
 
-Current strategy:
+Located in `gateway/run.py` (around line 2220). This is a **safety net** that
+runs before the agent processes a message. It prevents API failures when sessions
+grow too large between turns (e.g., overnight accumulation in Telegram/Discord).
 
-- cache the system prompt
-- cache the last 3 non-system messages
-- default TTL is 5 minutes unless explicitly extended
+- **Threshold**: Fixed at 85% of model context length
+- **Token source**: Prefers actual API-reported tokens from last turn; falls back
+  to rough character-based estimate (`estimate_messages_tokens_rough`)
+- **Fires**: Only when `len(history) >= 4` and compression is enabled
+- **Purpose**: Catch sessions that escaped the agent's own compressor
 
-This is implemented in `agent/prompt_caching.py`.
+The gateway hygiene threshold is intentionally higher than the agent's compressor.
+Setting it at 50% (same as the agent) caused premature compression on every turn
+in long gateway sessions.
 
-## Why prompt stability matters
+### 2. Agent ContextCompressor (50% threshold, configurable)
 
-Prompt caching only helps when the stable prefix remains stable. That is why Hermes avoids rebuilding or mutating the core system prompt mid-session unless it has to.
+Located in `agent/context_compressor.py`. This is the **primary compression
+system** that runs inside the agent's tool loop with access to accurate,
+API-reported token counts.
 
-## Compression trigger
 
-Hermes can compress context when conversations become large. Configuration defaults live in `config.yaml`, and the compressor also has runtime checks based on actual prompt token counts.
+## Configuration
 
-## Compression algorithm
+All compression settings are read from `config.yaml` under the `compression` key:
 
-The compressor protects:
+```yaml
+compression:
+  enabled: true              # Enable/disable compression (default: true)
+  threshold: 0.50            # Fraction of context window (default: 0.50 = 50%)
+  target_ratio: 0.20         # How much of threshold to keep as tail (default: 0.20)
+  protect_last_n: 20         # Minimum protected tail messages (default: 20)
+  summary_model: null        # Override model for summaries (default: uses auxiliary)
+```
 
-- the first N turns
-- the last N turns
+### Parameter Details
 
-and summarizes the middle section.
+| Parameter | Default | Range | Description |
+|-----------|---------|-------|-------------|
+| `threshold` | `0.50` | 0.0-1.0 | Compression triggers when prompt tokens ≥ `threshold × context_length` |
+| `target_ratio` | `0.20` | 0.10-0.80 | Controls tail protection token budget: `threshold_tokens × target_ratio` |
+| `protect_last_n` | `20` | ≥1 | Minimum number of recent messages always preserved |
+| `protect_first_n` | `3` | (hardcoded) | System prompt + first exchange always preserved |
 
-It also cleans up structural issues such as orphaned tool-call/result pairs so the API never receives invalid conversation structure after compression.
+### Computed Values (for a 200K context model at defaults)
 
-## Pre-compression memory flush
+```
+context_length       = 200,000
+threshold_tokens     = 200,000 × 0.50 = 100,000
+tail_token_budget    = 100,000 × 0.20 = 20,000
+max_summary_tokens   = min(200,000 × 0.05, 12,000) = 10,000
+```
 
-Before compression, Hermes can give the model one last chance to persist memory so facts are not lost when middle turns are summarized away.
 
-## Session lineage after compression
+## Compression Algorithm
 
-Compression can split the session into a new session ID while preserving parent lineage in the state DB.
+The `ContextCompressor.compress()` method follows a 4-phase algorithm:
 
-This lets Hermes continue operating with a smaller active context while retaining a searchable ancestry chain.
+### Phase 1: Prune Old Tool Results (cheap, no LLM call)
 
-## Re-injected state after compression
+Old tool results (>200 chars) outside the protected tail are replaced with:
+```
+[Old tool output cleared to save context space]
+```
 
-After compression, Hermes may re-inject compact operational state such as:
+This is a cheap pre-pass that saves significant tokens from verbose tool
+outputs (file contents, terminal output, search results).
 
-- todo snapshot
-- prior-read-files summary
+### Phase 2: Determine Boundaries
 
-## Related docs
+```
+┌─────────────────────────────────────────────────────────────┐
+│  Message list                                               │
+│                                                             │
+│  [0..2]  ← protect_first_n (system + first exchange)       │
+│  [3..N]  ← middle turns → SUMMARIZED                       │
+│  [N..end] ← tail (by token budget OR protect_last_n)       │
+│                                                             │
+└─────────────────────────────────────────────────────────────┘
+```
 
-- [Prompt Assembly](./prompt-assembly.md)
-- [Session Storage](./session-storage.md)
-- [Agent Loop Internals](./agent-loop.md)
+Tail protection is **token-budget based**: walks backward from the end,
+accumulating tokens until the budget is exhausted. Falls back to the fixed
+`protect_last_n` count if the budget would protect fewer messages.
+
+Boundaries are aligned to avoid splitting tool_call/tool_result groups.
+The `_align_boundary_backward()` method walks past consecutive tool results
+to find the parent assistant message, keeping groups intact.
+
+### Phase 3: Generate Structured Summary
+
+The middle turns are summarized using the auxiliary LLM with a structured
+template:
+
+```
+## Goal
+[What the user is trying to accomplish]
+
+## Constraints & Preferences
+[User preferences, coding style, constraints, important decisions]
+
+## Progress
+### Done
+[Completed work — specific file paths, commands run, results]
+### In Progress
+[Work currently underway]
+### Blocked
+[Any blockers or issues encountered]
+
+## Key Decisions
+[Important technical decisions and why]
+
+## Relevant Files
+[Files read, modified, or created — with brief note on each]
+
+## Next Steps
+[What needs to happen next]
+
+## Critical Context
+[Specific values, error messages, configuration details]
+```
+
+Summary budget scales with the amount of content being compressed:
+- Formula: `content_tokens × 0.20` (the `_SUMMARY_RATIO` constant)
+- Minimum: 2,000 tokens
+- Maximum: `min(context_length × 0.05, 12,000)` tokens
+
+### Phase 4: Assemble Compressed Messages
+
+The compressed message list is:
+1. Head messages (with a note appended to system prompt on first compression)
+2. Summary message (role chosen to avoid consecutive same-role violations)
+3. Tail messages (unmodified)
+
+Orphaned tool_call/tool_result pairs are cleaned up by `_sanitize_tool_pairs()`:
+- Tool results referencing removed calls → removed
+- Tool calls whose results were removed → stub result injected
+
+### Iterative Re-compression
+
+On subsequent compressions, the previous summary is passed to the LLM with
+instructions to **update** it rather than summarize from scratch. This preserves
+information across multiple compactions — items move from "In Progress" to "Done",
+new progress is added, and obsolete information is removed.
+
+The `_previous_summary` field on the compressor instance stores the last summary
+text for this purpose.
+
+
+## Before/After Example
+
+### Before Compression (45 messages, ~95K tokens)
+
+```
+[0] system:    "You are a helpful assistant..." (system prompt)
+[1] user:      "Help me set up a FastAPI project"
+[2] assistant: <tool_call> terminal: mkdir project </tool_call>
+[3] tool:      "directory created"
+[4] assistant: <tool_call> write_file: main.py </tool_call>
+[5] tool:      "file written (2.3KB)"
+    ... 30 more turns of file editing, testing, debugging ...
+[38] assistant: <tool_call> terminal: pytest </tool_call>
+[39] tool:      "8 passed, 2 failed\n..."  (5KB output)
+[40] user:      "Fix the failing tests"
+[41] assistant: <tool_call> read_file: tests/test_api.py </tool_call>
+[42] tool:      "import pytest\n..."  (3KB)
+[43] assistant: "I see the issue with the test fixtures..."
+[44] user:      "Great, also add error handling"
+```
+
+### After Compression (25 messages, ~45K tokens)
+
+```
+[0] system:    "You are a helpful assistant...
+               [Note: Some earlier conversation turns have been compacted...]"
+[1] user:      "Help me set up a FastAPI project"
+[2] assistant: "[CONTEXT COMPACTION] Earlier turns were compacted...
+
+               ## Goal
+               Set up a FastAPI project with tests and error handling
+
+               ## Progress
+               ### Done
+               - Created project structure: main.py, tests/, requirements.txt
+               - Implemented 5 API endpoints in main.py
+               - Wrote 10 test cases in tests/test_api.py
+               - 8/10 tests passing
+
+               ### In Progress
+               - Fixing 2 failing tests (test_create_user, test_delete_user)
+
+               ## Relevant Files
+               - main.py — FastAPI app with 5 endpoints
+               - tests/test_api.py — 10 test cases
+               - requirements.txt — fastapi, pytest, httpx
+
+               ## Next Steps
+               - Fix failing test fixtures
+               - Add error handling"
+[3] user:      "Fix the failing tests"
+[4] assistant: <tool_call> read_file: tests/test_api.py </tool_call>
+[5] tool:      "import pytest\n..."
+[6] assistant: "I see the issue with the test fixtures..."
+[7] user:      "Great, also add error handling"
+```
+
+
+## Prompt Caching (Anthropic)
+
+Source: `agent/prompt_caching.py`
+
+Reduces input token costs by ~75% on multi-turn conversations by caching the
+conversation prefix. Uses Anthropic's `cache_control` breakpoints.
+
+### Strategy: system_and_3
+
+Anthropic allows a maximum of 4 `cache_control` breakpoints per request. Hermes
+uses the "system_and_3" strategy:
+
+```
+Breakpoint 1: System prompt           (stable across all turns)
+Breakpoint 2: 3rd-to-last non-system message  ─┐
+Breakpoint 3: 2nd-to-last non-system message   ├─ Rolling window
+Breakpoint 4: Last non-system message          ─┘
+```
+
+### How It Works
+
+`apply_anthropic_cache_control()` deep-copies the messages and injects
+`cache_control` markers:
+
+```python
+# Cache marker format
+marker = {"type": "ephemeral"}
+# Or for 1-hour TTL:
+marker = {"type": "ephemeral", "ttl": "1h"}
+```
+
+The marker is applied differently based on content type:
+
+| Content Type | Where Marker Goes |
+|-------------|-------------------|
+| String content | Converted to `[{"type": "text", "text": ..., "cache_control": ...}]` |
+| List content | Added to the last element's dict |
+| None/empty | Added as `msg["cache_control"]` |
+| Tool messages | Added as `msg["cache_control"]` (native Anthropic only) |
+
+### Cache-Aware Design Patterns
+
+1. **Stable system prompt**: The system prompt is breakpoint 1 and cached across
+   all turns. Avoid mutating it mid-conversation (compression appends a note
+   only on the first compaction).
+
+2. **Message ordering matters**: Cache hits require prefix matching. Adding or
+   removing messages in the middle invalidates the cache for everything after.
+
+3. **Compression cache interaction**: After compression, the cache is invalidated
+   for the compressed region but the system prompt cache survives. The rolling
+   3-message window re-establishes caching within 1-2 turns.
+
+4. **TTL selection**: Default is `5m` (5 minutes). Use `1h` for long-running
+   sessions where the user takes breaks between turns.
+
+### Enabling Prompt Caching
+
+Prompt caching is automatically enabled when:
+- The model is an Anthropic Claude model (detected by model name)
+- The provider supports `cache_control` (native Anthropic API or OpenRouter)
+
+```yaml
+# config.yaml — TTL is configurable
+model:
+  cache_ttl: "5m"   # "5m" or "1h"
+```
+
+The CLI shows caching status at startup:
+```
+💾 Prompt caching: ENABLED (Claude via OpenRouter, 5m TTL)
+```
+
+
+## Context Pressure Warnings
+
+The agent emits context pressure warnings at 85% of the compression threshold
+(not 85% of context — 85% of the threshold which is itself 50% of context):
+
+```
+⚠️  Context is 85% to compaction threshold (42,500/50,000 tokens)
+```
+
+After compression, if usage drops below 85% of threshold, the warning state
+is cleared. If compression fails to reduce below the warning level (the
+conversation is too dense), the warning persists but compression won't
+re-trigger until the threshold is exceeded again.

website/docs/developer-guide/prompt-assembly.md

@@ -41,6 +41,163 @@ The cached system prompt is assembled in roughly this order:
 
 When `skip_context_files` is set (e.g., subagent delegation), SOUL.md is not loaded and the hardcoded `DEFAULT_AGENT_IDENTITY` is used instead.
 
+### Concrete example: assembled system prompt
+
+Here is a simplified view of what the final system prompt looks like when all layers are present (comments show the source of each section):
+
+```
+# Layer 1: Agent Identity (from ~/.hermes/SOUL.md)
+You are Hermes, an AI assistant created by Nous Research.
+You are an expert software engineer and researcher.
+You value correctness, clarity, and efficiency.
+...
+
+# Layer 2: Tool-aware behavior guidance
+You have persistent memory across sessions. Save durable facts using
+the memory tool: user preferences, environment details, tool quirks,
+and stable conventions. Memory is injected into every turn, so keep
+it compact and focused on facts that will still matter later.
+...
+When the user references something from a past conversation or you
+suspect relevant cross-session context exists, use session_search
+to recall it before asking them to repeat themselves.
+
+# Tool-use enforcement (for GPT/Codex models only)
+You MUST use your tools to take action — do not describe what you
+would do or plan to do without actually doing it.
+...
+
+# Layer 3: Honcho static block (when active)
+[Honcho personality/context data]
+
+# Layer 4: Optional system message (from config or API)
+[User-configured system message override]
+
+# Layer 5: Frozen MEMORY snapshot
+## Persistent Memory
+- User prefers Python 3.12, uses pyproject.toml
+- Default editor is nvim
+- Working on project "atlas" in ~/code/atlas
+- Timezone: US/Pacific
+
+# Layer 6: Frozen USER profile snapshot
+## User Profile
+- Name: Alice
+- GitHub: alice-dev
+
+# Layer 7: Skills index
+## Skills (mandatory)
+Before replying, scan the skills below. If one clearly matches
+your task, load it with skill_view(name) and follow its instructions.
+...
+<available_skills>
+  software-development:
+    - code-review: Structured code review workflow
+    - test-driven-development: TDD methodology
+  research:
+    - arxiv: Search and summarize arXiv papers
+</available_skills>
+
+# Layer 8: Context files (from project directory)
+# Project Context
+The following project context files have been loaded and should be followed:
+
+## AGENTS.md
+This is the atlas project. Use pytest for testing. The main
+entry point is src/atlas/main.py. Always run `make lint` before
+committing.
+
+# Layer 9: Timestamp + session
+Current time: 2026-03-30T14:30:00-07:00
+Session: abc123
+
+# Layer 10: Platform hint
+You are a CLI AI Agent. Try not to use markdown but simple text
+renderable inside a terminal.
+```
+
+## How SOUL.md appears in the prompt
+
+`SOUL.md` lives at `~/.hermes/SOUL.md` and serves as the agent's identity — the very first section of the system prompt. The loading logic in `prompt_builder.py` works as follows:
+
+```python
+# From agent/prompt_builder.py (simplified)
+def load_soul_md() -> Optional[str]:
+    soul_path = get_hermes_home() / "SOUL.md"
+    if not soul_path.exists():
+        return None
+    content = soul_path.read_text(encoding="utf-8").strip()
+    content = _scan_context_content(content, "SOUL.md")  # Security scan
+    content = _truncate_content(content, "SOUL.md")       # Cap at 20k chars
+    return content
+```
+
+When `load_soul_md()` returns content, it replaces the hardcoded `DEFAULT_AGENT_IDENTITY`. The `build_context_files_prompt()` function is then called with `skip_soul=True` to prevent SOUL.md from appearing twice (once as identity, once as a context file).
+
+If `SOUL.md` doesn't exist, the system falls back to:
+
+```
+You are Hermes Agent, an intelligent AI assistant created by Nous Research.
+You are helpful, knowledgeable, and direct. You assist users with a wide
+range of tasks including answering questions, writing and editing code,
+analyzing information, creative work, and executing actions via your tools.
+You communicate clearly, admit uncertainty when appropriate, and prioritize
+being genuinely useful over being verbose unless otherwise directed below.
+Be targeted and efficient in your exploration and investigations.
+```
+
+## How context files are injected
+
+`build_context_files_prompt()` uses a **priority system** — only one project context type is loaded (first match wins):
+
+```python
+# From agent/prompt_builder.py (simplified)
+def build_context_files_prompt(cwd=None, skip_soul=False):
+    cwd_path = Path(cwd).resolve()
+
+    # Priority: first match wins — only ONE project context loaded
+    project_context = (
+        _load_hermes_md(cwd_path)       # 1. .hermes.md / HERMES.md (walks to git root)
+        or _load_agents_md(cwd_path)    # 2. AGENTS.md (cwd only)
+        or _load_claude_md(cwd_path)    # 3. CLAUDE.md (cwd only)
+        or _load_cursorrules(cwd_path)  # 4. .cursorrules / .cursor/rules/*.mdc
+    )
+
+    sections = []
+    if project_context:
+        sections.append(project_context)
+
+    # SOUL.md from HERMES_HOME (independent of project context)
+    if not skip_soul:
+        soul_content = load_soul_md()
+        if soul_content:
+            sections.append(soul_content)
+
+    if not sections:
+        return ""
+
+    return (
+        "# Project Context\n\n"
+        "The following project context files have been loaded "
+        "and should be followed:\n\n"
+        + "\n".join(sections)
+    )
+```
+
+### Context file discovery details
+
+| Priority | Files | Search scope | Notes |
+|----------|-------|-------------|-------|
+| 1 | `.hermes.md`, `HERMES.md` | CWD up to git root | Hermes-native project config |
+| 2 | `AGENTS.md` | CWD only | Common agent instruction file |
+| 3 | `CLAUDE.md` | CWD only | Claude Code compatibility |
+| 4 | `.cursorrules`, `.cursor/rules/*.mdc` | CWD only | Cursor compatibility |
+
+All context files are:
+- **Security scanned** — checked for prompt injection patterns (invisible unicode, "ignore previous instructions", credential exfiltration attempts)
+- **Truncated** — capped at 20,000 characters using 70/20 head/tail ratio with a truncation marker
+- **YAML frontmatter stripped** — `.hermes.md` frontmatter is removed (reserved for future config overrides)
+
 ## API-call-time-only layers
 
 These are intentionally *not* persisted as part of the cached system prompt:

website/docs/developer-guide/session-storage.md

@@ -1,66 +1,388 @@
----
-sidebar_position: 8
-title: "Session Storage"
-description: "How Hermes stores sessions in SQLite, maintains lineage, and exposes recall/search"
----
-
 # Session Storage
 
-Hermes uses a SQLite-backed session store as the main source of truth for historical conversation state.
+Hermes Agent uses a SQLite database (`~/.hermes/state.db`) to persist session
+metadata, full message history, and model configuration across CLI and gateway
+sessions. This replaces the earlier per-session JSONL file approach.
 
