Handle null response output during parsing

[FU-max-boop]

Summary

Fixes #3325.

parse_response() currently assumes response.output is always iterable. Some streaming backends can send a final response.completed payload with output: null after emitting valid output-item events, which causes a bare TypeError in the response stream accumulator.

This treats a null output the same as an empty output list during parsing, so stream consumers can still complete and inspect the final response instead of crashing in the parser.

Tests

• uv run python -m pytest -o addopts="" tests/lib/responses/test_responses.py -q
• uv run ruff check src/openai/lib/_parsing/_responses.py tests/lib/responses/test_responses.py
• uv run ruff format --check src/openai/lib/_parsing/_responses.py tests/lib/responses/test_responses.py
• uv run pyright src/openai/lib/_parsing/_responses.py tests/lib/responses/test_responses.py

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: c5de32017b

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]

Preserve accumulated output when completed payload is null

When a streaming backend sends response.completed with output: null after earlier response.output_item.* events, this fallback parses the final payload as output=[]. The streaming accumulator sets _completed_response by calling parse_response(event.response, ...) on completion (src/openai/lib/streaming/responses/_responses.py:359-364), so ResponseCompletedEvent.response and get_final_response() lose all message/tool output that was already accumulated in the snapshot instead of letting consumers inspect it. In this null-output case, the parser/stream state should preserve the accumulated snapshot output rather than replacing it with an empty list.

Useful? React with 👍 / 👎.

[FU-max-boop]

Fixed in d9b78c4d: ResponseStreamState now preserves the accumulated snapshot output when the final response.completed payload has output: null, and the new streaming regression test covers the created/output_item/content_part/text_delta/completed sequence.