Language and Pixels to Action Space

A system that turns spoken or written instructions plus a camera feed into robot actions. An LLM breaks the instruction into a plan, LLaMA 3.2 Vision reads the scene for objects and obstacles, and a small API converts the plan into movement commands for a differential drive robot. It keeps short and long term memory with LangGraph and Redis, and there is a Streamlit interface to drive it.