fix: handle null response output in parser

[puneetdixit200]

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

Changes being requested

Fixes #3325.

parse_response now treats a null response.output as an empty output list before iterating, avoiding a TypeError when a completed response event has output: null. Added a regression test that constructs that response shape and verifies parsing returns output=[].

AI-assisted: yes.

Additional context & links

Verification run locally:

• .venv/bin/python -m pytest -o addopts='' tests/lib/responses/test_responses.py -q -> 6 passed
• .venv/bin/ruff format --check src/openai/lib/_parsing/_responses.py tests/lib/responses/test_responses.py -> 2 files already formatted
• .venv/bin/ruff check src/openai/lib/_parsing/_responses.py tests/lib/responses/test_responses.py -> all checks passed
• .venv/bin/mypy src/openai/lib/_parsing/_responses.py -> success, no issues found in 1 source file
• git diff --check -> passed

Not run: full ./scripts/test; the repository docs note most tests require the mock server, and this change is covered by the focused responses parser test file.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f3adae0750

ℹ️ About Codex in GitHub

Your team has set up Codex to 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 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Preserve streamed output when completed event omits it

When the Responses stream has already accumulated output items and the final response.completed event carries output: null (the case this change is trying to tolerate), this line converts the final response's output to []. I checked ResponseStreamState.accumulate_event: on response.completed it sets _completed_response = parse_response(... event.response ...) (src/openai/lib/streaming/responses/_responses.py:359-364), and get_final_response() returns that object, not the accumulated snapshot. As a result, streamed calls with text/tool output but a null final output no longer raise, but response.completed.response, get_final_response().output, and output_text are empty, silently dropping the data sent by the earlier stream events.

Useful? React with 👍 / 👎.