agentjido / req_llm / 6f338a2388026bf88f1c28a2f4e5124dfb72d26a

49%

Build Type

push

github

Committed by web-flow

Commit Message

Enhance response handling and usage normalization (#124)

* Enhance response handling and usage normalization

- Updated `ReqLLM.Response` to include a new usage normalization step, ensuring consistent field names between streaming and non-streaming responses.
- Modified `ReqLLM.StreamResponse` to utilize the normalized usage fields, improving the structure of response data.
- Adjusted tests to validate the new usage structure, ensuring accurate assertions for response usage metrics.
- Added handling for fixture replay to ensure correct response steps are executed during tests.

These changes improve the consistency and clarity of usage data across different response types.

* Fix Dialyzer type mismatch in decode_response/2

- Normalize model to string when building Response structs
- Add @dialyzer nowarn attribute for decode_response/2 minimal fixture request
- Extract model_string in decode_chat_response for consistent typing

Run Details

12 of 15 new or added lines in 4 files covered. (80.0%)

1 existing line in 1 file now uncovered.

3359 of 6593 relevant lines covered (50.95%)

556.05 hits per line