fix: preserve metadata for custom callbacks on codex/responses path (…

[saneroen]

Description

fix: preserve metadata for custom callbacks on codex/responses path (#21204)

• Use metadata or litellm_metadata when calling update_environment_variables in responses/main.py so metadata is not overwritten by None on the bridge path (completion -> responses API).
• Add tests for metadata in custom callback for codex models and for litellm_metadata in aresponses().

Relevant issues

Fixes #21204

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

• I have Added testing in the tests/litellm/ directory, Adding at least 1 test is a hard requirement - see details
• My PR passes all unit tests on make test-unit
• My PR's scope is as isolated as possible, it only solves 1 specific problem
• I have requested a Greptile review by commenting @greptileai and received a Confidence Score of at least 4/5 before requesting a maintainer review

CI (LiteLLM team)

CI status guideline:

• 50-55 passing tests: main is stable with minor issues.
• 45-49 passing tests: acceptable but needs attention
• <= 40 passing tests: unstable; be careful with your merges and assess the risk.

• Branch creation CI run
  Link:

• CI run for the last commit
  Link:

• Merge / cherry-pick CI run
  Links:

Type

🐛 Bug Fix

Changes

• litellm/responses/main.py: Preserve metadata (or litellm_metadata) when calling update_environment_variables on the responses API bridge path so it is not overwritten by None.
• tests/test_litellm/responses/test_metadata_codex_callback.py: Add tests that ensure metadata is passed to custom callbacks for codex models via acompletion() and for direct aresponses() with litellm_metadata.

[vercel]

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

Project	Deployment	Actions	Updated (UTC)
litellm	Ready	Preview, Comment	Feb 16, 2026 4:41pm

[greptile-apps]

Greptile Summary

Fixes a regression where user-supplied metadata was not reaching custom callbacks for codex models (e.g., gpt-5.1-codex) that route through the responses API bridge. The bridge converts metadata to litellm_metadata in kwargs, but responses/main.py only passed the metadata parameter (which is None on the bridge path) to update_environment_variables. The fix falls back to kwargs.get("litellm_metadata") when metadata is None.

• litellm/responses/main.py: Added metadata_for_callbacks = metadata or kwargs.get("litellm_metadata") or {} before the update_environment_variables call, ensuring the bridge path preserves metadata for callbacks.
• tests/test_litellm/responses/test_metadata_codex_callback.py: Added two mock-based async tests covering the codex acompletion() bridge path and the direct aresponses() path with litellm_metadata.
• Tests use proper HTTP mocking (no real network calls) but should clean up litellm.callbacks global state after each test to avoid interference in parallel test runs.

Confidence Score: 4/5

• This PR is safe to merge — it fixes a clear metadata propagation bug with a minimal, well-scoped change.
• The core fix is a single line of fallback logic that correctly handles the metadata flow through the codex bridge path. The change is low-risk since it only adds a fallback when metadata is None, and doesn't alter behavior when metadata is already set. Tests cover both paths. Minor concern is that tests don't clean up global litellm.callbacks state.
• tests/test_litellm/responses/test_metadata_codex_callback.py — needs litellm.callbacks cleanup to avoid test isolation issues.

Important Files Changed

Filename	Overview
litellm/responses/main.py	Adds fallback logic to preserve metadata from litellm_metadata kwargs when metadata parameter is None (codex bridge path). The fix is correct and well-targeted.
tests/test_litellm/responses/test_metadata_codex_callback.py	New test file with two async tests verifying metadata preservation for codex models and direct responses API calls. Tests use mocks (no real network calls) but don't clean up litellm.callbacks global state.

Sequence Diagram

sequenceDiagram
    participant User
    participant acompletion as litellm.acompletion()
    participant Bridge as responses_api_bridge
    participant Responses as responses/main.py
    participant Logging as LiteLLMLoggingObj
    participant Callback as CustomLogger

    User->>acompletion: metadata={"foo": "bar"}
    acompletion->>Bridge: litellm_params={metadata: {"foo": "bar"}}
    Bridge->>Bridge: _build_sanitized_litellm_params()
    Note over Bridge: Merges metadata → litellm_metadata<br/>Strips metadata from params
    Bridge->>Responses: responses(**request_data)<br/>litellm_metadata={"foo": "bar"}, metadata=None
    Responses->>Responses: metadata_for_callbacks = metadata or kwargs["litellm_metadata"]
    Note over Responses: FIX: Falls back to litellm_metadata<br/>when metadata is None
    Responses->>Logging: update_environment_variables(metadata=metadata_for_callbacks)
    Logging->>Callback: async_log_success_event(kwargs)<br/>kwargs["litellm_params"]["metadata"] = {"foo": "bar"}

Loading

_{Last reviewed commit: fb2ee17}

[greptile-apps]

_{2 files reviewed, 1 comment}

_{Edit Code Review Agent Settings | Greptile}

[jeromeroussin]

@saneroen Can we expect the metadata in kwargs['standard_logging_object'] as well?

[ghost]

Can we expect the metadata in kwargs['standard_logging_object'] as well?

since the slo is built in the logging object, i believe this should work