-Primary files:
+Source file: `hermes_state.py`
 
-- `hermes_state.py`
-- `gateway/session.py`
-- `tools/session_search_tool.py`
 
-## Main database
+## Architecture Overview
 
-The primary store lives at:
-
-```text
-~/.hermes/state.db
+```
+~/.hermes/state.db (SQLite, WAL mode)
+├── sessions          — Session metadata, token counts, billing
+├── messages          — Full message history per session
+├── messages_fts      — FTS5 virtual table for full-text search
+└── schema_version    — Single-row table tracking migration state
 ```
 
-It contains:
+Key design decisions:
+- **WAL mode** for concurrent readers + one writer (gateway multi-platform)
+- **FTS5 virtual table** for fast text search across all session messages
+- **Session lineage** via `parent_session_id` chains (compression-triggered splits)
+- **Source tagging** (`cli`, `telegram`, `discord`, etc.) for platform filtering
+- Batch runner and RL trajectories are NOT stored here (separate systems)
 
-- sessions
-- messages
-- metadata such as token counts and titles
-- lineage relationships
-- full-text search indexes
 
-## What is stored per session
+## SQLite Schema
 
-Examples of important session metadata:
+### Sessions Table
 
-- session ID
-- source/platform
-- title
-- created/updated timestamps
-- token counts
-- tool call counts
-- stored system prompt snapshot
-- parent session ID after compression splits
+```sql
+CREATE TABLE IF NOT EXISTS sessions (
+    id TEXT PRIMARY KEY,
+    source TEXT NOT NULL,
+    user_id TEXT,
+    model TEXT,
+    model_config TEXT,
+    system_prompt TEXT,
+    parent_session_id TEXT,
+    started_at REAL NOT NULL,
+    ended_at REAL,
+    end_reason TEXT,
+    message_count INTEGER DEFAULT 0,
+    tool_call_count INTEGER DEFAULT 0,
+    input_tokens INTEGER DEFAULT 0,
+    output_tokens INTEGER DEFAULT 0,
+    cache_read_tokens INTEGER DEFAULT 0,
+    cache_write_tokens INTEGER DEFAULT 0,
+    reasoning_tokens INTEGER DEFAULT 0,
+    billing_provider TEXT,
+    billing_base_url TEXT,
+    billing_mode TEXT,
+    estimated_cost_usd REAL,
+    actual_cost_usd REAL,
+    cost_status TEXT,
+    cost_source TEXT,
+    pricing_version TEXT,
+    title TEXT,
+    FOREIGN KEY (parent_session_id) REFERENCES sessions(id)
+);
 
-## Lineage
+CREATE INDEX IF NOT EXISTS idx_sessions_source ON sessions(source);
+CREATE INDEX IF NOT EXISTS idx_sessions_parent ON sessions(parent_session_id);
+CREATE INDEX IF NOT EXISTS idx_sessions_started ON sessions(started_at DESC);
+CREATE UNIQUE INDEX IF NOT EXISTS idx_sessions_title_unique
+    ON sessions(title) WHERE title IS NOT NULL;
+```
 
-When Hermes compresses a conversation, it can continue in a new session ID while preserving ancestry via `parent_session_id`.
+### Messages Table
 
-This means resuming/searching can follow session families instead of treating each compressed shard as unrelated.
+```sql
+CREATE TABLE IF NOT EXISTS messages (
+    id INTEGER PRIMARY KEY AUTOINCREMENT,
+    session_id TEXT NOT NULL REFERENCES sessions(id),
+    role TEXT NOT NULL,
+    content TEXT,
+    tool_call_id TEXT,
+    tool_calls TEXT,
+    tool_name TEXT,
+    timestamp REAL NOT NULL,
+    token_count INTEGER,
+    finish_reason TEXT,
+    reasoning TEXT,
+    reasoning_details TEXT,
+    codex_reasoning_items TEXT
+);
 
-## Gateway vs CLI persistence
+CREATE INDEX IF NOT EXISTS idx_messages_session ON messages(session_id, timestamp);
+```
 
-- CLI uses the state DB directly for resume/history/search
-- gateway keeps active-session mappings and may also maintain additional platform transcript/state files
-- some legacy JSON/JSONL artifacts still exist for compatibility, but SQLite is the main historical store
+Notes:
+- `tool_calls` is stored as a JSON string (serialized list of tool call objects)
+- `reasoning_details` and `codex_reasoning_items` are stored as JSON strings
+- `reasoning` stores the raw reasoning text for providers that expose it
+- Timestamps are Unix epoch floats (`time.time()`)
 
-## Session search
+### FTS5 Full-Text Search
 
-The `session_search` tool uses the session DB's search features to retrieve and summarize relevant past work.
+```sql
+CREATE VIRTUAL TABLE IF NOT EXISTS messages_fts USING fts5(
+    content,
+    content=messages,
+    content_rowid=id
+);
+```
 
-## Related docs
+The FTS5 table is kept in sync via three triggers that fire on INSERT, UPDATE,
+and DELETE of the `messages` table:
 
-- [Gateway Internals](./gateway-internals.md)
-- [Prompt Assembly](./prompt-assembly.md)
-- [Context Compression & Prompt Caching](./context-compression-and-caching.md)
+```sql
+CREATE TRIGGER IF NOT EXISTS messages_fts_insert AFTER INSERT ON messages BEGIN
+    INSERT INTO messages_fts(rowid, content) VALUES (new.id, new.content);
+END;
+
+CREATE TRIGGER IF NOT EXISTS messages_fts_delete AFTER DELETE ON messages BEGIN
+    INSERT INTO messages_fts(messages_fts, rowid, content)
+        VALUES('delete', old.id, old.content);
+END;
+
+CREATE TRIGGER IF NOT EXISTS messages_fts_update AFTER UPDATE ON messages BEGIN
+    INSERT INTO messages_fts(messages_fts, rowid, content)
+        VALUES('delete', old.id, old.content);
+    INSERT INTO messages_fts(rowid, content) VALUES (new.id, new.content);
+END;
+```
+
+
+## Schema Version and Migrations
+
+Current schema version: **6**
+
+The `schema_version` table stores a single integer. On initialization,
+`_init_schema()` checks the current version and applies migrations sequentially:
+
+| Version | Change |
+|---------|--------|
+| 1 | Initial schema (sessions, messages, FTS5) |
+| 2 | Add `finish_reason` column to messages |
+| 3 | Add `title` column to sessions |
+| 4 | Add unique index on `title` (NULLs allowed, non-NULL must be unique) |
+| 5 | Add billing columns: `cache_read_tokens`, `cache_write_tokens`, `reasoning_tokens`, `billing_provider`, `billing_base_url`, `billing_mode`, `estimated_cost_usd`, `actual_cost_usd`, `cost_status`, `cost_source`, `pricing_version` |
+| 6 | Add reasoning columns to messages: `reasoning`, `reasoning_details`, `codex_reasoning_items` |
+
+Each migration uses `ALTER TABLE ADD COLUMN` wrapped in try/except to handle
+the column-already-exists case (idempotent). The version number is bumped after
+each successful migration block.
+
+
+## Write Contention Handling
+
+Multiple hermes processes (gateway + CLI sessions + worktree agents) share one
+`state.db`. The `SessionDB` class handles write contention with:
+
+- **Short SQLite timeout** (1 second) instead of the default 30s
+- **Application-level retry** with random jitter (20-150ms, up to 15 retries)
+- **BEGIN IMMEDIATE** transactions to surface lock contention at transaction start
+- **Periodic WAL checkpoints** every 50 successful writes (PASSIVE mode)
+
+This avoids the "convoy effect" where SQLite's deterministic internal backoff
+causes all competing writers to retry at the same intervals.
+
+```
+_WRITE_MAX_RETRIES = 15
+_WRITE_RETRY_MIN_S = 0.020   # 20ms
+_WRITE_RETRY_MAX_S = 0.150   # 150ms
+_CHECKPOINT_EVERY_N_WRITES = 50
+```
+
+
+## Common Operations
+
+### Initialize
+
+```python
+from hermes_state import SessionDB
+
+db = SessionDB()                           # Default: ~/.hermes/state.db
+db = SessionDB(db_path=Path("/tmp/test.db"))  # Custom path
+```
+
+### Create and Manage Sessions
+
+```python
+# Create a new session
+db.create_session(
+    session_id="sess_abc123",
+    source="cli",
+    model="anthropic/claude-sonnet-4.6",
+    user_id="user_1",
+    parent_session_id=None,  # or previous session ID for lineage
+)
+
+# End a session
+db.end_session("sess_abc123", end_reason="user_exit")
+
+# Reopen a session (clear ended_at/end_reason)
+db.reopen_session("sess_abc123")
+```
+
+### Store Messages
+
+```python
+msg_id = db.append_message(
+    session_id="sess_abc123",
+    role="assistant",
+    content="Here's the answer...",
+    tool_calls=[{"id": "call_1", "function": {"name": "terminal", "arguments": "{}"}}],
+    token_count=150,
+    finish_reason="stop",
+    reasoning="Let me think about this...",
+)
+```
+
+### Retrieve Messages
+
+```python
+# Raw messages with all metadata
+messages = db.get_messages("sess_abc123")
+
+# OpenAI conversation format (for API replay)
+conversation = db.get_messages_as_conversation("sess_abc123")
+# Returns: [{"role": "user", "content": "..."}, {"role": "assistant", ...}]
+```
+
+### Session Titles
+
+```python
+# Set a title (must be unique among non-NULL titles)
+db.set_session_title("sess_abc123", "Fix Docker Build")
+
+# Resolve by title (returns most recent in lineage)
+session_id = db.resolve_session_by_title("Fix Docker Build")
+
+# Auto-generate next title in lineage
+next_title = db.get_next_title_in_lineage("Fix Docker Build")
+# Returns: "Fix Docker Build #2"
+```
+
+
+## Full-Text Search
+
+The `search_messages()` method supports FTS5 query syntax with automatic
+sanitization of user input.
+
+### Basic Search
+
+```python
+results = db.search_messages("docker deployment")
+```
+
+### FTS5 Query Syntax
+
+| Syntax | Example | Meaning |
+|--------|---------|---------|
+| Keywords | `docker deployment` | Both terms (implicit AND) |
+| Quoted phrase | `"exact phrase"` | Exact phrase match |
+| Boolean OR | `docker OR kubernetes` | Either term |
+| Boolean NOT | `python NOT java` | Exclude term |
+| Prefix | `deploy*` | Prefix match |
+
+### Filtered Search
+
+```python
+# Search only CLI sessions
+results = db.search_messages("error", source_filter=["cli"])
+
+# Exclude gateway sessions
+results = db.search_messages("bug", exclude_sources=["telegram", "discord"])
+
+# Search only user messages
+results = db.search_messages("help", role_filter=["user"])
+```
+
+### Search Results Format
+
+Each result includes:
+- `id`, `session_id`, `role`, `timestamp`
+- `snippet` — FTS5-generated snippet with `>>>match<<<` markers
+- `context` — 1 message before and after the match (content truncated to 200 chars)
+- `source`, `model`, `session_started` — from the parent session
+
+The `_sanitize_fts5_query()` method handles edge cases:
+- Strips unmatched quotes and special characters
+- Wraps hyphenated terms in quotes (`chat-send` → `"chat-send"`)
+- Removes dangling boolean operators (`hello AND` → `hello`)
+
+
+## Session Lineage
+
+Sessions can form chains via `parent_session_id`. This happens when context
+compression triggers a session split in the gateway.
+
+### Query: Find Session Lineage
+
+```sql
+-- Find all ancestors of a session
+WITH RECURSIVE lineage AS (
+    SELECT * FROM sessions WHERE id = ?
+    UNION ALL
+    SELECT s.* FROM sessions s
+    JOIN lineage l ON s.id = l.parent_session_id
+)
+SELECT id, title, started_at, parent_session_id FROM lineage;
+
+-- Find all descendants of a session
+WITH RECURSIVE descendants AS (
+    SELECT * FROM sessions WHERE id = ?
+    UNION ALL
+    SELECT s.* FROM sessions s
+    JOIN descendants d ON s.parent_session_id = d.id
+)
+SELECT id, title, started_at FROM descendants;
+```
+
+### Query: Recent Sessions with Preview
+
+```sql
+SELECT s.*,
+    COALESCE(
+        (SELECT SUBSTR(m.content, 1, 63)
+         FROM messages m
+         WHERE m.session_id = s.id AND m.role = 'user' AND m.content IS NOT NULL
+         ORDER BY m.timestamp, m.id LIMIT 1),
+        ''
+    ) AS preview,
+    COALESCE(
+        (SELECT MAX(m2.timestamp) FROM messages m2 WHERE m2.session_id = s.id),
+        s.started_at
+    ) AS last_active
+FROM sessions s
+ORDER BY s.started_at DESC
+LIMIT 20;
+```
+
+### Query: Token Usage Statistics
+
+```sql
+-- Total tokens by model
+SELECT model,
+       COUNT(*) as session_count,
+       SUM(input_tokens) as total_input,
+       SUM(output_tokens) as total_output,
+       SUM(estimated_cost_usd) as total_cost
+FROM sessions
+WHERE model IS NOT NULL
+GROUP BY model
+ORDER BY total_cost DESC;
+
+-- Sessions with highest token usage
+SELECT id, title, model, input_tokens + output_tokens AS total_tokens,
+       estimated_cost_usd
+FROM sessions
+ORDER BY total_tokens DESC
+LIMIT 10;
+```
+
+
+## Export and Cleanup
+
+```python
+# Export a single session with messages
+data = db.export_session("sess_abc123")
+
+# Export all sessions (with messages) as list of dicts
+all_data = db.export_all(source="cli")
+
+# Delete old sessions (only ended sessions)
+deleted_count = db.prune_sessions(older_than_days=90)
+deleted_count = db.prune_sessions(older_than_days=30, source="telegram")
+
+# Clear messages but keep the session record
+db.clear_messages("sess_abc123")
+
+# Delete session and all messages
+db.delete_session("sess_abc123")
+```
+
+
+## Database Location
+
+Default path: `~/.hermes/state.db`
+
+This is derived from `hermes_constants.get_hermes_home()` which resolves to
+`~/.hermes/` by default, or the value of `HERMES_HOME` environment variable.
+
+The database file, WAL file (`state.db-wal`), and shared-memory file
+(`state.db-shm`) are all created in the same directory.

website/docs/developer-guide/tools-runtime.md

@@ -22,6 +22,89 @@ Each tool module calls `registry.register(...)` at import time.
 
 `model_tools.py` is responsible for importing/discovering tool modules and building the schema list used by the model.
 
+### How `registry.register()` works
+
+Every tool file in `tools/` calls `registry.register()` at module level to declare itself. The function signature is:
+
+```python
+registry.register(
+    name="terminal",               # Unique tool name (used in API schemas)
+    toolset="terminal",            # Toolset this tool belongs to
+    schema={...},                  # OpenAI function-calling schema (description, parameters)
+    handler=handle_terminal,       # The function that executes when the tool is called
+    check_fn=check_terminal,       # Optional: returns True/False for availability
+    requires_env=["SOME_VAR"],     # Optional: env vars needed (for UI display)
+    is_async=False,                # Whether the handler is an async coroutine
+    description="Run commands",    # Human-readable description
+    emoji="💻",                    # Emoji for spinner/progress display
+)
+```
+
+Each call creates a `ToolEntry` stored in the singleton `ToolRegistry._tools` dict keyed by tool name. If a name collision occurs across toolsets, a warning is logged and the later registration wins.
+
+### Discovery: `_discover_tools()`
+
+When `model_tools.py` is imported, it calls `_discover_tools()` which imports every tool module in order:
+
+```python
+_modules = [
+    "tools.web_tools",
+    "tools.terminal_tool",
+    "tools.file_tools",
+    "tools.vision_tools",
+    "tools.mixture_of_agents_tool",
+    "tools.image_generation_tool",
+    "tools.skills_tool",
+    "tools.browser_tool",
+    "tools.cronjob_tools",
+    "tools.rl_training_tool",
+    "tools.tts_tool",
+    "tools.todo_tool",
+    "tools.memory_tool",
+    "tools.session_search_tool",
+    "tools.clarify_tool",
+    "tools.code_execution_tool",
+    "tools.delegate_tool",
+    "tools.process_registry",
+    "tools.send_message_tool",
+    "tools.honcho_tools",
+    "tools.homeassistant_tool",
+]
+```
+
+Each import triggers the module's `registry.register()` calls. Errors in optional tools (e.g., missing `fal_client` for image generation) are caught and logged — they don't prevent other tools from loading.
+
+After core tool discovery, MCP tools and plugin tools are also discovered:
+
+1. **MCP tools** — `tools.mcp_tool.discover_mcp_tools()` reads MCP server config and registers tools from external servers.
+2. **Plugin tools** — `hermes_cli.plugins.discover_plugins()` loads user/project/pip plugins that may register additional tools.
+
+## Tool availability checking (`check_fn`)
+
+Each tool can optionally provide a `check_fn` — a callable that returns `True` when the tool is available and `False` otherwise. Typical checks include:
+
+- **API key present** — e.g., `lambda: bool(os.environ.get("SERP_API_KEY"))` for web search
+- **Service running** — e.g., checking if the Honcho server is configured
+- **Binary installed** — e.g., verifying `playwright` is available for browser tools
+
+When `registry.get_definitions()` builds the schema list for the model, it runs each tool's `check_fn()`:
+
+```python
+# Simplified from registry.py
+if entry.check_fn:
+    try:
+        available = bool(entry.check_fn())
+    except Exception:
+        available = False   # Exceptions = unavailable
+    if not available:
+        continue            # Skip this tool entirely
+```
+
+Key behaviors:
+- Check results are **cached per-call** — if multiple tools share the same `check_fn`, it only runs once.
+- Exceptions in `check_fn()` are treated as "unavailable" (fail-safe).
+- The `is_toolset_available()` method checks whether a toolset's `check_fn` passes, used for UI display and toolset resolution.
+
 ## Toolset resolution
 
 Toolsets are named bundles of tools. Hermes resolves them through:
@@ -31,10 +114,108 @@ Toolsets are named bundles of tools. Hermes resolves them through:
 - dynamic MCP toolsets
 - curated special-purpose sets like `hermes-acp`
 
+### How `get_tool_definitions()` filters tools
+
+The main entry point is `model_tools.get_tool_definitions(enabled_toolsets, disabled_toolsets, quiet_mode)`:
+
+1. **If `enabled_toolsets` is provided** — only tools from those toolsets are included. Each toolset name is resolved via `resolve_toolset()` which expands composite toolsets into individual tool names.
+
