openai-v2: await AsyncStream.close on default async streaming path

[yulin-li]

Description

BaseStreamWrapper.close() is synchronous and calls self.stream.close() without awaiting it. On the default (non-experimental) async streaming path, async_chat_completions_create_v_old returns a LegacyChatStreamWrapper wrapping an openai.AsyncStream, whose close() is a coroutine. Calling it without await therefore:

1. emits RuntimeWarning: coroutine 'AsyncStream.close' was never awaited, and
2. never closes the underlying httpx response/connection, leaking it on every early-closed async stream.

This also means a correct caller doing await wrapper.close() got None back, so they could not close the real stream either.

The experimental path (gen_ai_latest_experimental opt-in) was already fixed by #4500 via util-genai's AsyncStreamWrapper, but the default path remained affected.

This PR adds AsyncLegacyChatStreamWrapper(LegacyChatStreamWrapper) with an awaitable close() that awaits the underlying AsyncStream.close(), and uses it for async streaming chat completions on the legacy path. This mirrors the existing sync/async split already present in response_wrappers.py (AsyncResponseStreamWrapper).

Fixes #4710

Type of change

• Bug fix (non-breaking change which fixes an issue)

How Has This Been Tested?

• Added test_async_chat_completion_streaming_close_awaits_underlying_stream, which asserts the legacy async wrapper's close() awaits the underlying AsyncStream.close() and finalizes the span.
• Updated test_async_chat_completion_streaming_not_complete to await response.close() on the legacy path.
• Ran the openai-v2 suite with -W error::RuntimeWarning; the coroutine ... was never awaited warning is gone and all tests pass.

Does This PR Require a Core Repo Change?

• Yes.
• No.

Checklist:

• Followed the style guidelines of this project
• Changelogs have been updated
• Unit tests have been added
• Documentation has been updated

[linux-foundation-easycla]

The committers listed above are authorized under a signed CLA.

• ✅ login: Copilot / name: Copilot (958c563)
• ✅ login: yulin-li / name: Yulin Li (958c563)

[xrmx]

I was going to cut a new release for openai-v2 thinking this was fixed in main, but isn't. So this LGTM and I think is small enough to merge even if the work has been moved to the genai repo.

[yulin-li]

Hi, I don't have permission to merge, could you please help to merge it?

[Copilot]

Pull request overview

Fixes a resource leak in the OpenAI v2 (legacy/default semconv) async streaming chat completions path by ensuring the wrapper’s close() properly awaits openai.AsyncStream.close(), preventing un-awaited coroutine warnings and leaked httpx responses/connections.

Changes:

• Introduce AsyncLegacyChatStreamWrapper with an async def close() that awaits the underlying AsyncStream.close().
• Update async streaming chat completion tests to always await response.close() on the legacy path and add a regression test for awaiting the underlying close.
• Add a towncrier changelog entry describing the fix.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
instrumentation-genai/opentelemetry-instrumentation-openai-v2/src/opentelemetry/instrumentation/openai_v2/patch.py	Return an async legacy wrapper for async streaming chat completions and add an awaitable close() implementation.
instrumentation-genai/opentelemetry-instrumentation-openai-v2/tests/test_async_chat_completions.py	Update legacy async streaming test to await close() and add a regression test asserting the underlying AsyncStream.close() is awaited.
instrumentation-genai/opentelemetry-instrumentation-openai-v2/.changelog/4751.fixed	Document the bug fix in the package changelog.