fix(langchain): cached token usage

[hassiebp]

Important

Fixes token usage calculation in langchain.py and adds a test for cached token usage handling.

• Behavior:
  • Fixes token usage calculation in _parse_usage_model() in langchain.py by adjusting input and output values when input_token_details and output_token_details are present.
  • Adds a test test_cached_token_usage() in test_langchain.py to verify correct handling of cached token usage.
• Misc:
  • No changes to constants or logging.
  • No renames or typo fixes.

^{This description was created by for 331cc43. It will automatically update as commits are pushed.}

Greptile Summary

Disclaimer: Experimental PR review

This PR fixes token usage calculations in the Langchain integration by correctly handling cached tokens and preventing double-counting when detailed token breakdowns are present.

• Modified langfuse/callback/langchain.py to subtract input/output token details from totals in _parse_usage_model function to prevent double-counting
• Added comprehensive test case in tests/test_langchain.py to verify token usage accuracy with caching enabled
• Ensures proper tracking of input_cache_read values and accurate cost calculations when using cached tokens

[greptile-apps]

_{2 file(s) reviewed, 3 comment(s)
Edit PR Review Bot Settings | Greptile}