+2. **If `disabled_toolsets` is provided** — start with ALL toolsets, then subtract the disabled ones.
+
+3. **If neither** — include all known toolsets.
+
+4. **Registry filtering** — the resolved tool name set is passed to `registry.get_definitions()`, which applies `check_fn` filtering and returns OpenAI-format schemas.
+
+5. **Dynamic schema patching** — after filtering, `execute_code` and `browser_navigate` schemas are dynamically adjusted to only reference tools that actually passed filtering (prevents model hallucination of unavailable tools).
+
+### Legacy toolset names
+
+Old toolset names with `_tools` suffixes (e.g., `web_tools`, `terminal_tools`) are mapped to their modern tool names via `_LEGACY_TOOLSET_MAP` for backward compatibility.
+
 ## Dispatch
 
 At runtime, tools are dispatched through the central registry, with agent-loop exceptions for some agent-level tools such as memory/todo/session-search handling.
 
+### Dispatch flow: model tool_call → handler execution
+
+When the model returns a `tool_call`, the flow is:
+
+```
+Model response with tool_call
+    ↓
+run_agent.py agent loop
+    ↓
+model_tools.handle_function_call(name, args, task_id, user_task)
+    ↓
+[Agent-loop tools?] → handled directly by agent loop (todo, memory, session_search, delegate_task)
+    ↓
+[Plugin pre-hook] → invoke_hook("pre_tool_call", ...)
+    ↓
+registry.dispatch(name, args, **kwargs)
+    ↓
+Look up ToolEntry by name
+    ↓
+[Async handler?] → bridge via _run_async()
+[Sync handler?]  → call directly
+    ↓
+Return result string (or JSON error)
+    ↓
+[Plugin post-hook] → invoke_hook("post_tool_call", ...)
+```
+
+### Error wrapping
+
+All tool execution is wrapped in error handling at two levels:
+
+1. **`registry.dispatch()`** — catches any exception from the handler and returns `{"error": "Tool execution failed: ExceptionType: message"}` as JSON.
+
+2. **`handle_function_call()`** — wraps the entire dispatch in a secondary try/except that returns `{"error": "Error executing tool_name: message"}`.
+
+This ensures the model always receives a well-formed JSON string, never an unhandled exception.
+
+### Agent-loop tools
+
+Four tools are intercepted before registry dispatch because they need agent-level state (TodoStore, MemoryStore, etc.):
+
+- `todo` — planning/task tracking
+- `memory` — persistent memory writes
+- `session_search` — cross-session recall
+- `delegate_task` — spawns subagent sessions
+
+These tools' schemas are still registered in the registry (for `get_tool_definitions`), but their handlers return a stub error if dispatch somehow reaches them directly.
+
+### Async bridging
+
+When a tool handler is async, `_run_async()` bridges it to the sync dispatch path:
+
+- **CLI path (no running loop)** — uses a persistent event loop to keep cached async clients alive
+- **Gateway path (running loop)** — spins up a disposable thread with `asyncio.run()`
+- **Worker threads (parallel tools)** — uses per-thread persistent loops stored in thread-local storage
+
+## The DANGEROUS_PATTERNS approval flow
+
+The terminal tool integrates a dangerous-command approval system defined in `tools/approval.py`:
+
+1. **Pattern detection** — `DANGEROUS_PATTERNS` is a list of `(regex, description)` tuples covering destructive operations:
+   - Recursive deletes (`rm -rf`)
+   - Filesystem formatting (`mkfs`, `dd`)
+   - SQL destructive operations (`DROP TABLE`, `DELETE FROM` without `WHERE`)
+   - System config overwrites (`> /etc/`)
+   - Service manipulation (`systemctl stop`)
+   - Remote code execution (`curl | sh`)
+   - Fork bombs, process kills, etc.
+
+2. **Detection** — before executing any terminal command, `detect_dangerous_command(command)` checks against all patterns.
+
+3. **Approval prompt** — if a match is found:
+   - **CLI mode** — an interactive prompt asks the user to approve, deny, or allow permanently
+   - **Gateway mode** — an async approval callback sends the request to the messaging platform
+   - **Smart approval** — optionally, an auxiliary LLM can auto-approve low-risk commands that match patterns (e.g., `rm -rf node_modules/` is safe but matches "recursive delete")
+
+4. **Session state** — approvals are tracked per-session. Once you approve "recursive delete" for a session, subsequent `rm -rf` commands don't re-prompt.
+
+5. **Permanent allowlist** — the "allow permanently" option writes the pattern to `config.yaml`'s `command_allowlist`, persisting across sessions.
+
 ## Terminal/runtime environments
 
 The terminal system supports multiple backends:

website/docs/developer-guide/trajectory-format.md

@@ -1,56 +1,233 @@
----
-sidebar_position: 10
-title: "Trajectories & Training Format"
-description: "How Hermes saves trajectories, normalizes tool calls, and produces training-friendly outputs"
----
+# Trajectory Format
 
-# Trajectories & Training Format
+Hermes Agent saves conversation trajectories in ShareGPT-compatible JSONL format
+for use as training data, debugging artifacts, and reinforcement learning datasets.
 
-Hermes can save conversation trajectories for training, evaluation, and batch data generation workflows.
+Source files: `agent/trajectory.py`, `run_agent.py` (lines 1788-1975), `batch_runner.py`
 
-Primary files:
 
-- `agent/trajectory.py`
-- `run_agent.py`
-- `batch_runner.py`
-- `trajectory_compressor.py`
+## File Naming Convention
 
-## What trajectories are for
+Trajectories are written to files in the current working directory:
 
-Trajectory outputs are used for:
+| File | When |
+|------|------|
+| `trajectory_samples.jsonl` | Conversations that completed successfully (`completed=True`) |
+| `failed_trajectories.jsonl` | Conversations that failed or were interrupted (`completed=False`) |
 
-- SFT data generation
-- debugging agent behavior
-- benchmark/evaluation artifact capture
-- post-processing and compression pipelines
+The batch runner (`batch_runner.py`) writes to a custom output file per batch
+(e.g., `batch_001_output.jsonl`) with additional metadata fields.
 
-## Normalization strategy
+You can override the filename via the `filename` parameter in `save_trajectory()`.
 
-Hermes converts live conversation structure into a training-friendly format.
 
-Important behaviors include:
+## JSONL Entry Format
 
-- representing reasoning in explicit markup
-- converting tool calls into structured XML-like regions for dataset compatibility
-- grouping tool outputs appropriately
-- separating successful and failed trajectories
+Each line in the file is a self-contained JSON object. There are two variants:
 
-## Persistence boundaries
+### CLI/Interactive Format (from `_save_trajectory`)
 
-Trajectory files do **not** blindly mirror all runtime prompt state.
+```json
+{
+  "conversations": [ ... ],
+  "timestamp": "2026-03-30T14:22:31.456789",
+  "model": "anthropic/claude-sonnet-4.6",
+  "completed": true
+}
+```
 
-Some prompt-time-only layers are intentionally excluded from persisted trajectory content so datasets are cleaner and less environment-specific.
+### Batch Runner Format (from `batch_runner.py`)
 
-## Batch runner
+```json
+{
+  "prompt_index": 42,
+  "conversations": [ ... ],
+  "metadata": { "prompt_source": "gsm8k", "difficulty": "hard" },
+  "completed": true,
+  "partial": false,
+  "api_calls": 7,
+  "toolsets_used": ["code_tools", "file_tools"],
+  "tool_stats": {
+    "terminal": {"count": 3, "success": 3, "failure": 0},
+    "read_file": {"count": 2, "success": 2, "failure": 0},
+    "write_file": {"count": 0, "success": 0, "failure": 0}
+  },
+  "tool_error_counts": {
+    "terminal": 0,
+    "read_file": 0,
+    "write_file": 0
+  }
+}
+```
 
-`batch_runner.py` emits richer metadata than single-session trajectory saving, including:
+The `tool_stats` and `tool_error_counts` dictionaries are normalized to include
+ALL possible tools (from `model_tools.TOOL_TO_TOOLSET_MAP`) with zero defaults,
+ensuring consistent schema across entries for HuggingFace dataset loading.
 
-- model/provider metadata
-- toolset info
-- partial/failure markers
-- tool statistics
 
-## Related docs
+## Conversations Array (ShareGPT Format)
 
-- [Environments, Benchmarks & Data Generation](./environments.md)
-- [Agent Loop Internals](./agent-loop.md)
+The `conversations` array uses ShareGPT role conventions:
+
+| API Role | ShareGPT `from` |
+|----------|-----------------|
+| system | `"system"` |
+| user | `"human"` |
+| assistant | `"gpt"` |
+| tool | `"tool"` |
+
+### Complete Example
+
+```json
+{
+  "conversations": [
+    {
+      "from": "system",
+      "value": "You are a function calling AI model. You are provided with function signatures within <tools> </tools> XML tags. You may call one or more functions to assist with the user query. If available tools are not relevant in assisting with user query, just respond in natural conversational language. Don't make assumptions about what values to plug into functions. After calling & executing the functions, you will be provided with function results within <tool_response> </tool_response> XML tags. Here are the available tools:\n<tools>\n[{\"name\": \"terminal\", \"description\": \"Execute shell commands\", \"parameters\": {\"type\": \"object\", \"properties\": {\"command\": {\"type\": \"string\"}}}, \"required\": null}]\n</tools>\nFor each function call return a JSON object, with the following pydantic model json schema for each:\n{'title': 'FunctionCall', 'type': 'object', 'properties': {'name': {'title': 'Name', 'type': 'string'}, 'arguments': {'title': 'Arguments', 'type': 'object'}}, 'required': ['name', 'arguments']}\nEach function call should be enclosed within <tool_call> </tool_call> XML tags.\nExample:\n<tool_call>\n{'name': <function-name>,'arguments': <args-dict>}\n</tool_call>"
+    },
+    {
+      "from": "human",
+      "value": "What Python version is installed?"
+    },
+    {
+      "from": "gpt",
+      "value": "<think>\nThe user wants to know the Python version. I should run python3 --version.\n</think>\n<tool_call>\n{\"name\": \"terminal\", \"arguments\": {\"command\": \"python3 --version\"}}\n</tool_call>"
+    },
+    {
+      "from": "tool",
+      "value": "<tool_response>\n{\"tool_call_id\": \"call_abc123\", \"name\": \"terminal\", \"content\": \"Python 3.11.6\"}\n</tool_response>"
+    },
+    {
+      "from": "gpt",
+      "value": "<think>\nGot the version. I can now answer the user.\n</think>\nPython 3.11.6 is installed on this system."
+    }
+  ],
+  "timestamp": "2026-03-30T14:22:31.456789",
+  "model": "anthropic/claude-sonnet-4.6",
+  "completed": true
+}
+```
+
+
+## Normalization Rules
+
+### Reasoning Content Markup
+
+The trajectory converter normalizes ALL reasoning into `<think>` tags, regardless
+of how the model originally produced it:
+
+1. **Native thinking tokens** (`msg["reasoning"]` field from providers like
+   Anthropic, OpenAI o-series): Wrapped as `<think>\n{reasoning}\n</think>\n`
+   and prepended before the content.
+
+2. **REASONING_SCRATCHPAD XML** (when native thinking is disabled and the model
+   reasons via system-prompt-instructed XML): `<REASONING_SCRATCHPAD>` tags are
+   converted to `<think>` via `convert_scratchpad_to_think()`.
+
+3. **Empty think blocks**: Every `gpt` turn is guaranteed to have a `<think>`
+   block. If no reasoning was produced, an empty block is inserted:
+   `<think>\n</think>\n` — this ensures consistent format for training data.
+
+### Tool Call Normalization
+
+Tool calls from the API format (with `tool_call_id`, function name, arguments as
+JSON string) are converted to XML-wrapped JSON:
+
+```
+<tool_call>
+{"name": "terminal", "arguments": {"command": "ls -la"}}
+</tool_call>
+```
+
+- Arguments are parsed from JSON strings back to objects (not double-encoded)
+- If JSON parsing fails (shouldn't happen — validated during conversation),
+  an empty `{}` is used with a warning logged
+- Multiple tool calls in one assistant turn produce multiple `<tool_call>` blocks
+  in a single `gpt` message
+
+### Tool Response Normalization
+
+All tool results following an assistant message are grouped into a single `tool`
+turn with XML-wrapped JSON responses:
+
+```
+<tool_response>
+{"tool_call_id": "call_abc123", "name": "terminal", "content": "output here"}
+</tool_response>
+```
+
+- If tool content looks like JSON (starts with `{` or `[`), it's parsed so the
+  content field contains a JSON object/array rather than a string
+- Multiple tool results are joined with newlines in one message
+- The tool name is matched by position against the parent assistant's `tool_calls`
+  array
+
+### System Message
+
+The system message is generated at save time (not taken from the conversation).
+It follows the Hermes function-calling prompt template with:
+
+- Preamble explaining the function-calling protocol
+- `<tools>` XML block containing the JSON tool definitions
+- Schema reference for `FunctionCall` objects
+- `<tool_call>` example
+
+Tool definitions include `name`, `description`, `parameters`, and `required`
+(set to `null` to match the canonical format).
+
+
+## Loading Trajectories
+
+Trajectories are standard JSONL — load with any JSON-lines reader:
+
+```python
+import json
+
+def load_trajectories(path: str):
+    """Load trajectory entries from a JSONL file."""
+    entries = []
+    with open(path, "r", encoding="utf-8") as f:
+        for line in f:
+            line = line.strip()
+            if line:
+                entries.append(json.loads(line))
+    return entries
+
+# Filter to successful completions only
+successful = [e for e in load_trajectories("trajectory_samples.jsonl")
+              if e.get("completed")]
+
+# Extract just the conversations for training
+training_data = [e["conversations"] for e in successful]
+```
+
+### Loading for HuggingFace Datasets
+
+```python
+from datasets import load_dataset
+
+ds = load_dataset("json", data_files="trajectory_samples.jsonl")
+```
+
+The normalized `tool_stats` schema ensures all entries have the same columns,
+preventing Arrow schema mismatch errors during dataset loading.
+
+
+## Controlling Trajectory Saving
+
+In the CLI, trajectory saving is controlled by:
+
+```yaml
+# config.yaml
+agent:
+  save_trajectories: true  # default: false
+```
+
+Or via the `--save-trajectories` flag. When the agent initializes with
+`save_trajectories=True`, the `_save_trajectory()` method is called at the end
+of each conversation turn.
+
+The batch runner always saves trajectories (that's its primary purpose).
+
+Samples with zero reasoning across all turns are automatically discarded by the
+batch runner to avoid polluting training data with non-reasoning examples.

website/docs/getting-started/quickstart.md

@@ -61,7 +61,7 @@ hermes setup       # Or configure everything at once
 | **Custom Endpoint** | VLLM, SGLang, Ollama, or any OpenAI-compatible API | Set base URL + API key |
 
 :::tip
-You can switch providers at any time with `hermes model` — no code changes, no lock-in. When configuring a custom endpoint, Hermes will prompt for the context window size and auto-detect it when possible. See [Context Length Detection](../user-guide/configuration.md#context-length-detection) for details.
+You can switch providers at any time with `hermes model` — no code changes, no lock-in. When configuring a custom endpoint, Hermes will prompt for the context window size and auto-detect it when possible. See [Context Length Detection](../integrations/providers.md#context-length-detection) for details.
 :::
 
 ## 3. Start Chatting

website/docs/getting-started/updating.md

@@ -20,6 +20,43 @@ This pulls the latest code, updates dependencies, and prompts you to configure a
 `hermes update` automatically detects new configuration options and prompts you to add them. If you skipped that prompt, you can manually run `hermes config check` to see missing options, then `hermes config migrate` to interactively add them.
 :::
 
+### What happens during an update
+
+When you run `hermes update`, the following steps occur:
+
+1. **Git pull** — pulls the latest code from the `main` branch and updates submodules
+2. **Dependency install** — runs `uv pip install -e ".[all]"` to pick up new or changed dependencies
+3. **Config migration** — detects new config options added since your version and prompts you to set them
+4. **Gateway auto-restart** — if the gateway service is running (systemd on Linux, launchd on macOS), it is **automatically restarted** after the update completes so the new code takes effect immediately
+
+Expected output looks like:
+
+```
+$ hermes update
+Updating Hermes Agent...
+📥 Pulling latest code...
+Already up to date.  (or: Updating abc1234..def5678)
+📦 Updating dependencies...
+✅ Dependencies updated
+🔍 Checking for new config options...
+✅ Config is up to date  (or: Found 2 new options — running migration...)
+🔄 Restarting gateway service...
+✅ Gateway restarted
+✅ Hermes Agent updated successfully!
+```
+
+### Checking your current version
+
+```bash
+hermes version
+```
+
+Compare against the latest release at the [GitHub releases page](https://github.com/NousResearch/hermes-agent/releases) or check for available updates:
+
+```bash
+hermes update --check
+```
+
 ### Updating from Messaging Platforms
 
 You can also update directly from Telegram, Discord, Slack, or WhatsApp by sending:
@@ -28,7 +65,7 @@ You can also update directly from Telegram, Discord, Slack, or WhatsApp by sendi
 /update
 ```
 
