Add Claude Sonnet 3.5 VisionAgentCoder

pyproject.toml

@@ -41,7 +41,6 @@ pillow-heif = "^0.16.0"
 pytube = "15.0.0"
 anthropic = "^0.31.0"
 pydantic = "2.7.4"
-eva-decord = "^0.6.1"
 av = "^11.0.0"
 
 [tool.poetry.group.dev.dependencies]

tests/integ/test_tools.py

@@ -22,6 +22,7 @@
  grounding_sam,
  ixc25_image_vqa,
  ixc25_video_vqa,
+ ixc25_temporal_localization,
  loca_visual_prompt_counting,
  loca_zero_shot_counting,
  ocr,
@@ -238,6 +239,17 @@ def test_ixc25_video_vqa() -> None:
  assert "cat" in result.strip()
 
 
+def test_ixc25_temporal_localization() -> None:
+ frames = [
+ np.array(Image.fromarray(ski.data.cat()).convert("RGB")) for _ in range(10)
+ ]
+ result = ixc25_temporal_localization(
+ prompt="What animal is in this video?",
+ frames=frames,
+ )
+ assert result == [True] * 10
+
+
 def test_ocr() -> None:
  img = ski.data.page()
  result = ocr(

tests/unit/tools/test_video.py

@@ -6,5 +6,5 @@ def test_extract_frames_from_video():
  video_path = "tests/data/video/test.mp4"
 
  # there are 48 frames at 24 fps in this video file
- res = extract_frames_from_video(video_path)
- assert len(res) == 2
+ res = extract_frames_from_video(video_path, fps=24)
+ assert len(res) == 48

vision_agent/agent/__init__.py

@@ -2,6 +2,7 @@
 from .vision_agent import VisionAgent
 from .vision_agent_coder import (
  AzureVisionAgentCoder,
+ ClaudeVisionAgentCoder,
  OllamaVisionAgentCoder,
  VisionAgentCoder,
 )

vision_agent/agent/agent_utils.py

@@ -14,36 +14,48 @@ def _extract_sub_json(json_str: str) -> Optional[Dict[str, Any]]:
  if match:
  json_str = match.group()
  try:
+ # remove trailing comma
+ trailing_bracket_pattern = r",\s+\}"
+ json_str = re.sub(trailing_bracket_pattern, "}", json_str, flags=re.DOTALL)
+
  json_dict = json.loads(json_str)
  return json_dict # type: ignore
  except json.JSONDecodeError:
  return None
  return None
 
 
+def _find_markdown_json(json_str: str) -> str:
+ pattern = r"```json(.*?)```"
+ match = re.search(pattern, json_str, re.DOTALL)
+ if match:
+ return match.group(1).strip()
+ return json_str
+
+
+def _strip_markdown_code(inp_str: str) -> str:
+ pattern = r"```python.*?```"
+ cleaned_str = re.sub(pattern, "", inp_str, flags=re.DOTALL)
+ return cleaned_str
+
+
 def extract_json(json_str: str) -> Dict[str, Any]:
+ json_str = json_str.replace("\n", " ").strip()
+
  try:
- json_str = json_str.replace("\n", " ")
- json_dict = json.loads(json_str)
+ return json.loads(json_str) # type: ignore
  except json.JSONDecodeError:
- if "```json" in json_str:
- json_str = json_str[json_str.find("```json") + len("```json") :]
- json_str = json_str[: json_str.find("```")]
- elif "```" in json_str:
- json_str = json_str[json_str.find("```") + len("```") :]
- # get the last ``` not one from an intermediate string
- json_str = json_str[: json_str.find("}```")]
- try:
- json_dict = json.loads(json_str)
- except json.JSONDecodeError as e:
- json_dict = _extract_sub_json(json_str)
- if json_dict is not None:
- return json_dict # type: ignore
- error_msg = f"Could not extract JSON from the given str: {json_str}"
+ json_orig = json_str
+ json_str = _strip_markdown_code(json_str)
+ json_str = _find_markdown_json(json_str)
+ json_dict = _extract_sub_json(json_str)
+
+ if json_dict is None:
+ error_msg = f"Could not extract JSON from the given str: {json_orig}"
  _LOGGER.exception(error_msg)
- raise ValueError(error_msg) from e
+ raise ValueError(error_msg)
 
- return json_dict # type: ignore
+  return json_dict
 
 
 def extract_code(code: str) -> str: