fix(langchain): token counts for vertex

[hassiebp]

Important

Adjusts token count calculations for Vertex AI in _parse_usage_model() to handle modality-specific token counts without going below zero.

• Behavior:
  • Adjusts token count calculations in _parse_usage_model() in CallbackHandler.py for Vertex AI.
  • Subtracts modality-specific token counts from input and output ensuring they do not go below zero.
  • Applies changes to prompt_tokens_details, candidates_tokens_details, and cache_tokens_details.

^{This description was created by for 33e09bc. You can customize this summary. It will automatically update as commits are pushed.}

Greptile Summary

Disclaimer: Experimental PR review

Updates token count handling for Vertex AI in the Langchain integration to ensure accurate usage statistics for multi-modal interactions.

• Fixed token count processing in langfuse/langchain/CallbackHandler.py to properly subtract prompt, candidate, and cache tokens from total counts
• Added support for multi-modal input/output token tracking in Vertex AI integrations
• Improved accuracy of usage statistics by correctly handling different token types (prompt, completion, cache)

[greptile-apps]

LGTM

_{1 file(s) reviewed, no comment(s)
Edit PR Review Bot Settings | Greptile}