yoniebans
f4b32301ce
Manual script that hits real Anthropic API to confirm: (1) >8000 px images are still rejected with the same error message, (2) our clamp produces an image Anthropic accepts. Run when threshold drift is suspected.
94 lines
3.3 KiB
Python
94 lines
3.3 KiB
Python
"""Manual verification that Anthropic's >8000px image rejection is real and
|
|
that our dimension clamp unblocks the call.
|
|
|
|
Not a pytest — a script you run manually when you suspect Anthropic moved the
|
|
threshold or the error message changed. Mirrors the pattern in
|
|
``scripts/benchmark_browser_eval.py``.
|
|
|
|
Usage:
|
|
ANTHROPIC_API_KEY=... .venv/bin/python scripts/verify_anthropic_pixel_cap.py
|
|
"""
|
|
from __future__ import annotations
|
|
|
|
import base64
|
|
import os
|
|
import sys
|
|
import tempfile
|
|
from io import BytesIO
|
|
from pathlib import Path
|
|
|
|
|
|
def main() -> int:
|
|
if not os.environ.get("ANTHROPIC_API_KEY"):
|
|
print("ANTHROPIC_API_KEY not set", file=sys.stderr)
|
|
return 2
|
|
|
|
# Make tools/ importable
|
|
repo_root = Path(__file__).resolve().parent.parent
|
|
sys.path.insert(0, str(repo_root))
|
|
|
|
from PIL import Image
|
|
import anthropic
|
|
from tools.vision_tools import _MAX_IMAGE_DIMENSION, _resize_image_for_vision
|
|
|
|
model = "claude-haiku-4-5-20251001"
|
|
client = anthropic.Anthropic()
|
|
|
|
with tempfile.TemporaryDirectory() as td:
|
|
raw_path = Path(td) / "oversize.png"
|
|
Image.new("RGB", (8500, 100), (128, 128, 128)).save(raw_path, "PNG")
|
|
raw_b64 = base64.b64encode(raw_path.read_bytes()).decode()
|
|
|
|
# Step 1: confirm Anthropic still rejects raw oversized image.
|
|
print(f"[1/2] Sending raw 8500x100 PNG to {model}...")
|
|
try:
|
|
client.messages.create(
|
|
model=model, max_tokens=64,
|
|
messages=[{"role": "user", "content": [
|
|
{"type": "image", "source": {
|
|
"type": "base64", "media_type": "image/png", "data": raw_b64,
|
|
}},
|
|
{"type": "text", "text": "describe"},
|
|
]}],
|
|
)
|
|
print(" UNEXPECTED: raw oversized image was accepted. "
|
|
"Anthropic may have relaxed the cap — investigate.")
|
|
return 1
|
|
except anthropic.BadRequestError as exc:
|
|
msg = str(exc).lower()
|
|
if "8000" in msg or "dimensions" in msg:
|
|
print(f" OK rejected as expected: {exc}")
|
|
else:
|
|
print(f" WARN unexpected error shape: {exc}")
|
|
print(" _is_image_size_error may need a refresh.")
|
|
return 1
|
|
|
|
# Step 2: confirm clamped image is accepted.
|
|
print(f"[2/2] Clamping + retrying...")
|
|
data_url = _resize_image_for_vision(raw_path, mime_type="image/png",
|
|
clamp_dimensions=True)
|
|
_, clamped_b64 = data_url.split(",", 1)
|
|
decoded = Image.open(BytesIO(base64.b64decode(clamped_b64)))
|
|
print(f" Clamped to {decoded.width}x{decoded.height} "
|
|
f"(cap={_MAX_IMAGE_DIMENSION})")
|
|
resp = client.messages.create(
|
|
model=model, max_tokens=64,
|
|
messages=[{"role": "user", "content": [
|
|
{"type": "image", "source": {
|
|
"type": "base64", "media_type": "image/png", "data": clamped_b64,
|
|
}},
|
|
{"type": "text", "text": "describe"},
|
|
]}],
|
|
)
|
|
if not resp.content:
|
|
print(" FAIL: clamped image accepted but response empty")
|
|
return 1
|
|
print(f" OK clamped image accepted")
|
|
|
|
print("\nAll checks passed.")
|
|
return 0
|
|
|
|
|
|
if __name__ == "__main__":
|
|
sys.exit(main())
|