-This pulls the latest code, updates dependencies, and restarts the gateway.
+This pulls the latest code, updates dependencies, and restarts the gateway. The bot will briefly go offline during the restart (typically 5–15 seconds) and then resume.
 
 ### Manual Update
 
@@ -51,6 +88,57 @@ hermes config check
 hermes config migrate   # Interactively add any missing options
 ```
 
+### Rollback instructions
+
+If an update introduces a problem, you can roll back to a previous version:
+
+```bash
+cd /path/to/hermes-agent
+
+# List recent versions
+git log --oneline -10
+
+# Roll back to a specific commit
+git checkout <commit-hash>
+git submodule update --init --recursive
+uv pip install -e ".[all]"
+
+# Restart the gateway if running
+hermes gateway restart
+```
+
+To roll back to a specific release tag:
+
+```bash
+git checkout v0.6.0
+git submodule update --init --recursive
+uv pip install -e ".[all]"
+```
+
+:::warning
+Rolling back may cause config incompatibilities if new options were added. Run `hermes config check` after rolling back and remove any unrecognized options from `config.yaml` if you encounter errors.
+:::
+
+### Note for Nix users
+
+If you installed via Nix flake, updates are managed through the Nix package manager:
+
+```bash
+# Update the flake input
+nix flake update hermes-agent
+
+# Or rebuild with the latest
+nix profile upgrade hermes-agent
+```
+
+Nix installations are immutable — rollback is handled by Nix's generation system:
+
+```bash
+nix profile rollback
+```
+
+See [Nix Setup](./nix-setup.md) for more details.
+
 ---
 
 ## Uninstalling

website/docs/guides/build-a-hermes-plugin.md

@@ -1,5 +1,8 @@
 ---
-sidebar_position: 10
+sidebar_position: 8
+sidebar_label: "Build a Plugin"
+title: "Build a Hermes Plugin"
+description: "Step-by-step guide to building a complete Hermes plugin with tools, hooks, data files, and skills"
 ---
 
 # Build a Hermes Plugin

website/docs/integrations/index.md

@@ -0,0 +1,82 @@
+---
+title: "Integrations"
+sidebar_label: "Overview"
+sidebar_position: 0
+---
+
+# Integrations
+
+Hermes Agent connects to external systems for AI inference, tool servers, IDE workflows, programmatic access, and more. These integrations extend what Hermes can do and where it can run.
+
+## AI Providers & Routing
+
+Hermes supports multiple AI inference providers out of the box. Use `hermes model` to configure interactively, or set them in `config.yaml`.
+
+- **[AI Providers](/docs/user-guide/features/provider-routing)** — OpenRouter, Anthropic, OpenAI, Google, and any OpenAI-compatible endpoint. Hermes auto-detects capabilities like vision, streaming, and tool use per provider.
+- **[Provider Routing](/docs/user-guide/features/provider-routing)** — Fine-grained control over which underlying providers handle your OpenRouter requests. Optimize for cost, speed, or quality with sorting, whitelists, blacklists, and explicit priority ordering.
+- **[Fallback Providers](/docs/user-guide/features/fallback-providers)** — Automatic failover to backup LLM providers when your primary model encounters errors. Includes primary model fallback and independent auxiliary task fallback for vision, compression, and web extraction.
+
+## Tool Servers (MCP)
+
+- **[MCP Servers](/docs/user-guide/features/mcp)** — Connect Hermes to external tool servers via Model Context Protocol. Access tools from GitHub, databases, file systems, browser stacks, internal APIs, and more without writing native Hermes tools. Supports both stdio and SSE transports, per-server tool filtering, and capability-aware resource/prompt registration.
+
+## Web Search Backends
+
+The `web_search`, `web_extract`, and `web_crawl` tools support four backend providers, configured via `config.yaml` or `hermes tools`:
+
+| Backend | Env Var | Search | Extract | Crawl |
+|---------|---------|--------|---------|-------|
+| **Firecrawl** (default) | `FIRECRAWL_API_KEY` | ✔ | ✔ | ✔ |
+| **Parallel** | `PARALLEL_API_KEY` | ✔ | ✔ | — |
+| **Tavily** | `TAVILY_API_KEY` | ✔ | ✔ | ✔ |
+| **Exa** | `EXA_API_KEY` | ✔ | ✔ | — |
+
+Quick setup example:
+
+```yaml
+web:
+  backend: firecrawl    # firecrawl | parallel | tavily | exa
+```
+
+If `web.backend` is not set, the backend is auto-detected from whichever API key is available. Self-hosted Firecrawl is also supported via `FIRECRAWL_API_URL`.
+
+## Browser Automation
+
+Hermes includes full browser automation with multiple backend options for navigating websites, filling forms, and extracting information:
+
+- **Browserbase** — Managed cloud browsers with anti-bot tooling, CAPTCHA solving, and residential proxies
+- **Browser Use** — Alternative cloud browser provider
+- **Local Chrome via CDP** — Connect to your running Chrome instance using `/browser connect`
+- **Local Chromium** — Headless local browser via the `agent-browser` CLI
+
+See [Browser Automation](/docs/user-guide/features/browser) for setup and usage.
+
+## Voice & TTS Providers
+
+Text-to-speech and speech-to-text across all messaging platforms:
+
+| Provider | Quality | Cost | API Key |
+|----------|---------|------|---------|
+| **Edge TTS** (default) | Good | Free | None needed |
+| **ElevenLabs** | Excellent | Paid | `ELEVENLABS_API_KEY` |
+| **OpenAI TTS** | Good | Paid | `VOICE_TOOLS_OPENAI_KEY` |
+| **NeuTTS** | Good | Free | None needed |
+
+Speech-to-text uses Whisper for voice message transcription on Telegram, Discord, and WhatsApp. See [Voice & TTS](/docs/user-guide/features/tts) and [Voice Mode](/docs/user-guide/features/voice-mode) for details.
+
+## IDE & Editor Integration
+
+- **[IDE Integration (ACP)](/docs/user-guide/features/acp)** — Use Hermes Agent inside ACP-compatible editors such as VS Code, Zed, and JetBrains. Hermes runs as an ACP server, rendering chat messages, tool activity, file diffs, and terminal commands inside your editor.
+
+## Programmatic Access
+
+- **[API Server](/docs/user-guide/features/api-server)** — Expose Hermes as an OpenAI-compatible HTTP endpoint. Any frontend that speaks the OpenAI format — Open WebUI, LobeChat, LibreChat, NextChat, ChatBox — can connect and use Hermes as a backend with its full toolset.
+
+## Memory & Personalization
+
+- **[Honcho Memory](/docs/user-guide/features/honcho)** — AI-native persistent memory for cross-session user modeling and personalization. Honcho adds deep user modeling via dialectic reasoning on top of Hermes's built-in memory system.
+
+## Training & Evaluation
+
+- **[RL Training](/docs/user-guide/features/rl-training)** — Generate trajectory data from agent sessions for reinforcement learning and model fine-tuning.
+- **[Batch Processing](/docs/user-guide/features/batch-processing)** — Run the agent across hundreds of prompts in parallel, generating structured ShareGPT-format trajectory data for training data generation or evaluation.

website/docs/integrations/providers.md

@@ -0,0 +1,643 @@
+---
+title: "AI Providers"
+sidebar_label: "AI Providers"
+sidebar_position: 1
+---
+
+# AI Providers
+
+This page covers setting up inference providers for Hermes Agent — from cloud APIs like OpenRouter and Anthropic, to self-hosted endpoints like Ollama and vLLM, to advanced routing and fallback configurations. You need at least one provider configured to use Hermes.
+
+## Inference Providers
+
+You need at least one way to connect to an LLM. Use `hermes model` to switch providers and models interactively, or configure directly:
+
+| Provider | Setup |
+|----------|-------|
+| **Nous Portal** | `hermes model` (OAuth, subscription-based) |
+| **OpenAI Codex** | `hermes model` (ChatGPT OAuth, uses Codex models) |
+| **GitHub Copilot** | `hermes model` (OAuth device code flow, `COPILOT_GITHUB_TOKEN`, `GH_TOKEN`, or `gh auth token`) |
+| **GitHub Copilot ACP** | `hermes model` (spawns local `copilot --acp --stdio`) |
+| **Anthropic** | `hermes model` (Claude Pro/Max via Claude Code auth, Anthropic API key, or manual setup-token) |
+| **OpenRouter** | `OPENROUTER_API_KEY` in `~/.hermes/.env` |
+| **AI Gateway** | `AI_GATEWAY_API_KEY` in `~/.hermes/.env` (provider: `ai-gateway`) |
+| **z.ai / GLM** | `GLM_API_KEY` in `~/.hermes/.env` (provider: `zai`) |
+| **Kimi / Moonshot** | `KIMI_API_KEY` in `~/.hermes/.env` (provider: `kimi-coding`) |
+| **MiniMax** | `MINIMAX_API_KEY` in `~/.hermes/.env` (provider: `minimax`) |
+| **MiniMax China** | `MINIMAX_CN_API_KEY` in `~/.hermes/.env` (provider: `minimax-cn`) |
+| **Alibaba Cloud** | `DASHSCOPE_API_KEY` in `~/.hermes/.env` (provider: `alibaba`, aliases: `dashscope`, `qwen`) |
+| **Kilo Code** | `KILOCODE_API_KEY` in `~/.hermes/.env` (provider: `kilocode`) |
+| **OpenCode Zen** | `OPENCODE_ZEN_API_KEY` in `~/.hermes/.env` (provider: `opencode-zen`) |
+| **OpenCode Go** | `OPENCODE_GO_API_KEY` in `~/.hermes/.env` (provider: `opencode-go`) |
+| **DeepSeek** | `DEEPSEEK_API_KEY` in `~/.hermes/.env` (provider: `deepseek`) |
+| **Hugging Face** | `HF_TOKEN` in `~/.hermes/.env` (provider: `huggingface`, aliases: `hf`) |
+| **Custom Endpoint** | `hermes model` (saved in `config.yaml`) or `OPENAI_BASE_URL` + `OPENAI_API_KEY` in `~/.hermes/.env` |
+
+:::tip Model key alias
+In the `model:` config section, you can use either `default:` or `model:` as the key name for your model ID. Both `model: { default: my-model }` and `model: { model: my-model }` work identically.
+:::
+
+:::info Codex Note
+The OpenAI Codex provider authenticates via device code (open a URL, enter a code). Hermes stores the resulting credentials in its own auth store under `~/.hermes/auth.json` and can import existing Codex CLI credentials from `~/.codex/auth.json` when present. No Codex CLI installation is required.
+:::
+
+:::warning
+Even when using Nous Portal, Codex, or a custom endpoint, some tools (vision, web summarization, MoA) use a separate "auxiliary" model — by default Gemini Flash via OpenRouter. An `OPENROUTER_API_KEY` enables these tools automatically. You can also configure which model and provider these tools use — see [Auxiliary Models](/docs/user-guide/configuration#auxiliary-models).
+:::
+
+### Anthropic (Native)
+
+Use Claude models directly through the Anthropic API — no OpenRouter proxy needed. Supports three auth methods:
+
+```bash
+# With an API key (pay-per-token)
+export ANTHROPIC_API_KEY=***
+hermes chat --provider anthropic --model claude-sonnet-4-6
+
+# Preferred: authenticate through `hermes model`
+# Hermes will use Claude Code's credential store directly when available
+hermes model
+
+# Manual override with a setup-token (fallback / legacy)
+export ANTHROPIC_TOKEN=***  # setup-token or manual OAuth token
+hermes chat --provider anthropic
+
+# Auto-detect Claude Code credentials (if you already use Claude Code)
+hermes chat --provider anthropic  # reads Claude Code credential files automatically
+```
+
+When you choose Anthropic OAuth through `hermes model`, Hermes prefers Claude Code's own credential store over copying the token into `~/.hermes/.env`. That keeps refreshable Claude credentials refreshable.
+
+Or set it permanently:
+```yaml
+model:
+  provider: "anthropic"
+  default: "claude-sonnet-4-6"
+```
+
+:::tip Aliases
+`--provider claude` and `--provider claude-code` also work as shorthand for `--provider anthropic`.
+:::
+
+### GitHub Copilot
+
+Hermes supports GitHub Copilot as a first-class provider with two modes:
+
+**`copilot` — Direct Copilot API** (recommended). Uses your GitHub Copilot subscription to access GPT-5.x, Claude, Gemini, and other models through the Copilot API.
+
+```bash
+hermes chat --provider copilot --model gpt-5.4
+```
+
+**Authentication options** (checked in this order):
+
+1. `COPILOT_GITHUB_TOKEN` environment variable
+2. `GH_TOKEN` environment variable
+3. `GITHUB_TOKEN` environment variable
+4. `gh auth token` CLI fallback
+
+If no token is found, `hermes model` offers an **OAuth device code login** — the same flow used by the Copilot CLI and opencode.
+
+:::warning Token types
+The Copilot API does **not** support classic Personal Access Tokens (`ghp_*`). Supported token types:
+
+| Type | Prefix | How to get |
+|------|--------|------------|
+| OAuth token | `gho_` | `hermes model` → GitHub Copilot → Login with GitHub |
+| Fine-grained PAT | `github_pat_` | GitHub Settings → Developer settings → Fine-grained tokens (needs **Copilot Requests** permission) |
+| GitHub App token | `ghu_` | Via GitHub App installation |
+
+If your `gh auth token` returns a `ghp_*` token, use `hermes model` to authenticate via OAuth instead.
+:::
+
+**API routing**: GPT-5+ models (except `gpt-5-mini`) automatically use the Responses API. All other models (GPT-4o, Claude, Gemini, etc.) use Chat Completions. Models are auto-detected from the live Copilot catalog.
+
+**`copilot-acp` — Copilot ACP agent backend**. Spawns the local Copilot CLI as a subprocess:
+
+```bash
+hermes chat --provider copilot-acp --model copilot-acp
+# Requires the GitHub Copilot CLI in PATH and an existing `copilot login` session
+```
+
+**Permanent config:**
+```yaml
+model:
+  provider: "copilot"
+  default: "gpt-5.4"
+```
+
+| Environment variable | Description |
+|---------------------|-------------|
+| `COPILOT_GITHUB_TOKEN` | GitHub token for Copilot API (first priority) |
+| `HERMES_COPILOT_ACP_COMMAND` | Override the Copilot CLI binary path (default: `copilot`) |
+| `HERMES_COPILOT_ACP_ARGS` | Override ACP args (default: `--acp --stdio`) |
+
+### First-Class Chinese AI Providers
+
+These providers have built-in support with dedicated provider IDs. Set the API key and use `--provider` to select:
+
+```bash
+# z.ai / ZhipuAI GLM
+hermes chat --provider zai --model glm-4-plus
+# Requires: GLM_API_KEY in ~/.hermes/.env
+
+# Kimi / Moonshot AI
+hermes chat --provider kimi-coding --model moonshot-v1-auto
+# Requires: KIMI_API_KEY in ~/.hermes/.env
+
+# MiniMax (global endpoint)
+hermes chat --provider minimax --model MiniMax-M2.7
+# Requires: MINIMAX_API_KEY in ~/.hermes/.env
+
+# MiniMax (China endpoint)
+hermes chat --provider minimax-cn --model MiniMax-M2.7
+# Requires: MINIMAX_CN_API_KEY in ~/.hermes/.env
+
+# Alibaba Cloud / DashScope (Qwen models)
+hermes chat --provider alibaba --model qwen3.5-plus
+# Requires: DASHSCOPE_API_KEY in ~/.hermes/.env
+```
+
+Or set the provider permanently in `config.yaml`:
+```yaml
+model:
+  provider: "zai"       # or: kimi-coding, minimax, minimax-cn, alibaba
+  default: "glm-4-plus"
+```
+
+Base URLs can be overridden with `GLM_BASE_URL`, `KIMI_BASE_URL`, `MINIMAX_BASE_URL`, `MINIMAX_CN_BASE_URL`, or `DASHSCOPE_BASE_URL` environment variables.
+
+### Hugging Face Inference Providers
+
+[Hugging Face Inference Providers](https://huggingface.co/docs/inference-providers) routes to 20+ open models through a unified OpenAI-compatible endpoint (`router.huggingface.co/v1`). Requests are automatically routed to the fastest available backend (Groq, Together, SambaNova, etc.) with automatic failover.
+
+```bash
+# Use any available model
+hermes chat --provider huggingface --model Qwen/Qwen3-235B-A22B-Thinking-2507
+# Requires: HF_TOKEN in ~/.hermes/.env
+
+# Short alias
+hermes chat --provider hf --model deepseek-ai/DeepSeek-V3.2
+```
+
+Or set it permanently in `config.yaml`:
+```yaml
+model:
+  provider: "huggingface"
+  default: "Qwen/Qwen3-235B-A22B-Thinking-2507"
+```
+
+Get your token at [huggingface.co/settings/tokens](https://huggingface.co/settings/tokens) — make sure to enable the "Make calls to Inference Providers" permission. Free tier included ($0.10/month credit, no markup on provider rates).
+
+You can append routing suffixes to model names: `:fastest` (default), `:cheapest`, or `:provider_name` to force a specific backend.
+
+The base URL can be overridden with `HF_BASE_URL`.
+
+## Custom & Self-Hosted LLM Providers
+
+Hermes Agent works with **any OpenAI-compatible API endpoint**. If a server implements `/v1/chat/completions`, you can point Hermes at it. This means you can use local models, GPU inference servers, multi-provider routers, or any third-party API.
+
+### General Setup
+
+Three ways to configure a custom endpoint:
+
+**Interactive setup (recommended):**
+```bash
+hermes model
+# Select "Custom endpoint (self-hosted / VLLM / etc.)"
+# Enter: API base URL, API key, Model name
+```
+
+**Manual config (`config.yaml`):**
+```yaml
+# In ~/.hermes/config.yaml
+model:
+  default: your-model-name
+  provider: custom
+  base_url: http://localhost:8000/v1
+  api_key: your-key-or-leave-empty-for-local
+```
+
+**Environment variables (`.env` file):**
+```bash
+# Add to ~/.hermes/.env
+OPENAI_BASE_URL=http://localhost:8000/v1
+OPENAI_API_KEY=your-key     # Any non-empty string for local servers
+LLM_MODEL=your-model-name
+```
+
+All three approaches end up in the same runtime path. `hermes model` persists provider, model, and base URL to `config.yaml` so later sessions keep using that endpoint even if env vars are not set.
+
+### Switching Models with `/model`
+
+Once a custom endpoint is configured, you can switch models mid-session:
