agentjido / req_llm / c5ff026cd32d2490fe5d5e3f7978bd7e01bee87a

49%

Build Type

push

github

Committed by web-flow

Commit Message

Refactor streaming from Req to Finch for production stability (#63)

* Streaming refactor

* Quality

* Formatting

* Fix tests

* Enhance Anthropic provider to ensure max_tokens is always present and update context encoding.

* Add reasoning effort control for GPT-5 models and enhance response handling

- Introduced a new `--reasoning-effort` option in the `req_llm.gen` mix task to specify reasoning levels (minimal, low, medium, high) for GPT-5 models.
- Updated response handling to include reasoning token counts in usage statistics and output.
- Enhanced the Anthropic and OpenAI providers to support reasoning effort parameters and adjust request options accordingly.
- Improved decoding logic to handle reasoning content in streaming responses.

These changes enhance the flexibility and clarity of AI interactions, particularly for models with reasoning capabilities.

* Formatting

* Example agent to test streaming tool calls

Run Details

317 of 904 new or added lines in 20 files covered. (35.07%)

19 existing lines in 3 files now uncovered.

1953 of 3271 relevant lines covered (59.71%)

296.15 hits per line