fix(tests): replace asyncio.sleep(1) with event-based wait to fix flaky metadata callback tests

[ishaan-jaff]

Relevant issues

Fixes flaky test test_metadata_passed_via_litellm_metadata_responses_api (CI test ID: litellm_mapped_tests_core/1262471).

What changed

The two tests in test_metadata_codex_callback.py used await asyncio.sleep(1) to wait for the custom callback to fire. The callback is dispatched via asyncio.create_task() in utils.py, so the fixed 1-second sleep races against the event loop scheduler — causing intermittent failures.

Fix: add an asyncio.Event to MetadataCaptureCallback that gets .set() inside async_log_success_event. Tests now do asyncio.wait_for(callback.event.wait(), timeout=5.0) — deterministic, no timing dependency.

Pre-Submission checklist

• Added at least 1 test (fixing existing tests to be reliable)
• make test-unit passes locally

Type

• New Feature
• Bug Fix
• Documentation
• Other

Changes

• tests/test_litellm/responses/test_metadata_codex_callback.py: replace asyncio.sleep(1) with asyncio.Event-based wait in both test functions

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
litellm	Ready	Preview, Comment	Feb 21, 2026 8:33pm

[greptile-apps]

Greptile Summary

Fixes flaky metadata callback tests by replacing asyncio.sleep(1) with an asyncio.Event-based synchronization mechanism. The callback handler dispatches via an async logging worker queue, making fixed-time sleeps unreliable. The new approach uses asyncio.wait_for(callback.event.wait(), timeout=5.0) which is deterministic — it proceeds as soon as the callback fires, with a 5-second timeout as a safety net.

• Adds asyncio.Event to MetadataCaptureCallback.__init__ and sets it in async_log_success_event
• Replaces await asyncio.sleep(1) with await asyncio.wait_for(callback.event.wait(), timeout=5.0) in both test functions
• No functional changes to production code; all changes are test-only

Confidence Score: 5/5

• This PR is safe to merge — it only modifies test code with a well-understood synchronization pattern.
• The change is test-only, minimal in scope, and replaces a known-flaky pattern (fixed sleep) with a deterministic event-based wait. The asyncio.Event + wait_for approach is the standard Python idiom for this kind of async synchronization. No production code is affected.
• No files require special attention.

Important Files Changed

Filename	Overview
tests/test_litellm/responses/test_metadata_codex_callback.py	Replaces flaky asyncio.sleep(1) with event-based wait using asyncio.Event in both test functions. Clean, correct fix with no issues found.

Sequence Diagram

sequenceDiagram
    participant Test as Test Function
    participant LiteLLM as litellm.acompletion
    participant Worker as LoggingWorker
    participant CB as MetadataCaptureCallback

    Test->>LiteLLM: await acompletion(model, messages, metadata)
    LiteLLM->>Worker: enqueue(async_success_handler)
    LiteLLM-->>Test: return result
    Note over Test: Before: await asyncio.sleep(1) — race condition
    Note over Test: After: await asyncio.wait_for(event.wait(), 5s)
    Worker->>CB: await async_log_success_event(kwargs, ...)
    CB->>CB: captured_kwargs = kwargs
    CB->>CB: event.set()
    CB-->>Test: event unblocks wait_for
    Test->>Test: assert metadata preserved

Loading

_{Last reviewed commit: 61236c0}

[greptile-apps]

_{1 file reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}