+
+```
+/model custom:qwen-2.5          # Switch to a model on your custom endpoint
+/model custom                    # Auto-detect the model from the endpoint
+/model openrouter:claude-sonnet-4 # Switch back to a cloud provider
+```
+
+If you have **named custom providers** configured (see below), use the triple syntax:
+
+```
+/model custom:local:qwen-2.5    # Use the "local" custom provider with model qwen-2.5
+/model custom:work:llama3       # Use the "work" custom provider with llama3
+```
+
+When switching providers, Hermes persists the base URL and provider to config so the change survives restarts. When switching away from a custom endpoint to a built-in provider, the stale base URL is automatically cleared.
+
+:::tip
+`/model custom` (bare, no model name) queries your endpoint's `/models` API and auto-selects the model if exactly one is loaded. Useful for local servers running a single model.
+:::
+
+Everything below follows this same pattern — just change the URL, key, and model name.
+
+---
+
+### Ollama — Local Models, Zero Config
+
+[Ollama](https://ollama.com/) runs open-weight models locally with one command. Best for: quick local experimentation, privacy-sensitive work, offline use.
+
+```bash
+# Install and run a model
+ollama pull llama3.1:70b
+ollama serve   # Starts on port 11434
+
+# Configure Hermes
+OPENAI_BASE_URL=http://localhost:11434/v1
+OPENAI_API_KEY=ollama           # Any non-empty string
+LLM_MODEL=llama3.1:70b
+```
+
+Ollama's OpenAI-compatible endpoint supports chat completions, streaming, and tool calling (for supported models). No GPU required for smaller models — Ollama handles CPU inference automatically.
+
+:::tip
+List available models with `ollama list`. Pull any model from the [Ollama library](https://ollama.com/library) with `ollama pull <model>`.
+:::
+
+---
+
+### vLLM — High-Performance GPU Inference
+
+[vLLM](https://docs.vllm.ai/) is the standard for production LLM serving. Best for: maximum throughput on GPU hardware, serving large models, continuous batching.
+
+```bash
+# Start vLLM server
+pip install vllm
+vllm serve meta-llama/Llama-3.1-70B-Instruct \
+  --port 8000 \
+  --tensor-parallel-size 2    # Multi-GPU
+
+# Configure Hermes
+OPENAI_BASE_URL=http://localhost:8000/v1
+OPENAI_API_KEY=dummy
+LLM_MODEL=meta-llama/Llama-3.1-70B-Instruct
+```
+
+vLLM supports tool calling, structured output, and multi-modal models. Use `--enable-auto-tool-choice` and `--tool-call-parser hermes` for Hermes-format tool calling with NousResearch models.
+
+---
+
+### SGLang — Fast Serving with RadixAttention
+
+[SGLang](https://github.com/sgl-project/sglang) is an alternative to vLLM with RadixAttention for KV cache reuse. Best for: multi-turn conversations (prefix caching), constrained decoding, structured output.
+
+```bash
+# Start SGLang server
+pip install "sglang[all]"
+python -m sglang.launch_server \
+  --model meta-llama/Llama-3.1-70B-Instruct \
+  --port 8000 \
+  --tp 2
+
+# Configure Hermes
+OPENAI_BASE_URL=http://localhost:8000/v1
+OPENAI_API_KEY=dummy
+LLM_MODEL=meta-llama/Llama-3.1-70B-Instruct
+```
+
+---
+
+### llama.cpp / llama-server — CPU & Metal Inference
+
+[llama.cpp](https://github.com/ggml-org/llama.cpp) runs quantized models on CPU, Apple Silicon (Metal), and consumer GPUs. Best for: running models without a datacenter GPU, Mac users, edge deployment.
+
+```bash
+# Build and start llama-server
+cmake -B build && cmake --build build --config Release
+./build/bin/llama-server \
+  -m models/llama-3.1-8b-instruct-Q4_K_M.gguf \
+  --port 8080 --host 0.0.0.0
+
+# Configure Hermes
+OPENAI_BASE_URL=http://localhost:8080/v1
+OPENAI_API_KEY=dummy
+LLM_MODEL=llama-3.1-8b-instruct
+```
+
+:::tip
+Download GGUF models from [Hugging Face](https://huggingface.co/models?library=gguf). Q4_K_M quantization offers the best balance of quality vs. memory usage.
+:::
+
+---
+
+### LiteLLM Proxy — Multi-Provider Gateway
+
+[LiteLLM](https://docs.litellm.ai/) is an OpenAI-compatible proxy that unifies 100+ LLM providers behind a single API. Best for: switching between providers without config changes, load balancing, fallback chains, budget controls.
+
+```bash
+# Install and start
+pip install "litellm[proxy]"
+litellm --model anthropic/claude-sonnet-4 --port 4000
+
+# Or with a config file for multiple models:
+litellm --config litellm_config.yaml --port 4000
+
+# Configure Hermes
+OPENAI_BASE_URL=http://localhost:4000/v1
+OPENAI_API_KEY=sk-your-litellm-key
+LLM_MODEL=anthropic/claude-sonnet-4
+```
+
+Example `litellm_config.yaml` with fallback:
+```yaml
+model_list:
+  - model_name: "best"
+    litellm_params:
+      model: anthropic/claude-sonnet-4
+      api_key: sk-ant-...
+  - model_name: "best"
+    litellm_params:
+      model: openai/gpt-4o
+      api_key: sk-...
+router_settings:
+  routing_strategy: "latency-based-routing"
+```
+
+---
+
+### ClawRouter — Cost-Optimized Routing
+
+[ClawRouter](https://github.com/BlockRunAI/ClawRouter) by BlockRunAI is a local routing proxy that auto-selects models based on query complexity. It classifies requests across 14 dimensions and routes to the cheapest model that can handle the task. Payment is via USDC cryptocurrency (no API keys).
+
+```bash
+# Install and start
+npx @blockrun/clawrouter    # Starts on port 8402
+
+# Configure Hermes
+OPENAI_BASE_URL=http://localhost:8402/v1
+OPENAI_API_KEY=dummy
+LLM_MODEL=blockrun/auto     # or: blockrun/eco, blockrun/premium, blockrun/agentic
+```
+
+Routing profiles:
+| Profile | Strategy | Savings |
+|---------|----------|---------|
+| `blockrun/auto` | Balanced quality/cost | 74-100% |
+| `blockrun/eco` | Cheapest possible | 95-100% |
+| `blockrun/premium` | Best quality models | 0% |
+| `blockrun/free` | Free models only | 100% |
+| `blockrun/agentic` | Optimized for tool use | varies |
+
+:::note
+ClawRouter requires a USDC-funded wallet on Base or Solana for payment. All requests route through BlockRun's backend API. Run `npx @blockrun/clawrouter doctor` to check wallet status.
+:::
+
+---
+
+### Other Compatible Providers
+
+Any service with an OpenAI-compatible API works. Some popular options:
+
+| Provider | Base URL | Notes |
+|----------|----------|-------|
+| [Together AI](https://together.ai) | `https://api.together.xyz/v1` | Cloud-hosted open models |
+| [Groq](https://groq.com) | `https://api.groq.com/openai/v1` | Ultra-fast inference |
+| [DeepSeek](https://deepseek.com) | `https://api.deepseek.com/v1` | DeepSeek models |
+| [Fireworks AI](https://fireworks.ai) | `https://api.fireworks.ai/inference/v1` | Fast open model hosting |
+| [Cerebras](https://cerebras.ai) | `https://api.cerebras.ai/v1` | Wafer-scale chip inference |
+| [Mistral AI](https://mistral.ai) | `https://api.mistral.ai/v1` | Mistral models |
+| [OpenAI](https://openai.com) | `https://api.openai.com/v1` | Direct OpenAI access |
+| [Azure OpenAI](https://azure.microsoft.com) | `https://YOUR.openai.azure.com/` | Enterprise OpenAI |
+| [LocalAI](https://localai.io) | `http://localhost:8080/v1` | Self-hosted, multi-model |
+| [Jan](https://jan.ai) | `http://localhost:1337/v1` | Desktop app with local models |
+
+```bash
+# Example: Together AI
+OPENAI_BASE_URL=https://api.together.xyz/v1
+OPENAI_API_KEY=your-together-key
+LLM_MODEL=meta-llama/Llama-3.1-70B-Instruct-Turbo
+```
+
+---
+
+### Context Length Detection
+
+Hermes uses a multi-source resolution chain to detect the correct context window for your model and provider:
+
+1. **Config override** — `model.context_length` in config.yaml (highest priority)
+2. **Custom provider per-model** — `custom_providers[].models.<id>.context_length`
+3. **Persistent cache** — previously discovered values (survives restarts)
+4. **Endpoint `/models`** — queries your server's API (local/custom endpoints)
+5. **Anthropic `/v1/models`** — queries Anthropic's API for `max_input_tokens` (API-key users only)
+6. **OpenRouter API** — live model metadata from OpenRouter
+7. **Nous Portal** — suffix-matches Nous model IDs against OpenRouter metadata
+8. **[models.dev](https://models.dev)** — community-maintained registry with provider-specific context lengths for 3800+ models across 100+ providers
+9. **Fallback defaults** — broad model family patterns (128K default)
+
+For most setups this works out of the box. The system is provider-aware — the same model can have different context limits depending on who serves it (e.g., `claude-opus-4.6` is 1M on Anthropic direct but 128K on GitHub Copilot).
+
+To set the context length explicitly, add `context_length` to your model config:
+
+```yaml
+model:
+  default: "qwen3.5:9b"
+  base_url: "http://localhost:8080/v1"
+  context_length: 131072  # tokens
+```
+
+For custom endpoints, you can also set context length per model:
+
+```yaml
+custom_providers:
+  - name: "My Local LLM"
+    base_url: "http://localhost:11434/v1"
+    models:
+      qwen3.5:27b:
+        context_length: 32768
+      deepseek-r1:70b:
+        context_length: 65536
+```
+
+`hermes model` will prompt for context length when configuring a custom endpoint. Leave it blank for auto-detection.
+
+:::tip When to set this manually
+- You're using Ollama with a custom `num_ctx` that's lower than the model's maximum
+- You want to limit context below the model's maximum (e.g., 8k on a 128k model to save VRAM)
+- You're running behind a proxy that doesn't expose `/v1/models`
+:::
+
+---
+
+### Named Custom Providers
+
+If you work with multiple custom endpoints (e.g., a local dev server and a remote GPU server), you can define them as named custom providers in `config.yaml`:
+
+```yaml
+custom_providers:
+  - name: local
+    base_url: http://localhost:8080/v1
+    # api_key omitted — Hermes uses "no-key-required" for keyless local servers
+  - name: work
+    base_url: https://gpu-server.internal.corp/v1
+    api_key: corp-api-key
+    api_mode: chat_completions   # optional, auto-detected from URL
+  - name: anthropic-proxy
+    base_url: https://proxy.example.com/anthropic
+    api_key: proxy-key
+    api_mode: anthropic_messages  # for Anthropic-compatible proxies
+```
+
+Switch between them mid-session with the triple syntax:
+
+```
+/model custom:local:qwen-2.5       # Use the "local" endpoint with qwen-2.5
+/model custom:work:llama3-70b      # Use the "work" endpoint with llama3-70b
+/model custom:anthropic-proxy:claude-sonnet-4  # Use the proxy
+```
+
+You can also select named custom providers from the interactive `hermes model` menu.
+
+---
+
+### Choosing the Right Setup
+
+| Use Case | Recommended |
+|----------|-------------|
+| **Just want it to work** | OpenRouter (default) or Nous Portal |
+| **Local models, easy setup** | Ollama |
+| **Production GPU serving** | vLLM or SGLang |
+| **Mac / no GPU** | Ollama or llama.cpp |
+| **Multi-provider routing** | LiteLLM Proxy or OpenRouter |
+| **Cost optimization** | ClawRouter or OpenRouter with `sort: "price"` |
+| **Maximum privacy** | Ollama, vLLM, or llama.cpp (fully local) |
+| **Enterprise / Azure** | Azure OpenAI with custom endpoint |
+| **Chinese AI models** | z.ai (GLM), Kimi/Moonshot, or MiniMax (first-class providers) |
+
+:::tip
+You can switch between providers at any time with `hermes model` — no restart required. Your conversation history, memory, and skills carry over regardless of which provider you use.
+:::
+
+## Optional API Keys
+
+| Feature | Provider | Env Variable |
+|---------|----------|--------------|
+| Web scraping | [Firecrawl](https://firecrawl.dev/) | `FIRECRAWL_API_KEY`, `FIRECRAWL_API_URL` |
+| Browser automation | [Browserbase](https://browserbase.com/) | `BROWSERBASE_API_KEY`, `BROWSERBASE_PROJECT_ID` |
+| Image generation | [FAL](https://fal.ai/) | `FAL_KEY` |
+| Premium TTS voices | [ElevenLabs](https://elevenlabs.io/) | `ELEVENLABS_API_KEY` |
+| OpenAI TTS + voice transcription | [OpenAI](https://platform.openai.com/api-keys) | `VOICE_TOOLS_OPENAI_KEY` |
+| RL Training | [Tinker](https://tinker-console.thinkingmachines.ai/) + [WandB](https://wandb.ai/) | `TINKER_API_KEY`, `WANDB_API_KEY` |
+| Cross-session user modeling | [Honcho](https://honcho.dev/) | `HONCHO_API_KEY` |
+
+### Self-Hosting Firecrawl
+
+By default, Hermes uses the [Firecrawl cloud API](https://firecrawl.dev/) for web search and scraping. If you prefer to run Firecrawl locally, you can point Hermes at a self-hosted instance instead. See Firecrawl's [SELF_HOST.md](https://github.com/firecrawl/firecrawl/blob/main/SELF_HOST.md) for complete setup instructions.
+
+**What you get:** No API key required, no rate limits, no per-page costs, full data sovereignty.
+
+**What you lose:** The cloud version uses Firecrawl's proprietary "Fire-engine" for advanced anti-bot bypassing (Cloudflare, CAPTCHAs, IP rotation). Self-hosted uses basic fetch + Playwright, so some protected sites may fail. Search uses DuckDuckGo instead of Google.
+
+**Setup:**
+
+1. Clone and start the Firecrawl Docker stack (5 containers: API, Playwright, Redis, RabbitMQ, PostgreSQL — requires ~4-8 GB RAM):
+   ```bash
+   git clone https://github.com/firecrawl/firecrawl
+   cd firecrawl
+   # In .env, set: USE_DB_AUTHENTICATION=false, HOST=0.0.0.0, PORT=3002
+   docker compose up -d
+   ```
+
+2. Point Hermes at your instance (no API key needed):
+   ```bash
+   hermes config set FIRECRAWL_API_URL http://localhost:3002
+   ```
+
+You can also set both `FIRECRAWL_API_KEY` and `FIRECRAWL_API_URL` if your self-hosted instance has authentication enabled.
+
+## OpenRouter Provider Routing
+
+When using OpenRouter, you can control how requests are routed across providers. Add a `provider_routing` section to `~/.hermes/config.yaml`:
+
+```yaml
+provider_routing:
+  sort: "throughput"          # "price" (default), "throughput", or "latency"
+  # only: ["anthropic"]      # Only use these providers
+  # ignore: ["deepinfra"]    # Skip these providers
+  # order: ["anthropic", "google"]  # Try providers in this order
+  # require_parameters: true  # Only use providers that support all request params
+  # data_collection: "deny"   # Exclude providers that may store/train on data
+```
+
+**Shortcuts:** Append `:nitro` to any model name for throughput sorting (e.g., `anthropic/claude-sonnet-4:nitro`), or `:floor` for price sorting.
+
+## Fallback Model
+
+Configure a backup provider:model that Hermes switches to automatically when your primary model fails (rate limits, server errors, auth failures):
+
+```yaml
+fallback_model:
+  provider: openrouter                    # required
+  model: anthropic/claude-sonnet-4        # required
+  # base_url: http://localhost:8000/v1    # optional, for custom endpoints
+  # api_key_env: MY_CUSTOM_KEY           # optional, env var name for custom endpoint API key
+```
+
+When activated, the fallback swaps the model and provider mid-session without losing your conversation. It fires **at most once** per session.
+
+Supported providers: `openrouter`, `nous`, `openai-codex`, `copilot`, `anthropic`, `huggingface`, `zai`, `kimi-coding`, `minimax`, `minimax-cn`, `custom`.
+
+:::tip
+Fallback is configured exclusively through `config.yaml` — there are no environment variables for it. For full details on when it triggers, supported providers, and how it interacts with auxiliary tasks and delegation, see [Fallback Providers](/docs/user-guide/features/fallback-providers).
+:::
+
+## Smart Model Routing
+
+Optional cheap-vs-strong routing lets Hermes keep your main model for complex work while sending very short/simple turns to a cheaper model.
+
+```yaml
+smart_model_routing:
+  enabled: true
+  max_simple_chars: 160
+  max_simple_words: 28
+  cheap_model:
+    provider: openrouter
+    model: google/gemini-2.5-flash
+    # base_url: http://localhost:8000/v1  # optional custom endpoint
+    # api_key_env: MY_CUSTOM_KEY          # optional env var name for that endpoint's API key
+```
+
+How it works:
+- If a turn is short, single-line, and does not look code/tool/debug heavy, Hermes may route it to `cheap_model`
+- If the turn looks complex, Hermes stays on your primary model/provider
+- If the cheap route cannot be resolved cleanly, Hermes falls back to the primary model automatically
+
+This is intentionally conservative. It is meant for quick, low-stakes turns like:
+- short factual questions
+- quick rewrites
+- lightweight summaries
+
+It will avoid routing prompts that look like:
+- coding/debugging work
+- tool-heavy requests
+- long or multi-line analysis asks
+
+Use this when you want lower latency or cost without fully changing your default model.
+
+---
+
+## See Also
+
+- [Configuration](/docs/user-guide/configuration) — General configuration (directory structure, config precedence, terminal backends, memory, compression, and more)
+- [Environment Variables](/docs/reference/environment-variables) — Complete reference of all environment variables

website/docs/reference/faq.md

@@ -254,7 +254,7 @@ custom_providers:
         context_length: 32768
 ```
 
