[Bug Fix] Spend Tracking Bug Fix, don't modify in memory default litellm params

[ishaan-jaff]

[Bug Fix] Spend Tracking Bug Fix, don't modify in memory default litellm params

When a user invokes the /responses endpoint and then immediately calls /chat/completions, the cost/spend for the /chat/completions call was not being recorded.

What caused the issue:

• The /responses handler passes metadata_variable_name="litellm_metadata", so the helper moves the defaults under that key instead of "metadata".
• Because we operated directly on the router’s shared default_litellm_params dict, that rename stuck for all future calls.
• The original code popped "metadata" out of self.default_litellm_params and re‑inserted it under the new name in place.
• Subsequent calls to _update_kwargs_with_default_litellm_params saw only litellm_metadata in the defaults, so they merged into the wrong field.

Fix made

By copying default_litellm_params and popping/merging metadata on that local copy, we avoid mutating the router’s shared defaults dict. Metadata now always lands under the correct metadata_variable_name per call, so downstream spend‑tracking reliably picks it up.

Relevant issues

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
• I have added a screenshot of my new test passing locally
• My PR passes all unit tests on (make test-unit)[https://docs.litellm.ai/docs/extras/contributing_code]
• My PR's scope is as isolated as possible, it only solves 1 specific problem

Type

🆕 New Feature
✅ Test

Changes

[vercel]

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

Name	Status	Preview	Comments	Updated (UTC)
litellm	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Apr 19, 2025 9:12pm