fix(langfuse): Handle null usage values to prevent validation errors

[AlanPonnachan]

Handle null usage values to prevent validation errors

Relevant issues

Fixes #16330

Pre-Submission checklist

Please complete all items before asking a Lite-LLM maintainer to review your PR

• I have Added testing in the tests/litellm/integrations/test_langfuse.py 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
• My PR's scope is as isolated as possible, it only solves 1 specific problem

Type

🐛 Bug Fix
✅ Test

Changes

This pull request resolves a pydantic.ValidationError that occurred when sending data to the Langfuse callback.

Problem:
When an upstream provider (e.g., an OpenAI-compatible endpoint) returned a usage object with null values for token counts, these None values were passed directly to the Langfuse client, which requires integer values for these fields.

Solution:
The _log_langfuse_v2 method in litellm/integrations/langfuse/langfuse.py has been updated to robustly handle None values for all token fields (prompt_tokens, completion_tokens, total_tokens, etc.). It now coalesces any None value to 0 before constructing the usage and usage_details payloads for Langfuse. This ensures the data is always compliant with the Langfuse schema.

Testing:
A new unit test, test_log_langfuse_v2_handles_null_usage_values, has been added to tests/test_litellm/integrations/test_langfuse.py. This test:

1. Simulates an API response with None values in the usage object.
2. Asserts that the values sent to the Langfuse client are correctly converted to 0.
3. Is isolated using unittest.mock.patch to avoid side effects and ensure it only tests the intended logic, consistent with the existing testing style in the file.

[vercel]

@AlanPonnachan is attempting to deploy a commit to the CLERKIEAI Team on Vercel.

A member of the Team first needs to authorize it.