Enables text-only LLMs to see images on demand by integrating Qwen vision models via MCP, supporting stateful multi-turn conversations with automatic history trimming.
