[Dhruv Shaswat (Dhruv-Sh-git)]

Fixes #9779. "fix: streamed chat completions with n>1 produce separate messages" --body "

What problem does this PR solve?

When a Chat Completions request uses n > 1, the OpenAI streaming client merged deltas from all choices into a single ChatResponseUpdate. This caused the final response to contain interleaved text instead of separate messages.
For example:

• Expected: [A1A2, B1B2]
• Actual: [A1B1A2B2]
  The non-streaming path already returns a separate message for each choice.

Root cause

_parse_response_update_from_openai() aggregated contents from all chunk.choices using contents.extend(choice_contents). This lost the separation between parallel completions before ChatResponse.from_updates could assemble the final messages.

Changes made

• Updated the streaming loop in _chat_completion_client.py to emit one ChatResponseUpdate per choice, using choice.index to keep each completion independent.
• Added a usage-only ChatResponseUpdate when chunk.usage is present.
• Preserved the tool-call identity mapping for each choice.
• Left the existing parsing helpers unchanged.
• Added a regression test in test_chat_stream_merge_fix.py that verifies two streamed choices produce separate final messages.

Why this fixes the issue

Each choice now produces its own ChatResponseUpdate with a distinct message ID. ChatResponse.from_updates can therefore group updates correctly and build one assistant message per choice, matching the non-streaming behavior.

Testing

• Added regression coverage for multiple streamed choices.
• Existing tests passed locally with uv run pytest -q (0 failures).

Additional notes

• No public API changes.
• No new runtime dependencies.
• The .NET implementation is unchanged."

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota.

[Eduard van Valkenburg (eavanvalkenburg)]

We do not support N > 1, closing this.