-See [Context Length Detection](../user-guide/configuration.md#context-length-detection) for how auto-detection works and all override options.
+See [Context Length Detection](../integrations/providers.md#context-length-detection) for how auto-detection works and all override options.
 
 ---
 

website/docs/user-guide/checkpoints-and-rollback.md

@@ -1,5 +1,6 @@
 ---
 sidebar_position: 8
+sidebar_label: "Checkpoints & Rollback"
 title: "Checkpoints and /rollback"
 description: "Filesystem safety nets for destructive operations using shadow git repos and automatic snapshots"
 ---

website/docs/user-guide/configuration.md

@@ -71,632 +71,7 @@ delegation:
 
 Multiple references in a single value work: `url: "${HOST}:${PORT}"`. If a referenced variable is not set, the placeholder is kept verbatim (`${UNDEFINED_VAR}` stays as-is). Only the `${VAR}` syntax is supported — bare `$VAR` is not expanded.
 
-## Inference Providers
-
-You need at least one way to connect to an LLM. Use `hermes model` to switch providers and models interactively, or configure directly:
-
-| Provider | Setup |
-|----------|-------|
-| **Nous Portal** | `hermes model` (OAuth, subscription-based) |
-| **OpenAI Codex** | `hermes model` (ChatGPT OAuth, uses Codex models) |
-| **GitHub Copilot** | `hermes model` (OAuth device code flow, `COPILOT_GITHUB_TOKEN`, `GH_TOKEN`, or `gh auth token`) |
-| **GitHub Copilot ACP** | `hermes model` (spawns local `copilot --acp --stdio`) |
-| **Anthropic** | `hermes model` (Claude Pro/Max via Claude Code auth, Anthropic API key, or manual setup-token) |
-| **OpenRouter** | `OPENROUTER_API_KEY` in `~/.hermes/.env` |
-| **AI Gateway** | `AI_GATEWAY_API_KEY` in `~/.hermes/.env` (provider: `ai-gateway`) |
-| **z.ai / GLM** | `GLM_API_KEY` in `~/.hermes/.env` (provider: `zai`) |
-| **Kimi / Moonshot** | `KIMI_API_KEY` in `~/.hermes/.env` (provider: `kimi-coding`) |
-| **MiniMax** | `MINIMAX_API_KEY` in `~/.hermes/.env` (provider: `minimax`) |
-| **MiniMax China** | `MINIMAX_CN_API_KEY` in `~/.hermes/.env` (provider: `minimax-cn`) |
-| **Alibaba Cloud** | `DASHSCOPE_API_KEY` in `~/.hermes/.env` (provider: `alibaba`, aliases: `dashscope`, `qwen`) |
-| **Kilo Code** | `KILOCODE_API_KEY` in `~/.hermes/.env` (provider: `kilocode`) |
-| **OpenCode Zen** | `OPENCODE_ZEN_API_KEY` in `~/.hermes/.env` (provider: `opencode-zen`) |
-| **OpenCode Go** | `OPENCODE_GO_API_KEY` in `~/.hermes/.env` (provider: `opencode-go`) |
-| **DeepSeek** | `DEEPSEEK_API_KEY` in `~/.hermes/.env` (provider: `deepseek`) |
-| **Hugging Face** | `HF_TOKEN` in `~/.hermes/.env` (provider: `huggingface`, aliases: `hf`) |
-| **Custom Endpoint** | `hermes model` (saved in `config.yaml`) or `OPENAI_BASE_URL` + `OPENAI_API_KEY` in `~/.hermes/.env` |
-
-:::tip Model key alias
-In the `model:` config section, you can use either `default:` or `model:` as the key name for your model ID. Both `model: { default: my-model }` and `model: { model: my-model }` work identically.
-:::
-
-:::info Codex Note
-The OpenAI Codex provider authenticates via device code (open a URL, enter a code). Hermes stores the resulting credentials in its own auth store under `~/.hermes/auth.json` and can import existing Codex CLI credentials from `~/.codex/auth.json` when present. No Codex CLI installation is required.
-:::
-
-:::warning
-Even when using Nous Portal, Codex, or a custom endpoint, some tools (vision, web summarization, MoA) use a separate "auxiliary" model — by default Gemini Flash via OpenRouter. An `OPENROUTER_API_KEY` enables these tools automatically. You can also configure which model and provider these tools use — see [Auxiliary Models](#auxiliary-models) below.
-:::
-
-### Anthropic (Native)
-
-Use Claude models directly through the Anthropic API — no OpenRouter proxy needed. Supports three auth methods:
-
-```bash
-# With an API key (pay-per-token)
-export ANTHROPIC_API_KEY=***
-hermes chat --provider anthropic --model claude-sonnet-4-6
-
-# Preferred: authenticate through `hermes model`
-# Hermes will use Claude Code's credential store directly when available
-hermes model
-
-# Manual override with a setup-token (fallback / legacy)
-export ANTHROPIC_TOKEN=***  # setup-token or manual OAuth token
-hermes chat --provider anthropic
-
-# Auto-detect Claude Code credentials (if you already use Claude Code)
-hermes chat --provider anthropic  # reads Claude Code credential files automatically
-```
-
-When you choose Anthropic OAuth through `hermes model`, Hermes prefers Claude Code's own credential store over copying the token into `~/.hermes/.env`. That keeps refreshable Claude credentials refreshable.
-
-Or set it permanently:
-```yaml
-model:
-  provider: "anthropic"
-  default: "claude-sonnet-4-6"
-```
-
-:::tip Aliases
-`--provider claude` and `--provider claude-code` also work as shorthand for `--provider anthropic`.
-:::
-
-### GitHub Copilot
-
-Hermes supports GitHub Copilot as a first-class provider with two modes:
-
-**`copilot` — Direct Copilot API** (recommended). Uses your GitHub Copilot subscription to access GPT-5.x, Claude, Gemini, and other models through the Copilot API.
-
-```bash
-hermes chat --provider copilot --model gpt-5.4
-```
-
-**Authentication options** (checked in this order):
-
-1. `COPILOT_GITHUB_TOKEN` environment variable
-2. `GH_TOKEN` environment variable
-3. `GITHUB_TOKEN` environment variable
-4. `gh auth token` CLI fallback
-
-If no token is found, `hermes model` offers an **OAuth device code login** — the same flow used by the Copilot CLI and opencode.
-
-:::warning Token types
-The Copilot API does **not** support classic Personal Access Tokens (`ghp_*`). Supported token types:
-
-| Type | Prefix | How to get |
-|------|--------|------------|
-| OAuth token | `gho_` | `hermes model` → GitHub Copilot → Login with GitHub |
-| Fine-grained PAT | `github_pat_` | GitHub Settings → Developer settings → Fine-grained tokens (needs **Copilot Requests** permission) |
-| GitHub App token | `ghu_` | Via GitHub App installation |
-
-If your `gh auth token` returns a `ghp_*` token, use `hermes model` to authenticate via OAuth instead.
-:::
-
-**API routing**: GPT-5+ models (except `gpt-5-mini`) automatically use the Responses API. All other models (GPT-4o, Claude, Gemini, etc.) use Chat Completions. Models are auto-detected from the live Copilot catalog.
-
-**`copilot-acp` — Copilot ACP agent backend**. Spawns the local Copilot CLI as a subprocess:
-
-```bash
-hermes chat --provider copilot-acp --model copilot-acp
-# Requires the GitHub Copilot CLI in PATH and an existing `copilot login` session
-```
-
-**Permanent config:**
-```yaml
-model:
-  provider: "copilot"
-  default: "gpt-5.4"
-```
-
-| Environment variable | Description |
-|---------------------|-------------|
-| `COPILOT_GITHUB_TOKEN` | GitHub token for Copilot API (first priority) |
-| `HERMES_COPILOT_ACP_COMMAND` | Override the Copilot CLI binary path (default: `copilot`) |
-| `HERMES_COPILOT_ACP_ARGS` | Override ACP args (default: `--acp --stdio`) |
-
-### First-Class Chinese AI Providers
-
-These providers have built-in support with dedicated provider IDs. Set the API key and use `--provider` to select:
-
-```bash
-# z.ai / ZhipuAI GLM
-hermes chat --provider zai --model glm-4-plus
-# Requires: GLM_API_KEY in ~/.hermes/.env
-
-# Kimi / Moonshot AI
-hermes chat --provider kimi-coding --model moonshot-v1-auto
-# Requires: KIMI_API_KEY in ~/.hermes/.env
-
-# MiniMax (global endpoint)
-hermes chat --provider minimax --model MiniMax-M2.7
-# Requires: MINIMAX_API_KEY in ~/.hermes/.env
-
-# MiniMax (China endpoint)
-hermes chat --provider minimax-cn --model MiniMax-M2.7
-# Requires: MINIMAX_CN_API_KEY in ~/.hermes/.env
-
-# Alibaba Cloud / DashScope (Qwen models)
-hermes chat --provider alibaba --model qwen3.5-plus
-# Requires: DASHSCOPE_API_KEY in ~/.hermes/.env
-```
-
-Or set the provider permanently in `config.yaml`:
-```yaml
-model:
-  provider: "zai"       # or: kimi-coding, minimax, minimax-cn, alibaba
-  default: "glm-4-plus"
-```
-
-Base URLs can be overridden with `GLM_BASE_URL`, `KIMI_BASE_URL`, `MINIMAX_BASE_URL`, `MINIMAX_CN_BASE_URL`, or `DASHSCOPE_BASE_URL` environment variables.
-
-### Hugging Face Inference Providers
-
-[Hugging Face Inference Providers](https://huggingface.co/docs/inference-providers) routes to 20+ open models through a unified OpenAI-compatible endpoint (`router.huggingface.co/v1`). Requests are automatically routed to the fastest available backend (Groq, Together, SambaNova, etc.) with automatic failover.
-
-```bash
-# Use any available model
-hermes chat --provider huggingface --model Qwen/Qwen3-235B-A22B-Thinking-2507
-# Requires: HF_TOKEN in ~/.hermes/.env
-
-# Short alias
-hermes chat --provider hf --model deepseek-ai/DeepSeek-V3.2
-```
-
-Or set it permanently in `config.yaml`:
-```yaml
-model:
-  provider: "huggingface"
-  default: "Qwen/Qwen3-235B-A22B-Thinking-2507"
-```
-
-Get your token at [huggingface.co/settings/tokens](https://huggingface.co/settings/tokens) — make sure to enable the "Make calls to Inference Providers" permission. Free tier included ($0.10/month credit, no markup on provider rates).
-
-You can append routing suffixes to model names: `:fastest` (default), `:cheapest`, or `:provider_name` to force a specific backend.
-
-The base URL can be overridden with `HF_BASE_URL`.
-
-## Custom & Self-Hosted LLM Providers
-
-Hermes Agent works with **any OpenAI-compatible API endpoint**. If a server implements `/v1/chat/completions`, you can point Hermes at it. This means you can use local models, GPU inference servers, multi-provider routers, or any third-party API.
-
-### General Setup
-
-Three ways to configure a custom endpoint:
-
-**Interactive setup (recommended):**
-```bash
-hermes model
-# Select "Custom endpoint (self-hosted / VLLM / etc.)"
-# Enter: API base URL, API key, Model name
-```
-
-**Manual config (`config.yaml`):**
-```yaml
-# In ~/.hermes/config.yaml
-model:
-  default: your-model-name
-  provider: custom
-  base_url: http://localhost:8000/v1
-  api_key: your-key-or-leave-empty-for-local
-```
-
-**Environment variables (`.env` file):**
-```bash
-# Add to ~/.hermes/.env
-OPENAI_BASE_URL=http://localhost:8000/v1
-OPENAI_API_KEY=your-key     # Any non-empty string for local servers
-LLM_MODEL=your-model-name
-```
-
-All three approaches end up in the same runtime path. `hermes model` persists provider, model, and base URL to `config.yaml` so later sessions keep using that endpoint even if env vars are not set.
-
-### Switching Models with `/model`
-
-Once a custom endpoint is configured, you can switch models mid-session:
-
-```
-/model custom:qwen-2.5          # Switch to a model on your custom endpoint
-/model custom                    # Auto-detect the model from the endpoint
-/model openrouter:claude-sonnet-4 # Switch back to a cloud provider
-```
-
-If you have **named custom providers** configured (see below), use the triple syntax:
-
-```
-/model custom:local:qwen-2.5    # Use the "local" custom provider with model qwen-2.5
-/model custom:work:llama3       # Use the "work" custom provider with llama3
-```
-
-When switching providers, Hermes persists the base URL and provider to config so the change survives restarts. When switching away from a custom endpoint to a built-in provider, the stale base URL is automatically cleared.
-
-:::tip
-`/model custom` (bare, no model name) queries your endpoint's `/models` API and auto-selects the model if exactly one is loaded. Useful for local servers running a single model.
-:::
-
-Everything below follows this same pattern — just change the URL, key, and model name.
-
----
-
-### Ollama — Local Models, Zero Config
-
-[Ollama](https://ollama.com/) runs open-weight models locally with one command. Best for: quick local experimentation, privacy-sensitive work, offline use.
-
-```bash
-# Install and run a model
-ollama pull llama3.1:70b
-ollama serve   # Starts on port 11434
-
-# Configure Hermes
-OPENAI_BASE_URL=http://localhost:11434/v1
-OPENAI_API_KEY=ollama           # Any non-empty string
-LLM_MODEL=llama3.1:70b
-```
-
-Ollama's OpenAI-compatible endpoint supports chat completions, streaming, and tool calling (for supported models). No GPU required for smaller models — Ollama handles CPU inference automatically.
-
-:::tip
-List available models with `ollama list`. Pull any model from the [Ollama library](https://ollama.com/library) with `ollama pull <model>`.
-:::
-
----
-
-### vLLM — High-Performance GPU Inference
-
-[vLLM](https://docs.vllm.ai/) is the standard for production LLM serving. Best for: maximum throughput on GPU hardware, serving large models, continuous batching.
-
-```bash
-# Start vLLM server
-pip install vllm
-vllm serve meta-llama/Llama-3.1-70B-Instruct \
-  --port 8000 \
-  --tensor-parallel-size 2    # Multi-GPU
-
-# Configure Hermes
-OPENAI_BASE_URL=http://localhost:8000/v1
-OPENAI_API_KEY=dummy
-LLM_MODEL=meta-llama/Llama-3.1-70B-Instruct
-```
-
-vLLM supports tool calling, structured output, and multi-modal models. Use `--enable-auto-tool-choice` and `--tool-call-parser hermes` for Hermes-format tool calling with NousResearch models.
-
----
-
-### SGLang — Fast Serving with RadixAttention
-
-[SGLang](https://github.com/sgl-project/sglang) is an alternative to vLLM with RadixAttention for KV cache reuse. Best for: multi-turn conversations (prefix caching), constrained decoding, structured output.
-
-```bash
-# Start SGLang server
-pip install "sglang[all]"
-python -m sglang.launch_server \
-  --model meta-llama/Llama-3.1-70B-Instruct \
-  --port 8000 \
-  --tp 2
-
-# Configure Hermes
-OPENAI_BASE_URL=http://localhost:8000/v1
-OPENAI_API_KEY=dummy
-LLM_MODEL=meta-llama/Llama-3.1-70B-Instruct
-```
-
----
-
-### llama.cpp / llama-server — CPU & Metal Inference
-
-[llama.cpp](https://github.com/ggml-org/llama.cpp) runs quantized models on CPU, Apple Silicon (Metal), and consumer GPUs. Best for: running models without a datacenter GPU, Mac users, edge deployment.
-
-```bash
-# Build and start llama-server
-cmake -B build && cmake --build build --config Release
-./build/bin/llama-server \
-  -m models/llama-3.1-8b-instruct-Q4_K_M.gguf \
-  --port 8080 --host 0.0.0.0
-
-# Configure Hermes
-OPENAI_BASE_URL=http://localhost:8080/v1
-OPENAI_API_KEY=dummy
-LLM_MODEL=llama-3.1-8b-instruct
-```
-
-:::tip
-Download GGUF models from [Hugging Face](https://huggingface.co/models?library=gguf). Q4_K_M quantization offers the best balance of quality vs. memory usage.
-:::
-
----
-
-### LiteLLM Proxy — Multi-Provider Gateway
-
-[LiteLLM](https://docs.litellm.ai/) is an OpenAI-compatible proxy that unifies 100+ LLM providers behind a single API. Best for: switching between providers without config changes, load balancing, fallback chains, budget controls.
-
-```bash
-# Install and start
-pip install "litellm[proxy]"
-litellm --model anthropic/claude-sonnet-4 --port 4000
-
-# Or with a config file for multiple models:
-litellm --config litellm_config.yaml --port 4000
-
-# Configure Hermes
-OPENAI_BASE_URL=http://localhost:4000/v1
-OPENAI_API_KEY=sk-your-litellm-key
-LLM_MODEL=anthropic/claude-sonnet-4
-```
-
-Example `litellm_config.yaml` with fallback:
-```yaml
-model_list:
-  - model_name: "best"
-    litellm_params:
-      model: anthropic/claude-sonnet-4
-      api_key: sk-ant-...
-  - model_name: "best"
-    litellm_params:
-      model: openai/gpt-4o
-      api_key: sk-...
-router_settings:
-  routing_strategy: "latency-based-routing"
-```
-
----
-
-### ClawRouter — Cost-Optimized Routing
-
-[ClawRouter](https://github.com/BlockRunAI/ClawRouter) by BlockRunAI is a local routing proxy that auto-selects models based on query complexity. It classifies requests across 14 dimensions and routes to the cheapest model that can handle the task. Payment is via USDC cryptocurrency (no API keys).
-
-```bash
-# Install and start
-npx @blockrun/clawrouter    # Starts on port 8402
-
-# Configure Hermes
-OPENAI_BASE_URL=http://localhost:8402/v1
-OPENAI_API_KEY=dummy
-LLM_MODEL=blockrun/auto     # or: blockrun/eco, blockrun/premium, blockrun/agentic
-```
-
-Routing profiles:
-| Profile | Strategy | Savings |
-|---------|----------|---------|
-| `blockrun/auto` | Balanced quality/cost | 74-100% |
-| `blockrun/eco` | Cheapest possible | 95-100% |
-| `blockrun/premium` | Best quality models | 0% |
-| `blockrun/free` | Free models only | 100% |
-| `blockrun/agentic` | Optimized for tool use | varies |
-
-:::note
-ClawRouter requires a USDC-funded wallet on Base or Solana for payment. All requests route through BlockRun's backend API. Run `npx @blockrun/clawrouter doctor` to check wallet status.
-:::
-
----
-
-### Other Compatible Providers
-
-Any service with an OpenAI-compatible API works. Some popular options:
-
-| Provider | Base URL | Notes |
-|----------|----------|-------|
-| [Together AI](https://together.ai) | `https://api.together.xyz/v1` | Cloud-hosted open models |
-| [Groq](https://groq.com) | `https://api.groq.com/openai/v1` | Ultra-fast inference |
-| [DeepSeek](https://deepseek.com) | `https://api.deepseek.com/v1` | DeepSeek models |
-| [Fireworks AI](https://fireworks.ai) | `https://api.fireworks.ai/inference/v1` | Fast open model hosting |
-| [Cerebras](https://cerebras.ai) | `https://api.cerebras.ai/v1` | Wafer-scale chip inference |
-| [Mistral AI](https://mistral.ai) | `https://api.mistral.ai/v1` | Mistral models |
-| [OpenAI](https://openai.com) | `https://api.openai.com/v1` | Direct OpenAI access |
-| [Azure OpenAI](https://azure.microsoft.com) | `https://YOUR.openai.azure.com/` | Enterprise OpenAI |
-| [LocalAI](https://localai.io) | `http://localhost:8080/v1` | Self-hosted, multi-model |
-| [Jan](https://jan.ai) | `http://localhost:1337/v1` | Desktop app with local models |
-
-```bash
-# Example: Together AI
-OPENAI_BASE_URL=https://api.together.xyz/v1
-OPENAI_API_KEY=your-together-key
-LLM_MODEL=meta-llama/Llama-3.1-70B-Instruct-Turbo
-```
-
----
-
-### Context Length Detection
-
-Hermes uses a multi-source resolution chain to detect the correct context window for your model and provider:
-
-1. **Config override** — `model.context_length` in config.yaml (highest priority)
-2. **Custom provider per-model** — `custom_providers[].models.<id>.context_length`
-3. **Persistent cache** — previously discovered values (survives restarts)
-4. **Endpoint `/models`** — queries your server's API (local/custom endpoints)
-5. **Anthropic `/v1/models`** — queries Anthropic's API for `max_input_tokens` (API-key users only)
-6. **OpenRouter API** — live model metadata from OpenRouter
-7. **Nous Portal** — suffix-matches Nous model IDs against OpenRouter metadata
-8. **[models.dev](https://models.dev)** — community-maintained registry with provider-specific context lengths for 3800+ models across 100+ providers
-9. **Fallback defaults** — broad model family patterns (128K default)
-
-For most setups this works out of the box. The system is provider-aware — the same model can have different context limits depending on who serves it (e.g., `claude-opus-4.6` is 1M on Anthropic direct but 128K on GitHub Copilot).
-
-To set the context length explicitly, add `context_length` to your model config:
-
-```yaml
-model:
-  default: "qwen3.5:9b"
-  base_url: "http://localhost:8080/v1"
-  context_length: 131072  # tokens
-```
-
-For custom endpoints, you can also set context length per model:
-
-```yaml
-custom_providers:
-  - name: "My Local LLM"
-    base_url: "http://localhost:11434/v1"
-    models:
-      qwen3.5:27b:
-        context_length: 32768
-      deepseek-r1:70b:
-        context_length: 65536
-```
-
-`hermes model` will prompt for context length when configuring a custom endpoint. Leave it blank for auto-detection.
-
-:::tip When to set this manually
-- You're using Ollama with a custom `num_ctx` that's lower than the model's maximum
-- You want to limit context below the model's maximum (e.g., 8k on a 128k model to save VRAM)
-- You're running behind a proxy that doesn't expose `/v1/models`
-:::
-
----
-
-### Named Custom Providers
-
-If you work with multiple custom endpoints (e.g., a local dev server and a remote GPU server), you can define them as named custom providers in `config.yaml`:
-
-```yaml
-custom_providers:
-  - name: local
-    base_url: http://localhost:8080/v1
-    # api_key omitted — Hermes uses "no-key-required" for keyless local servers
-  - name: work
-    base_url: https://gpu-server.internal.corp/v1
-    api_key: corp-api-key
-    api_mode: chat_completions   # optional, auto-detected from URL
-  - name: anthropic-proxy
-    base_url: https://proxy.example.com/anthropic
-    api_key: proxy-key
-    api_mode: anthropic_messages  # for Anthropic-compatible proxies
-```
-
-Switch between them mid-session with the triple syntax:
-
-```
-/model custom:local:qwen-2.5       # Use the "local" endpoint with qwen-2.5
-/model custom:work:llama3-70b      # Use the "work" endpoint with llama3-70b
-/model custom:anthropic-proxy:claude-sonnet-4  # Use the proxy
-```
-
-You can also select named custom providers from the interactive `hermes model` menu.
-
----
-
-### Choosing the Right Setup
-
-| Use Case | Recommended |
-|----------|-------------|
-| **Just want it to work** | OpenRouter (default) or Nous Portal |
-| **Local models, easy setup** | Ollama |
-| **Production GPU serving** | vLLM or SGLang |
-| **Mac / no GPU** | Ollama or llama.cpp |
-| **Multi-provider routing** | LiteLLM Proxy or OpenRouter |
-| **Cost optimization** | ClawRouter or OpenRouter with `sort: "price"` |
-| **Maximum privacy** | Ollama, vLLM, or llama.cpp (fully local) |
-| **Enterprise / Azure** | Azure OpenAI with custom endpoint |
-| **Chinese AI models** | z.ai (GLM), Kimi/Moonshot, or MiniMax (first-class providers) |
-
-:::tip
-You can switch between providers at any time with `hermes model` — no restart required. Your conversation history, memory, and skills carry over regardless of which provider you use.
-:::
-
-## Optional API Keys
-
-| Feature | Provider | Env Variable |
-|---------|----------|--------------|
-| Web scraping | [Firecrawl](https://firecrawl.dev/) | `FIRECRAWL_API_KEY`, `FIRECRAWL_API_URL` |
-| Browser automation | [Browserbase](https://browserbase.com/) | `BROWSERBASE_API_KEY`, `BROWSERBASE_PROJECT_ID` |
-| Image generation | [FAL](https://fal.ai/) | `FAL_KEY` |
-| Premium TTS voices | [ElevenLabs](https://elevenlabs.io/) | `ELEVENLABS_API_KEY` |
-| OpenAI TTS + voice transcription | [OpenAI](https://platform.openai.com/api-keys) | `VOICE_TOOLS_OPENAI_KEY` |
-| RL Training | [Tinker](https://tinker-console.thinkingmachines.ai/) + [WandB](https://wandb.ai/) | `TINKER_API_KEY`, `WANDB_API_KEY` |
-| Cross-session user modeling | [Honcho](https://honcho.dev/) | `HONCHO_API_KEY` |
-
-### Self-Hosting Firecrawl
-
-By default, Hermes uses the [Firecrawl cloud API](https://firecrawl.dev/) for web search and scraping. If you prefer to run Firecrawl locally, you can point Hermes at a self-hosted instance instead. See Firecrawl's [SELF_HOST.md](https://github.com/firecrawl/firecrawl/blob/main/SELF_HOST.md) for complete setup instructions.
-
-**What you get:** No API key required, no rate limits, no per-page costs, full data sovereignty.
-
-**What you lose:** The cloud version uses Firecrawl's proprietary "Fire-engine" for advanced anti-bot bypassing (Cloudflare, CAPTCHAs, IP rotation). Self-hosted uses basic fetch + Playwright, so some protected sites may fail. Search uses DuckDuckGo instead of Google.
-
-**Setup:**
-
-1. Clone and start the Firecrawl Docker stack (5 containers: API, Playwright, Redis, RabbitMQ, PostgreSQL — requires ~4-8 GB RAM):
-   ```bash
-   git clone https://github.com/firecrawl/firecrawl
-   cd firecrawl
-   # In .env, set: USE_DB_AUTHENTICATION=false, HOST=0.0.0.0, PORT=3002
-   docker compose up -d
-   ```
-
-2. Point Hermes at your instance (no API key needed):
-   ```bash
-   hermes config set FIRECRAWL_API_URL http://localhost:3002
-   ```
-
-You can also set both `FIRECRAWL_API_KEY` and `FIRECRAWL_API_URL` if your self-hosted instance has authentication enabled.
-
-## OpenRouter Provider Routing
-
-When using OpenRouter, you can control how requests are routed across providers. Add a `provider_routing` section to `~/.hermes/config.yaml`:
-
-```yaml
-provider_routing:
-  sort: "throughput"          # "price" (default), "throughput", or "latency"
-  # only: ["anthropic"]      # Only use these providers
-  # ignore: ["deepinfra"]    # Skip these providers
-  # order: ["anthropic", "google"]  # Try providers in this order
-  # require_parameters: true  # Only use providers that support all request params
-  # data_collection: "deny"   # Exclude providers that may store/train on data
-```
-
-**Shortcuts:** Append `:nitro` to any model name for throughput sorting (e.g., `anthropic/claude-sonnet-4:nitro`), or `:floor` for price sorting.
-
-## Fallback Model
-
-Configure a backup provider:model that Hermes switches to automatically when your primary model fails (rate limits, server errors, auth failures):
-
-```yaml
-fallback_model:
-  provider: openrouter                    # required
-  model: anthropic/claude-sonnet-4        # required
-  # base_url: http://localhost:8000/v1    # optional, for custom endpoints
-  # api_key_env: MY_CUSTOM_KEY           # optional, env var name for custom endpoint API key
-```
-
-When activated, the fallback swaps the model and provider mid-session without losing your conversation. It fires **at most once** per session.
-
-Supported providers: `openrouter`, `nous`, `openai-codex`, `copilot`, `anthropic`, `huggingface`, `zai`, `kimi-coding`, `minimax`, `minimax-cn`, `custom`.
-
-:::tip
-Fallback is configured exclusively through `config.yaml` — there are no environment variables for it. For full details on when it triggers, supported providers, and how it interacts with auxiliary tasks and delegation, see [Fallback Providers](/docs/user-guide/features/fallback-providers).
-:::
-
-## Smart Model Routing
-
-Optional cheap-vs-strong routing lets Hermes keep your main model for complex work while sending very short/simple turns to a cheaper model.
-
-```yaml
-smart_model_routing:
-  enabled: true
-  max_simple_chars: 160
-  max_simple_words: 28
-  cheap_model:
-    provider: openrouter
-    model: google/gemini-2.5-flash
-    # base_url: http://localhost:8000/v1  # optional custom endpoint
-    # api_key_env: MY_CUSTOM_KEY          # optional env var name for that endpoint's API key
-```
-
-How it works:
-- If a turn is short, single-line, and does not look code/tool/debug heavy, Hermes may route it to `cheap_model`
-- If the turn looks complex, Hermes stays on your primary model/provider
-- If the cheap route cannot be resolved cleanly, Hermes falls back to the primary model automatically
-
-This is intentionally conservative. It is meant for quick, low-stakes turns like:
-- short factual questions
-- quick rewrites
-- lightweight summaries
-
-It will avoid routing prompts that look like:
-- coding/debugging work
-- tool-heavy requests
-- long or multi-line analysis asks
-
-Use this when you want lower latency or cost without fully changing your default model.
+For AI provider setup (OpenRouter, Anthropic, Copilot, custom endpoints, self-hosted LLMs, fallback models, etc.), see [AI Providers](/docs/integrations/providers).
 
 ## Terminal Backend Configuration
 
@@ -1192,7 +567,7 @@ Each auxiliary task has a configurable `timeout` (in seconds). Defaults: vision
 :::
 
 :::info
-Context compression has its own top-level `compression:` block with `summary_provider`, `summary_model`, and `summary_base_url` — see [Context Compression](#context-compression) above. The fallback model uses a `fallback_model:` block — see [Fallback Model](#fallback-model) above. All three follow the same provider/model/base_url pattern.
+Context compression has its own top-level `compression:` block with `summary_provider`, `summary_model`, and `summary_base_url` — see [Context Compression](#context-compression) above. The fallback model uses a `fallback_model:` block — see [Fallback Model](/docs/integrations/providers#fallback-model). All three follow the same provider/model/base_url pattern.
 :::
 
 ### Changing the Vision Model
@@ -1485,12 +860,15 @@ When enabled, responses appear token-by-token inside a streaming box. Tool calls
 ```yaml
 streaming:
   enabled: true           # Enable progressive message editing
+  transport: edit         # "edit" (progressive message editing) or "off"
   edit_interval: 0.3      # Seconds between message edits
   buffer_threshold: 40    # Characters before forcing an edit flush
   cursor: " ▉"            # Cursor shown during streaming
 ```
 
-When enabled, the bot sends a message on the first token, then progressively edits it as more tokens arrive. Platforms that don't support message editing (Signal, Email) gracefully skip streaming and deliver the final response normally.
+When enabled, the bot sends a message on the first token, then progressively edits it as more tokens arrive. Platforms that don't support message editing (Signal, Email, Home Assistant) are auto-detected on the first attempt — streaming is gracefully disabled for that session with no flood of messages.
+
+**Overflow handling:** If the streamed text exceeds the platform's message length limit (~4096 chars), the current message is finalized and a new one starts automatically.
 
