github.com/OpenHands/agent-canvas/pull/1573 this improves the UX of the vision tool for non vision LLMs. Would appreciate a 👀 See screenshot.