Extract google model usage using genai-prices #3466
Conversation
| usage=RequestUsage( | ||
| input_tokens=46, | ||
| output_tokens=528, | ||
| output_tokens=1429, |
There was a problem hiding this comment.
All these increases are because toolUsePromptTokenCount is meant to be counted too, see pydantic/genai-prices#200. For example here 1429-528==901 which is the value of tool_use_prompt_tokens just below.
There was a problem hiding this comment.
Pull Request Overview
This PR upgrades the genai-prices dependency from version 0.0.35 to 0.0.40 and refactors Google model usage extraction to leverage the new library's capabilities. The manual audio token calculation logic has been replaced with a call to RequestUsage.extract(), which automatically handles usage extraction from provider responses.
Key changes:
- Upgraded genai-prices from 0.0.35 to 0.0.40
- Refactored
_metadata_as_usage()to useRequestUsage.extract()method - Updated test expectations for output token counts to reflect the new calculation method
Reviewed Changes
Copilot reviewed 3 out of 4 changed files in this pull request and generated no comments.
| File | Description |
|---|---|
| uv.lock | Updated genai-prices package from 0.0.35 to 0.0.40 with new checksums and added Windows wheels for ruamel.yaml.clib |
| pydantic_ai_slim/pyproject.toml | Updated genai-prices dependency requirement from >=0.0.35 to >=0.0.40 |
| pydantic_ai_slim/pydantic_ai/models/google.py | Refactored _metadata_as_usage() to accept provider parameters and use RequestUsage.extract(), removing manual audio token calculation logic |
| tests/models/test_google.py | Updated test assertions with new output token values and added provider/provider_url parameters to function calls |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Docs Preview
|
No description provided.