Files
cherry-studio/src
suyao 163ae9c085 fix: properly handle thinking mode switching in qwenThinkingMiddleware
Fix issue #11612 where Ollama couldn't disable Qwen3 thinking mode.
The middleware now properly removes existing /think or /no_think suffixes
before adding the correct suffix based on current reasoning_effort setting.

Previously, messages with existing suffixes were skipped, causing thinking
mode to persist even when user switched back to non-thinking mode.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-01 20:43:14 +08:00
..