fix: handle null response output text

[he-yufeng]

• I understand that this repository is auto-generated and my pull request may not be merged

Changes being requested

• Allow ResponseOutputText.text to be None when the API returns a nullable output_text.text item.
• Make Response.output_text skip null text items instead of failing or appending a non-string value.
• Make responses.parse(...) leave null output_text items unparsed while still parsing later non-null output text items.
• Add regression coverage for both Response.output_text and the structured parse path.

Additional context & links

Fixes #3063.

I know the repository is generated; this PR is intended to pin down the runtime behavior and regression tests for the nullable API payload shape.

Verification:

• PYTHONPATH=src python -m pytest tests\lib\responses\test_responses.py -q
• python -m ruff check src\openai\types\responses\response_output_text.py src\openai\types\responses\response.py src\openai\lib\_parsing\_responses.py tests\lib\responses\test_responses.py
• python -m pyright src\openai\types\responses\response_output_text.py src\openai\types\responses\response.py src\openai\lib\_parsing\_responses.py tests\lib\responses\test_responses.py
• python -m ruff format --check src\openai\types\responses\response_output_text.py src\openai\types\responses\response.py src\openai\lib\_parsing\_responses.py tests\lib\responses\test_responses.py
• git diff --check

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 897ce0a518

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Use v1-compatible validation in test helper

In the test-pydantic-v1 nox session, noxfile.py installs pydantic<2 and runs the whole pytest suite, but this helper now calls Response.model_validate(...) unconditionally. The SDK's BaseModel does not shim model_validate for pydantic v1, so that CI path will fail with AttributeError before these regressions run. Use the repo's compatibility helper or Response.parse_obj(...) here instead.

Useful? React with 👍 / 👎.

[he-yufeng]

Rebased onto latest main — the only conflict was in the test file (both sides added tests near the same spot), resolved by keeping both my null-output-text cases and the new program_items parametrize test. The source change is unchanged: output_text / parse_response now skip output_text items whose text is null instead of tripping over them. Ready for review whenever someone has a moment.

[he-yufeng]

A note on why this keeps showing conflicts: the edits under src/openai/types/responses/ land in Stainless-generated files, so every spec regeneration collides with them. The durable part of this fix is the lib/_parsing side (skip null text items while parsing later ones). I can rebase this down to just the lib change plus tests and leave the nullable output_text.text type to the spec side, if that's the preferred split. Holding off on further rebases until then since each one conflicts again on the generated files.

[he-yufeng]

Re-verified against current main: ResponseOutputText.text is still a required str, so a Responses payload item that arrives with text: null still fails parsing instead of being tolerated. The null-tolerant handling and its test still apply. (Sibling #3401 landed recently; this one is the same error class on the output side.)

[marcuswood-oai]

Thanks for working on this! We’re closing this for now because we haven’t established a case where Responses structured-output parsing should accept null or empty text. If you’re hitting this in practice, please open an issue with a minimal reproduction and the API/provider you’re using—we’re happy to revisit.