What's Changed
- ChatReqLLM: send OpenAI reasoning items back in tool loops by @georgeguimaraes in #673
- ChatReqLLM: keep reasoning details through persistence and streaming by @brainlid in #674
- ChatReqLLM: give each thinking block its own signature by @brainlid in #675
- ChatReqLLM: send signed thinking back in its original order by @brainlid in #676
- ChatReqLLM: keep each streamed thinking block as its own signed part by @brainlid in #677
- Set Gemini tool call status correctly and remove dropped complete keys by @nelsonkopliku in #671
- ChatReqLLM: keep a content-less refusal as a content_filtered message by @brainlid in #678
- ChatVertexAI: report finishReason as the message status by @brainlid in #679
- ChatReqLLM: require req_llm 1.25 and read phase only from each part by @brainlid in #680
- prep for v0.15.1 by @brainlid in #681
Full Changelog: v0.15.0...v0.15.1