fix: handle null response.output in parse_response

[xodn348]

Fixes #3325

What

parse_response() in src/openai/lib/_parsing/_responses.py crashes with TypeError: 'NoneType' object is not iterable when response.output is null. This can happen when the server sends a response.completed event with output: null while valid response.output_item.done events were streamed earlier. The fix coerces None to [] before iteration.

Verification

• Build: pass
• Tests: pass (5 passed)
• Lint: pass (ruff)
• Type check: pass (pyright, 0 errors)

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 74a244f293

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 completion output is null

When a streamed Responses request ends with response.completed carrying output: null after earlier output item/content events, this fallback parses the completion payload as an empty ParsedResponse. ResponseStreamState.accumulate_event passes event.response directly into parse_response, and get_final_response()/the completed event then expose _completed_response, so callers receive empty output/output_text even though the accumulated snapshot contains the generated items. The null case needs to reuse or merge the accumulated stream snapshot instead of coercing it to [] here.

Useful? React with 👍 / 👎.