5d67e76788
Ollama's Qwen3 thinking models stream the chain-of-thought in a `reasoning` field with an empty `content`. The OpenAI-compatible provider only recognised `reasoning_content`/`thinking`, so reasoning-only responses surfaced no text at all. Add `reasoning` to both the streaming (`ChunkDelta`) and non-streaming (`ChatMessage`) shapes and fall back to it when extracting thinking, preserving the existing thinking/text content-block separation. Covered by two new tests for the streaming and non-streaming paths. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> |
||
|---|---|---|
| .. | ||
| benches | ||
| src | ||
| tests | ||
| Cargo.toml | ||