 :::note
 Streaming is disabled by default. Enable it in `~/.hermes/config.yaml` to try the streaming UX.
@@ -1554,23 +932,6 @@ Usage: type `/status`, `/disk`, `/update`, or `/gpu` in the CLI or any messaging
 - **Type** — only `exec` is supported (runs a shell command); other types show an error
 - **Works everywhere** — CLI, Telegram, Discord, Slack, WhatsApp, Signal, Email, Home Assistant
 
-## Gateway Streaming
-
-Enable progressive token delivery on messaging platforms. When streaming is enabled, responses appear character-by-character in Telegram, Discord, and Slack via message editing, rather than waiting for the full response.
-
-```yaml
-streaming:
-  enabled: false              # Enable streaming token delivery (default: off)
-  transport: edit             # "edit" (progressive message editing) or "off"
-  edit_interval: 0.3          # Min seconds between message edits
-  buffer_threshold: 40        # Characters accumulated before forcing an edit
-  cursor: " ▉"               # Cursor character shown during streaming
-```
-
-**Platform support:** Telegram, Discord, and Slack support edit-based streaming. Platforms that don't support message editing (Signal, Email, Home Assistant) are auto-detected on the first attempt — streaming is gracefully disabled for that session with no flood of messages.
-
-**Overflow handling:** If the streamed text exceeds the platform's message length limit (~4096 chars), the current message is finalized and a new one starts automatically.
-
 ## Human Delay
 
 Simulate human-like response pacing in messaging platforms:
@@ -1725,7 +1086,7 @@ Setting `approvals.mode: off` disables all safety checks for terminal commands.
 
 ## Checkpoints
 
-Automatic filesystem snapshots before destructive file operations. See the [Checkpoints feature page](/docs/user-guide/features/checkpoints) for details.
+Automatic filesystem snapshots before destructive file operations. See the [Checkpoints & Rollback](/docs/user-guide/checkpoints-and-rollback) for details.
 
 ```yaml
 checkpoints:

website/docs/user-guide/docker.md

@@ -1,10 +1,17 @@
+---
+sidebar_position: 7
+title: "Docker"
+description: "Running Hermes Agent in Docker and using Docker as a terminal backend"
+---
+
 # Hermes Agent — Docker
 
-Want to run Hermes Agent, but without installing packages on your host? This'll sort you out.
+There are two distinct ways Docker intersects with Hermes Agent:
 
-This will let you run the agent in a container, with the most relevant modes outlined below.
+1. **Running Hermes IN Docker** — the agent itself runs inside a container (this page's primary focus)
+2. **Docker as a terminal backend** — the agent runs on your host but executes commands inside a Docker sandbox (see [Configuration → terminal.backend](./configuration.md))
 
-The container stores all user data (config, API keys, sessions, skills, memories) in a single directory mounted from the host at `/opt/data`. The image itself is stateless and can be upgraded by pulling a new version without losing any configuration.
+This page covers option 1. The container stores all user data (config, API keys, sessions, skills, memories) in a single directory mounted from the host at `/opt/data`. The image itself is stateless and can be upgraded by pulling a new version without losing any configuration.
 
 ## Quick start
 
@@ -41,6 +48,110 @@ docker run -it --rm \
   nousresearch/hermes-agent
 ```
 
+## Persistent volumes
+
+The `/opt/data` volume is the single source of truth for all Hermes state. It maps to your host's `~/.hermes/` directory and contains:
+
+| Path | Contents |
+|------|----------|
+| `.env` | API keys and secrets |
+| `config.yaml` | All Hermes configuration |
+| `SOUL.md` | Agent personality/identity |
+| `sessions/` | Conversation history |
+| `memories/` | Persistent memory store |
+| `skills/` | Installed skills |
+| `cron/` | Scheduled job definitions |
+| `hooks/` | Event hooks |
+| `logs/` | Runtime logs |
+| `skins/` | Custom CLI skins |
+
+:::warning
+Never run two Hermes containers against the same data directory simultaneously — session files and memory stores are not designed for concurrent access.
+:::
+
+## Environment variable forwarding
+
+API keys are read from `/opt/data/.env` inside the container. You can also pass environment variables directly:
+
+```sh
+docker run -it --rm \
+  -v ~/.hermes:/opt/data \
+  -e ANTHROPIC_API_KEY="sk-ant-..." \
+  -e OPENAI_API_KEY="sk-..." \
+  nousresearch/hermes-agent
+```
+
+Direct `-e` flags override values from `.env`. This is useful for CI/CD or secrets-manager integrations where you don't want keys on disk.
+
+## Docker Compose example
+
+For persistent gateway deployment, a `docker-compose.yaml` is convenient:
+
+```yaml
+version: "3.8"
+services:
+  hermes:
+    image: nousresearch/hermes-agent:latest
+    container_name: hermes
+    restart: unless-stopped
+    command: gateway run
+    volumes:
+      - ~/.hermes:/opt/data
+    # Uncomment to forward specific env vars instead of using .env file:
+    # environment:
+    #   - ANTHROPIC_API_KEY=${ANTHROPIC_API_KEY}
+    #   - OPENAI_API_KEY=${OPENAI_API_KEY}
+    #   - TELEGRAM_BOT_TOKEN=${TELEGRAM_BOT_TOKEN}
+    deploy:
+      resources:
+        limits:
+          memory: 4G
+          cpus: "2.0"
+```
+
+Start with `docker compose up -d` and view logs with `docker compose logs -f hermes`.
+
+## Resource limits
+
+The Hermes container needs moderate resources. Recommended minimums:
+
+| Resource | Minimum | Recommended |
+|----------|---------|-------------|
+| Memory | 1 GB | 2–4 GB |
+| CPU | 1 core | 2 cores |
+| Disk (data volume) | 500 MB | 2+ GB (grows with sessions/skills) |
+
+Browser automation (Playwright/Chromium) is the most memory-hungry feature. If you don't need browser tools, 1 GB is sufficient. With browser tools active, allocate at least 2 GB.
+
+Set limits in Docker:
+
+```sh
+docker run -d \
+  --name hermes \
+  --restart unless-stopped \
+  --memory=4g --cpus=2 \
+  -v ~/.hermes:/opt/data \
+  nousresearch/hermes-agent gateway run
+```
+
+## What the Dockerfile does
+
+The official image is based on `debian:13.4` and includes:
+
+- Python 3 with all Hermes dependencies (`pip install -e ".[all]"`)
+- Node.js + npm (for browser automation and WhatsApp bridge)
+- Playwright with Chromium (`npx playwright install --with-deps chromium`)
+- ripgrep and ffmpeg as system utilities
+- The WhatsApp bridge (`scripts/whatsapp-bridge/`)
+
+The entrypoint script (`docker/entrypoint.sh`) bootstraps the data volume on first run:
+- Creates the directory structure (`sessions/`, `memories/`, `skills/`, etc.)
+- Copies `.env.example` → `.env` if no `.env` exists
+- Copies default `config.yaml` if missing
+- Copies default `SOUL.md` if missing
+- Syncs bundled skills using a manifest-based approach (preserves user edits)
+- Then runs `hermes` with whatever arguments you pass
+
 ## Upgrading
 
 Pull the latest image and recreate the container. Your data directory is untouched.
@@ -52,7 +163,14 @@ docker run -d \
   --name hermes \
   --restart unless-stopped \
   -v ~/.hermes:/opt/data \
-  nousresearch/hermes-agent
+  nousresearch/hermes-agent gateway run
+```
+
+Or with Docker Compose:
+
+```sh
+docker compose pull
+docker compose up -d
 ```
 
 ## Skills and credential files
@@ -60,3 +178,47 @@ docker run -d \
 When using Docker as the execution environment (not the methods above, but when the agent runs commands inside a Docker sandbox), Hermes automatically bind-mounts the skills directory (`~/.hermes/skills/`) and any credential files declared by skills into the container as read-only volumes. This means skill scripts, templates, and references are available inside the sandbox without manual configuration.
 
 The same syncing happens for SSH and Modal backends — skills and credential files are uploaded via rsync or the Modal mount API before each command.
+
+## Troubleshooting
+
+### Container exits immediately
+
+Check logs: `docker logs hermes`. Common causes:
+- Missing or invalid `.env` file — run interactively first to complete setup
+- Port conflicts if running with exposed ports
+
+### "Permission denied" errors
+
+The container runs as root by default. If your host `~/.hermes/` was created by a non-root user, permissions should work. If you get errors, ensure the data directory is writable:
+
+```sh
+chmod -R 755 ~/.hermes
+```
+
+### Browser tools not working
+
+Playwright needs shared memory. Add `--shm-size=1g` to your Docker run command:
+
+```sh
+docker run -d \
+  --name hermes \
+  --shm-size=1g \
+  -v ~/.hermes:/opt/data \
+  nousresearch/hermes-agent gateway run
+```
+
+### Gateway not reconnecting after network issues
+
+The `--restart unless-stopped` flag handles most transient failures. If the gateway is stuck, restart the container:
+
+```sh
+docker restart hermes
+```
+
+### Checking container health
+
+```sh
+docker logs --tail 50 hermes          # Recent logs
+docker exec hermes hermes version     # Verify version
+docker stats hermes                    # Resource usage
+```

website/docs/user-guide/features/api-server.md

@@ -8,7 +8,7 @@ description: "Expose hermes-agent as an OpenAI-compatible API for any frontend"
 
 The API server exposes hermes-agent as an OpenAI-compatible HTTP endpoint. Any frontend that speaks the OpenAI format — Open WebUI, LobeChat, LibreChat, NextChat, ChatBox, and hundreds more — can connect to hermes-agent and use it as a backend.
 
-Your agent handles requests with its full toolset (terminal, file operations, web search, memory, skills) and returns the final response. Tool calls execute invisibly server-side.
+Your agent handles requests with its full toolset (terminal, file operations, web search, memory, skills) and returns the final response. When streaming, tool progress indicators appear inline so frontends can show what the agent is doing.
 
 ## Quick Start
 
@@ -85,6 +85,8 @@ Standard OpenAI Chat Completions format. Stateless — the full conversation is
 
 **Streaming** (`"stream": true`): Returns Server-Sent Events (SSE) with token-by-token response chunks. When streaming is enabled in config, tokens are emitted live as the LLM generates them. When disabled, the full response is sent as a single SSE chunk.
 
+**Tool progress in streams**: When the agent calls tools during a streaming request, brief progress indicators are injected into the content stream as the tools start executing (e.g. `` `💻 pwd` ``, `` `🔍 Python docs` ``). These appear as inline markdown before the agent's response text, giving frontends like Open WebUI real-time visibility into tool execution.
+
 ### POST /v1/responses
 
 OpenAI Responses API format. Supports server-side conversation state via `previous_response_id` — the server stores full conversation history (including tool calls and results) so multi-turn context is preserved without the client managing it.

website/docs/user-guide/features/checkpoints.md

@@ -1,30 +0,0 @@
-# Filesystem Checkpoints
-
-Hermes automatically snapshots your working directory before making file changes, giving you a safety net to roll back if something goes wrong. Checkpoints are **enabled by default**.
-
-## Quick Reference
-
-| Command | Description |
-|---------|-------------|
-| `/rollback` | List all checkpoints with change stats |
-| `/rollback <N>` | Restore to checkpoint N (also undoes last chat turn) |
-| `/rollback diff <N>` | Preview diff between checkpoint N and current state |
-| `/rollback <N> <file>` | Restore a single file from checkpoint N |
-
-## What Triggers Checkpoints
-
-- **File tools** — `write_file` and `patch`
-- **Destructive terminal commands** — `rm`, `mv`, `sed -i`, output redirects (`>`), `git reset`/`clean`
-
-## Configuration
-
-```yaml
-# ~/.hermes/config.yaml
-checkpoints:
-  enabled: true          # default: true
-  max_snapshots: 50      # max checkpoints per directory
-```
-
-## Learn More
-
-For the full guide — how shadow repos work, diff previews, file-level restore, conversation undo, safety guards, and best practices — see **[Checkpoints and /rollback](../checkpoints-and-rollback.md)**.

website/docs/user-guide/features/context-references.md

@@ -1,5 +1,6 @@
 ---
 sidebar_position: 9
+sidebar_label: "Context References"
 title: "Context References"
 description: "Inline @-syntax for attaching files, folders, git diffs, and URLs directly into your messages"
 ---

website/docs/user-guide/features/mcp.md

@@ -168,9 +168,7 @@ So a server that exposes callable tools but no resources/prompts will not get th
 
 ## Per-server filtering
 
-This is the main feature added by the PR work.
-
-You can now control which tools each MCP server contributes to Hermes.
+You can control which tools each MCP server contributes to Hermes, allowing fine-grained management of your tool namespace.
 
 ### Disable a server entirely
 

website/docs/user-guide/features/overview.md

@@ -0,0 +1,49 @@
+---
+title: "Features Overview"
+sidebar_label: "Overview"
+sidebar_position: 1
+---
+
+# Features Overview
+
+Hermes Agent includes a rich set of capabilities that extend far beyond basic chat. From persistent memory and file-aware context to browser automation and voice conversations, these features work together to make Hermes a powerful autonomous assistant.
+
+## Core
+
+- **[Tools & Toolsets](tools.md)** — Tools are functions that extend the agent's capabilities. They're organized into logical toolsets that can be enabled or disabled per platform, covering web search, terminal execution, file editing, memory, delegation, and more.
+- **[Skills System](skills.md)** — On-demand knowledge documents the agent can load when needed. Skills follow a progressive disclosure pattern to minimize token usage and are compatible with the [agentskills.io](https://agentskills.io/specification) open standard.
+- **[Persistent Memory](memory.md)** — Bounded, curated memory that persists across sessions. Hermes remembers your preferences, projects, environment, and things it has learned via `MEMORY.md` and `USER.md`.
+- **[Context Files](context-files.md)** — Hermes automatically discovers and loads project context files (`.hermes.md`, `AGENTS.md`, `CLAUDE.md`, `SOUL.md`, `.cursorrules`) that shape how it behaves in your project.
+- **[Context References](context-references.md)** — Type `@` followed by a reference to inject files, folders, git diffs, and URLs directly into your messages. Hermes expands the reference inline and appends the content automatically.
+- **[Checkpoints](../checkpoints-and-rollback.md)** — Hermes automatically snapshots your working directory before making file changes, giving you a safety net to roll back with `/rollback` if something goes wrong.
+
+## Automation
+
+- **[Scheduled Tasks (Cron)](cron.md)** — Schedule tasks to run automatically with natural language or cron expressions. Jobs can attach skills, deliver results to any platform, and support pause/resume/edit operations.
+- **[Subagent Delegation](delegation.md)** — The `delegate_task` tool spawns child agent instances with isolated context, restricted toolsets, and their own terminal sessions. Run up to 3 concurrent subagents for parallel workstreams.
+- **[Code Execution](code-execution.md)** — The `execute_code` tool lets the agent write Python scripts that call Hermes tools programmatically, collapsing multi-step workflows into a single LLM turn via sandboxed RPC execution.
+- **[Event Hooks](hooks.md)** — Run custom code at key lifecycle points. Gateway hooks handle logging, alerts, and webhooks; plugin hooks handle tool interception, metrics, and guardrails.
+- **[Batch Processing](batch-processing.md)** — Run the Hermes agent across hundreds or thousands of prompts in parallel, generating structured ShareGPT-format trajectory data for training data generation or evaluation.
+
+## Media & Web
+
+- **[Voice Mode](voice-mode.md)** — Full voice interaction across CLI and messaging platforms. Talk to the agent using your microphone, hear spoken replies, and have live voice conversations in Discord voice channels.
+- **[Browser Automation](browser.md)** — Full browser automation with multiple backends: Browserbase cloud, Browser Use cloud, local Chrome via CDP, or local Chromium. Navigate websites, fill forms, and extract information.
+- **[Vision & Image Paste](vision.md)** — Multimodal vision support. Paste images from your clipboard into the CLI and ask the agent to analyze, describe, or work with them using any vision-capable model.
+- **[Image Generation](image-generation.md)** — Generate images from text prompts using FAL.ai's FLUX 2 Pro model with automatic 2x upscaling via the Clarity Upscaler.
+- **[Voice & TTS](tts.md)** — Text-to-speech output and voice message transcription across all messaging platforms, with four provider options: Edge TTS (free), ElevenLabs, OpenAI TTS, and NeuTTS.
+
+## Integrations
+
+- **[Provider Routing](provider-routing.md)** — Fine-grained control over which AI providers handle your requests. Optimize for cost, speed, or quality with sorting, whitelists, blacklists, and priority ordering.
+- **[Fallback Providers](fallback-providers.md)** — Automatic failover to backup LLM providers when your primary model encounters errors, including independent fallback for auxiliary tasks like vision and compression.
+- **[API Server](api-server.md)** — Expose Hermes as an OpenAI-compatible HTTP endpoint. Connect any frontend that speaks the OpenAI format — Open WebUI, LobeChat, LibreChat, and more.
+- **[IDE Integration (ACP)](acp.md)** — Use Hermes inside ACP-compatible editors such as VS Code, Zed, and JetBrains. Chat, tool activity, file diffs, and terminal commands render inside your editor.
+- **[Honcho Memory](honcho.md)** — AI-native persistent memory for cross-session user modeling and personalization via dialectic reasoning.
+- **[RL Training](rl-training.md)** — Generate trajectory data from agent sessions for reinforcement learning and model fine-tuning.
+
+## Customization
+
+- **[Personality & SOUL.md](personality.md)** — Fully customizable agent personality. `SOUL.md` is the primary identity file — the first thing in the system prompt — and you can swap in built-in or custom `/personality` presets per session.
+- **[Skins & Themes](skins.md)** — Customize the CLI's visual presentation: banner colors, spinner faces and verbs, response-box labels, branding text, and the tool activity prefix.
+- **[Plugins](plugins.md)** — Add custom tools, hooks, and integrations without modifying core code. Drop a directory into `~/.hermes/plugins/` with a `plugin.yaml` and Python code.

website/docs/user-guide/features/plugins.md

@@ -1,5 +1,8 @@
 ---
-sidebar_position: 20
+sidebar_position: 11
+sidebar_label: "Plugins"
+title: "Plugins"
+description: "Extend Hermes with custom tools, hooks, and integrations via the plugin system"
 ---
 
 # Plugins
@@ -22,6 +25,56 @@ Drop a directory into `~/.hermes/plugins/` with a `plugin.yaml` and Python code:
 
 Start Hermes — your tools appear alongside built-in tools. The model can call them immediately.
 
+### Minimal working example
+
+Here is a complete plugin that adds a `hello_world` tool and logs every tool call via a hook.
+
+**`~/.hermes/plugins/hello-world/plugin.yaml`**
+
+```yaml
+name: hello-world
+version: "1.0"
+description: A minimal example plugin
+```
+
+**`~/.hermes/plugins/hello-world/__init__.py`**
+
+```python
+"""Minimal Hermes plugin — registers a tool and a hook."""
+
+
+def register(ctx):
+    # --- Tool: hello_world ---
+    schema = {
+        "name": "hello_world",
+        "description": "Returns a friendly greeting for the given name.",
+        "parameters": {
+            "type": "object",
+            "properties": {
+                "name": {
+                    "type": "string",
+                    "description": "Name to greet",
+                }
+            },
+            "required": ["name"],
+        },
+    }
+
+    def handle_hello(params):
+        name = params.get("name", "World")
+        return f"Hello, {name}! 👋  (from the hello-world plugin)"
+
+    ctx.register_tool("hello_world", schema, handle_hello)
+
+    # --- Hook: log every tool call ---
+    def on_tool_call(tool_name, params, result):
+        print(f"[hello-world] tool called: {tool_name}")
+
+    ctx.register_hook("post_tool_call", on_tool_call)
+```
+
+Drop both files into `~/.hermes/plugins/hello-world/`, restart Hermes, and the model can immediately call `hello_world`. The hook prints a log line after every tool invocation.
+
 Project-local plugins under `./.hermes/plugins/` are disabled by default. Enable them only for trusted repositories by setting `HERMES_ENABLE_PROJECT_PLUGINS=true` before starting Hermes.
 
 ## What plugins can do

website/docs/user-guide/features/skins.md

@@ -30,28 +30,150 @@ display:
 
 ## Built-in skins
 
-| Skin | Description | Agent branding |
-|------|-------------|----------------|
-| `default` | Classic Hermes — gold and kawaii | `Hermes Agent` |
-| `ares` | War-god theme — crimson and bronze | `Ares Agent` |
-| `mono` | Monochrome — clean grayscale | `Hermes Agent` |
-| `slate` | Cool blue — developer-focused | `Hermes Agent` |
-| `poseidon` | Ocean-god theme — deep blue and seafoam | `Poseidon Agent` |
-| `sisyphus` | Sisyphean theme — austere grayscale with persistence | `Sisyphus Agent` |
-| `charizard` | Volcanic theme — burnt orange and ember | `Charizard Agent` |
+| Skin | Description | Agent branding | Visual character |
+|------|-------------|----------------|------------------|
+| `default` | Classic Hermes — gold and kawaii | `Hermes Agent` | Warm gold borders, cornsilk text, kawaii faces in spinners. The familiar caduceus banner. Clean and inviting. |
+| `ares` | War-god theme — crimson and bronze | `Ares Agent` | Deep crimson borders with bronze accents. Aggressive spinner verbs ("forging", "marching", "tempering steel"). Custom sword-and-shield ASCII art banner. |
+| `mono` | Monochrome — clean grayscale | `Hermes Agent` | All grays — no color. Borders are `#555555`, text is `#c9d1d9`. Ideal for minimal terminal setups or screen recordings. |
+| `slate` | Cool blue — developer-focused | `Hermes Agent` | Royal blue borders (`#4169e1`), soft blue text. Calm and professional. No custom spinner — uses default faces. |
+| `poseidon` | Ocean-god theme — deep blue and seafoam | `Poseidon Agent` | Deep blue to seafoam gradient. Ocean-themed spinners ("charting currents", "sounding the depth"). Trident ASCII art banner. |
+| `sisyphus` | Sisyphean theme — austere grayscale with persistence | `Sisyphus Agent` | Light grays with stark contrast. Boulder-themed spinners ("pushing uphill", "resetting the boulder", "enduring the loop"). Boulder-and-hill ASCII art banner. |
+| `charizard` | Volcanic theme — burnt orange and ember | `Charizard Agent` | Warm burnt orange to ember gradient. Fire-themed spinners ("banking into the draft", "measuring burn"). Dragon-silhouette ASCII art banner. |
 
-## What a skin can customize
+## Complete list of configurable keys
 
-| Area | Keys |
-|------|------|
-| Banner + response colors | `colors.banner_*`, `colors.response_border` |
-| Spinner animation | `spinner.waiting_faces`, `spinner.thinking_faces`, `spinner.thinking_verbs`, `spinner.wings` |
-| Branding text | `branding.agent_name`, `branding.welcome`, `branding.response_label`, `branding.prompt_symbol` |
-| Tool activity prefix | `tool_prefix` |
+### Colors (`colors:`)
+
+Controls all color values throughout the CLI. Values are hex color strings.
+
+| Key | Description | Default (`default` skin) |
+|-----|-------------|--------------------------|
+| `banner_border` | Panel border around the startup banner | `#CD7F32` (bronze) |
+| `banner_title` | Title text color in the banner | `#FFD700` (gold) |
+| `banner_accent` | Section headers in the banner (Available Tools, etc.) | `#FFBF00` (amber) |
+| `banner_dim` | Muted text in the banner (separators, secondary labels) | `#B8860B` (dark goldenrod) |
+| `banner_text` | Body text in the banner (tool names, skill names) | `#FFF8DC` (cornsilk) |
+| `ui_accent` | General UI accent color (highlights, active elements) | `#FFBF00` |
+| `ui_label` | UI labels and tags | `#4dd0e1` (teal) |
+| `ui_ok` | Success indicators (checkmarks, completion) | `#4caf50` (green) |
+| `ui_error` | Error indicators (failures, blocked) | `#ef5350` (red) |
+| `ui_warn` | Warning indicators (caution, approval prompts) | `#ffa726` (orange) |
+| `prompt` | Interactive prompt text color | `#FFF8DC` |
+| `input_rule` | Horizontal rule above the input area | `#CD7F32` |
+| `response_border` | Border around the agent's response box (ANSI escape) | `#FFD700` |
+| `session_label` | Session label color | `#DAA520` |
+| `session_border` | Session ID dim border color | `#8B8682` |
+
+### Spinner (`spinner:`)
+
+Controls the animated spinner shown while waiting for API responses.
+
+| Key | Type | Description | Example |
+|-----|------|-------------|---------|
+| `waiting_faces` | list of strings | Faces cycled while waiting for API response | `["(⚔)", "(⛨)", "(▲)"]` |
+| `thinking_faces` | list of strings | Faces cycled during model reasoning | `["(⚔)", "(⌁)", "(<>)"]` |
+| `thinking_verbs` | list of strings | Verbs shown in spinner messages | `["forging", "plotting", "hammering plans"]` |
+| `wings` | list of [left, right] pairs | Decorative brackets around the spinner | `[["⟪⚔", "⚔⟫"], ["⟪▲", "▲⟫"]]` |
+
+When spinner values are empty (like in `default` and `mono`), hardcoded defaults from `display.py` are used.
+
+### Branding (`branding:`)
+
+Text strings used throughout the CLI interface.
+
+| Key | Description | Default |
+|-----|-------------|---------|
+| `agent_name` | Name shown in banner title and status display | `Hermes Agent` |
+| `welcome` | Welcome message shown at CLI startup | `Welcome to Hermes Agent! Type your message or /help for commands.` |
+| `goodbye` | Message shown on exit | `Goodbye! ⚕` |
+| `response_label` | Label on the response box header | ` ⚕ Hermes ` |
+| `prompt_symbol` | Symbol before the user input prompt | `❯ ` |
+| `help_header` | Header text for the `/help` command output | `(^_^)? Available Commands` |
+
+### Other top-level keys
+
+| Key | Type | Description | Default |
+|-----|------|-------------|---------|
+| `tool_prefix` | string | Character prefixed to tool output lines in the CLI | `┊` |
+| `tool_emojis` | dict | Per-tool emoji overrides for spinners and progress (`{tool_name: emoji}`) | `{}` |
+| `banner_logo` | string | Rich-markup ASCII art logo (replaces the default HERMES_AGENT banner) | `""` |
+| `banner_hero` | string | Rich-markup hero art (replaces the default caduceus art) | `""` |
 
 ## Custom skins
 
-Create YAML files under `~/.hermes/skins/`. User skins inherit missing values from the built-in `default` skin.
+Create YAML files under `~/.hermes/skins/`. User skins inherit missing values from the built-in `default` skin, so you only need to specify the keys you want to change.
+
+### Full custom skin YAML template
+
+```yaml
+# ~/.hermes/skins/mytheme.yaml
+# Complete skin template — all keys shown. Delete any you don't need;
+# missing values automatically inherit from the 'default' skin.
+
+name: mytheme
+description: My custom theme
+
+colors:
+  banner_border: "#CD7F32"
+  banner_title: "#FFD700"
+  banner_accent: "#FFBF00"
+  banner_dim: "#B8860B"
+  banner_text: "#FFF8DC"
+  ui_accent: "#FFBF00"
+  ui_label: "#4dd0e1"
+  ui_ok: "#4caf50"
+  ui_error: "#ef5350"
+  ui_warn: "#ffa726"
+  prompt: "#FFF8DC"
+  input_rule: "#CD7F32"
+  response_border: "#FFD700"
+  session_label: "#DAA520"
+  session_border: "#8B8682"
+
+spinner:
+  waiting_faces:
+    - "(⚔)"
+    - "(⛨)"
+    - "(▲)"
+  thinking_faces:
+    - "(⚔)"
+    - "(⌁)"
+    - "(<>)"
+  thinking_verbs:
+    - "processing"
+    - "analyzing"
+    - "computing"
+    - "evaluating"
+  wings:
+    - ["⟪⚡", "⚡⟫"]
+    - ["⟪●", "●⟫"]
+
+branding:
+  agent_name: "My Agent"
+  welcome: "Welcome to My Agent! Type your message or /help for commands."
+  goodbye: "See you later! ⚡"
+  response_label: " ⚡ My Agent "
+  prompt_symbol: "⚡ ❯ "
+  help_header: "(⚡) Available Commands"
+
+tool_prefix: "┊"
+
+# Per-tool emoji overrides (optional)
+tool_emojis:
+  terminal: "⚔"
+  web_search: "🔮"
+  read_file: "📄"
+
+# Custom ASCII art banners (optional, Rich markup supported)
+# banner_logo: |
+#   [bold #FFD700] MY AGENT [/]
+# banner_hero: |
+#   [#FFD700]  Custom art here  [/]
+```
+
+### Minimal custom skin example
+
+Since everything inherits from `default`, a minimal skin only needs to change what's different:
 
