Chat with an already-loaded Qwen2.5-VL or Qwen3-VL text encoder instead of
loading a separate LLM -- multi-turn, vision-capable chat sharing the same
VRAM, exposed as both a sidebar panel and an OpenAI-compatible
/v1/chat/completions endpoint.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
