fix: CI issues for openai agents

[satyadevai]

Closes #2376

---

Note

Handle output as string or list for tool/function call outputs, extracting input_text into message.contents.*.message_content.text; update tests accordingly.

• Span attribute extraction (OpenAI agents):
  • Enhance /_processor.py to parse tool/function call output as either str or list.
    • For list outputs, extract input_text items into message.contents.{i}.message_content.text.
    • Leave TODOs for input_image and input_file types.
• Tests:
  • Update expectations for empty/None outputs and add coverage for list-based outputs in tests/test_span_attribute_helpers.py.

^{Written by Cursor Bugbot for commit 96af527. This will update automatically on new commits. Configure here.}

[caroger]

this is brittle, and test would fail again when agno 2.2 is released because it would use to recorded vcr yaml from older response again, correct?

  # Parse version into tuple for comparison
    version_parts = tuple(map(int, agno.__version__.split(".")[:2]))  # Major.minor only
    vcr_path = "team_coordinate_run21.yaml" if version_parts >= (2, 1) else "team_coordinate_run.yaml"

[satyadevai]

This change is not required as we are instrmenting the latest package of agno, also moved the changes related agno to different MR

[caroger]

do a type check here instead of runtime check.

assert never

[satyadevai]

@caroger, could you please confirm whether the attribute name is correct?. Based on my understanding, FunctionCallOutput can contain either a string or a List[Union[ResponseInputTextContentParam, ResponseInputImageContentParam, ResponseInputFileContentParam]]. I’ve assumed it will be a list of contents, so I’m passing these values as contents.