 ```yaml
 name: cyberpunk
@@ -78,4 +200,7 @@ tool_prefix: "▏"
 
 - Built-in skins load from `hermes_cli/skin_engine.py`.
 - Unknown skins automatically fall back to `default`.
-- `/skin` updates the active CLI theme immediately for the current session.
+- `/skin` updates the active CLI theme immediately for the current session.
+- User skins in `~/.hermes/skins/` take precedence over built-in skins with the same name.
+- Skin changes via `/skin` are session-only. To make a skin your permanent default, set it in `config.yaml`.
+- The `banner_logo` and `banner_hero` fields support Rich console markup (e.g., `[bold #FF0000]text[/]`) for colored ASCII art.

website/docs/user-guide/git-worktrees.md

@@ -1,5 +1,6 @@
 ---
-sidebar_position: 9
+sidebar_position: 3
+sidebar_label: "Git Worktrees"
 title: "Git Worktrees"
 description: "Run multiple Hermes agents safely on the same repository using git worktrees and isolated checkouts"
 ---

website/docs/user-guide/messaging/index.md

@@ -10,6 +10,26 @@ Chat with Hermes from Telegram, Discord, Slack, WhatsApp, Signal, SMS, Email, Ho
 
 For the full voice feature set — including CLI microphone mode, spoken replies in messaging, and Discord voice-channel conversations — see [Voice Mode](/docs/user-guide/features/voice-mode) and [Use Voice Mode with Hermes](/docs/guides/use-voice-mode-with-hermes).
 
+## Platform Comparison
+
+| Platform | Voice | Images | Files | Threads | Reactions | Typing | Streaming |
+|----------|:-----:|:------:|:-----:|:-------:|:---------:|:------:|:---------:|
+| Telegram | ✅ | ✅ | ✅ | ✅ | — | ✅ | ✅ |
+| Discord | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |
+| Slack | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |
+| WhatsApp | — | ✅ | ✅ | — | — | ✅ | ✅ |
+| Signal | — | ✅ | ✅ | — | — | ✅ | ✅ |
+| SMS | — | — | — | — | — | — | — |
+| Email | — | ✅ | ✅ | ✅ | — | — | — |
+| Home Assistant | — | — | — | — | — | — | — |
+| Mattermost | ✅ | ✅ | ✅ | ✅ | — | ✅ | ✅ |
+| Matrix | ✅ | ✅ | ✅ | ✅ | — | ✅ | ✅ |
+| DingTalk | — | — | — | — | — | ✅ | ✅ |
+| Feishu/Lark | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |
+| WeCom | ✅ | ✅ | ✅ | — | — | ✅ | ✅ |
+
+**Voice** = TTS audio replies and/or voice message transcription. **Images** = send/receive images. **Files** = send/receive file attachments. **Threads** = threaded conversations. **Reactions** = emoji reactions on messages. **Typing** = typing indicator while processing. **Streaming** = progressive message updates via editing.
+
 ## Architecture
 
 ```mermaid

website/docs/user-guide/messaging/open-webui.md

@@ -147,12 +147,16 @@ When you send a message in Open WebUI:
 1. Open WebUI sends a `POST /v1/chat/completions` request with your message and conversation history
 2. Hermes Agent creates an AIAgent instance with its full toolset
 3. The agent processes your request — it may call tools (terminal, file operations, web search, etc.)
-4. Tool calls happen invisibly server-side
-5. The agent's final text response is returned to Open WebUI
+4. As tools execute, **inline progress messages stream to the UI** so you can see what the agent is doing (e.g. `` `💻 ls -la` ``, `` `🔍 Python 3.12 release` ``)
+5. The agent's final text response streams back to Open WebUI
 6. Open WebUI displays the response in its chat interface
 
 Your agent has access to all the same tools and capabilities as when using the CLI or Telegram — the only difference is the frontend.
 
+:::tip Tool Progress
+With streaming enabled (the default), you'll see brief inline indicators as tools run — the tool emoji and its key argument. These appear in the response stream before the agent's final answer, giving you visibility into what's happening behind the scenes.
+:::
+
 ## Configuration Reference
 
 ### Hermes Agent (API server)

website/docs/user-guide/messaging/sms.md

@@ -1,5 +1,6 @@
 ---
 sidebar_position: 8
+sidebar_label: "SMS (Twilio)"
 title: "SMS (Twilio)"
 description: "Set up Hermes Agent as an SMS chatbot via Twilio"
 ---

website/docs/user-guide/skills/godmode.md

@@ -1,4 +1,6 @@
 ---
+sidebar_position: 1
+sidebar_label: "G0DM0D3 (Godmode)"
 title: "G0DM0D3 — Godmode Jailbreaking"
 description: "Automated LLM jailbreaking using G0DM0D3 techniques — system prompt templates, input obfuscation, and multi-model racing"
 ---

website/sidebars.ts

@@ -16,61 +16,37 @@ const sidebars: SidebarsConfig = {
     },
     {
       type: 'category',
-      label: 'Guides & Tutorials',
-      collapsed: true,
-      items: [
-        'guides/tips',
-        'guides/daily-briefing-bot',
-        'guides/team-telegram-assistant',
-        'guides/python-library',
-        'guides/use-mcp-with-hermes',
-        'guides/use-soul-with-hermes',
-        'guides/use-voice-mode-with-hermes',
-        'guides/migrate-from-openclaw',
-      ],
-    },
-    {
-      type: 'category',
-      label: 'User Guide',
+      label: 'Using Hermes',
       collapsed: true,
       items: [
         'user-guide/cli',
         'user-guide/configuration',
         'user-guide/sessions',
-        'user-guide/security',
-        'user-guide/docker',
         'user-guide/profiles',
+        'user-guide/git-worktrees',
+        'user-guide/docker',
+        'user-guide/security',
+        'user-guide/checkpoints-and-rollback',
+      ],
+    },
+    {
+      type: 'category',
+      label: 'Features',
+      collapsed: true,
+      items: [
+        'user-guide/features/overview',
         {
           type: 'category',
-          label: 'Messaging Gateway',
-          items: [
-            'user-guide/messaging/index',
-            'user-guide/messaging/telegram',
-            'user-guide/messaging/discord',
-            'user-guide/messaging/slack',
-            'user-guide/messaging/whatsapp',
-            'user-guide/messaging/signal',
-            'user-guide/messaging/email',
-            'user-guide/messaging/homeassistant',
-            'user-guide/messaging/mattermost',
-            'user-guide/messaging/matrix',
-            'user-guide/messaging/dingtalk',
-              'user-guide/messaging/feishu',
-              'user-guide/messaging/wecom',
-            'user-guide/messaging/open-webui',
-            'user-guide/messaging/webhooks',
-          ],
-        },
-        {
-          type: 'category',
-          label: 'Core Features',
+          label: 'Core',
           items: [
             'user-guide/features/tools',
             'user-guide/features/skills',
             'user-guide/features/memory',
             'user-guide/features/context-files',
+            'user-guide/features/context-references',
             'user-guide/features/personality',
             'user-guide/features/skins',
+            'user-guide/features/plugins',
           ],
         },
         {
@@ -81,11 +57,12 @@ const sidebars: SidebarsConfig = {
             'user-guide/features/delegation',
             'user-guide/features/code-execution',
             'user-guide/features/hooks',
+            'user-guide/features/batch-processing',
           ],
         },
         {
           type: 'category',
-          label: 'Web & Media',
+          label: 'Media & Web',
           items: [
             'user-guide/features/voice-mode',
             'user-guide/features/browser',
@@ -94,23 +71,10 @@ const sidebars: SidebarsConfig = {
             'user-guide/features/tts',
           ],
         },
-        {
-          type: 'category',
-          label: 'Integrations',
-          items: [
-            'user-guide/features/api-server',
-            'user-guide/features/acp',
-            'user-guide/features/mcp',
-            'user-guide/features/honcho',
-            'user-guide/features/provider-routing',
-            'user-guide/features/fallback-providers',
-          ],
-        },
         {
           type: 'category',
           label: 'Advanced',
           items: [
-            'user-guide/features/batch-processing',
             'user-guide/features/rl-training',
           ],
         },
@@ -125,25 +89,98 @@ const sidebars: SidebarsConfig = {
     },
     {
       type: 'category',
-      label: 'Developer Guide',
+      label: 'Messaging Platforms',
+      collapsed: true,
+      items: [
+        'user-guide/messaging/index',
+        'user-guide/messaging/telegram',
+        'user-guide/messaging/discord',
+        'user-guide/messaging/slack',
+        'user-guide/messaging/whatsapp',
+        'user-guide/messaging/signal',
+        'user-guide/messaging/email',
+        'user-guide/messaging/sms',
+        'user-guide/messaging/homeassistant',
+        'user-guide/messaging/mattermost',
+        'user-guide/messaging/matrix',
+        'user-guide/messaging/dingtalk',
+        'user-guide/messaging/feishu',
+        'user-guide/messaging/wecom',
+        'user-guide/messaging/open-webui',
+        'user-guide/messaging/webhooks',
+      ],
+    },
+    {
+      type: 'category',
+      label: 'Integrations',
+      collapsed: true,
+      items: [
+        'integrations/index',
+        'integrations/providers',
+        'user-guide/features/mcp',
+        'user-guide/features/acp',
+        'user-guide/features/api-server',
+        'user-guide/features/honcho',
+        'user-guide/features/provider-routing',
+        'user-guide/features/fallback-providers',
+      ],
+    },
+    {
+      type: 'category',
+      label: 'Guides & Tutorials',
+      collapsed: true,
+      items: [
+        'guides/tips',
+        'guides/build-a-hermes-plugin',
+        'guides/daily-briefing-bot',
+        'guides/team-telegram-assistant',
+        'guides/python-library',
+        'guides/use-mcp-with-hermes',
+        'guides/use-soul-with-hermes',
+        'guides/use-voice-mode-with-hermes',
+        'guides/migrate-from-openclaw',
+      ],
+    },
+    {
+      type: 'category',
+      label: 'Developer Guide',
+      collapsed: true,
       items: [
-        'developer-guide/architecture',
-        'developer-guide/agent-loop',
-        'developer-guide/provider-runtime',
-        'developer-guide/adding-providers',
-        'developer-guide/prompt-assembly',
-        'developer-guide/context-compression-and-caching',
-        'developer-guide/gateway-internals',
-        'developer-guide/session-storage',
-        'developer-guide/tools-runtime',
-        'developer-guide/acp-internals',
-        'developer-guide/trajectory-format',
-        'developer-guide/cron-internals',
-        'developer-guide/environments',
-        'developer-guide/adding-tools',
-        'developer-guide/creating-skills',
-        'developer-guide/extending-the-cli',
         'developer-guide/contributing',
+        {
+          type: 'category',
+          label: 'Architecture',
+          items: [
+            'developer-guide/architecture',
+            'developer-guide/agent-loop',
+            'developer-guide/prompt-assembly',
+            'developer-guide/context-compression-and-caching',
+            'developer-guide/gateway-internals',
+            'developer-guide/session-storage',
+            'developer-guide/provider-runtime',
+          ],
+        },
+        {
+          type: 'category',
+          label: 'Extending',
+          items: [
+            'developer-guide/adding-tools',
+            'developer-guide/adding-providers',
+            'developer-guide/creating-skills',
+            'developer-guide/extending-the-cli',
+          ],
+        },
+        {
+          type: 'category',
+          label: 'Internals',
+          items: [
+            'developer-guide/tools-runtime',
+            'developer-guide/acp-internals',
+            'developer-guide/cron-internals',
+            'developer-guide/environments',
+            'developer-guide/trajectory-format',
+          ],
+        },
       ],
     },
     {
@@ -152,13 +189,13 @@ const sidebars: SidebarsConfig = {
       items: [
         'reference/cli-commands',
         'reference/slash-commands',
+        'reference/profile-commands',
+        'reference/environment-variables',
         'reference/tools-reference',
         'reference/toolsets-reference',
         'reference/mcp-config-reference',
         'reference/skills-catalog',
         'reference/optional-skills-catalog',
-        'reference/profile-commands',
-        'reference/environment-variables',
         'reference/faq',
       ],
     },