fix: openai chat message dump logic

[dongs0104]

What does the PR do?

I found an issue with the OpenAI message format where user messages that are not strings are automatically converted to strings by the code, resulting in a [ChatCompletionResponseMessageContentPart(text="~~~")] string conversion problem. I have fixed this issue.

Checklist

• I have read the Contribution guidelines and signed the Contributor License
  Agreement
• PR title reflects the change and is of format <commit_type>: <Title>
• Changes are described in the pull request.
• Related issues are referenced.
• Populated github labels field
• Added test plan and verified test passes.
• Verified that the PR passes existing CI.
• I ran pre-commit locally (pre-commit install, pre-commit run --all)
• Verified copyright is correct on all changed files.
• Added succinct git squash message before merging ref.
• All template sections are filled out.
• Optional: Additional screenshots for behavior/output changes with before/after.

Commit Type:

Check the conventional commit type
box here and add the label to the github PR.

• build
• ci
• docs
• feat
• fix
• perf
• refactor
• revert
• style
• test

Related PRs:

#7561

Where should the reviewer start?

@rmccorm4

Test plan:

Caveats:

Background

To fix the issue of ChatCompletionResponseMessageContentPart being mixed with the output

Related Issues: (use one of the action keywords Closes / Fixes / Resolves / Relates to)

• closes GitHub issue: #xxx

[rmccorm4]

Hi @dongs0104, thanks for contributing this fix!

1. Have you emailed a signed CLA as described here: https://github.com/triton-inference-server/server/blob/main/CONTRIBUTING.md#contributor-license-agreement-cla?
2. Can you share an example request/payload that would produce this string conversion problem before your fix so we can add a test case? If you'd like to add as simple test case, you can refer to here: https://github.com/triton-inference-server/server/blob/main/python/openai/tests/test_chat_completions.py.

[dongs0104]

@rmccorm4 i signed it and add test case

[rmccorm4]

Started pipeline: 21452750

[rmccorm4]

Hi @dongs0104,

Have you tested this is passing locally?

I get a schema validation error on your new test case (return code 422) because content is expected to be a string or List[dict], but you're passing a dict directly:

server/python/openai/openai_frontend/schemas/openai.py

Line 659 in bcff3da

content: Union[str, List[ChatCompletionRequestMessageContentPart]] = Field(

Your test case:

    def test_chat_completions_user_prompt_dict(self, client, model: str):
        messages = [
            {
                "role": "user",
                "content": {"type": "text", "text": "What is machine learning?"},
            }
        ]

Expected Schema: (note the list [ ])

    def test_chat_completions_user_prompt_dict(self, client, model: str):
        # Pass a List of Dicts for content to expose the 'type' field
        messages = [
            {
                "role": "user",
                "content": [{"type": "text", "text": "What is machine learning?"}],
            }
        ]

I think this should get the test to pass.

[dongs0104]

@rmccorm4 i tested it after your comment thanks

[nv-kmcgill53]

CLA has been accepted. This PR can be merged at Triton's discretion.

[rmccorm4]

Thanks for your contribution @dongs0104!