This unusual stack is built to create a thinking machine rather than a moving one, using ROS as the nervous system for AI-driven logic.
Some ROS packages provide Docker containers. Almost all are ROS CI conform. The repository link at the bottom is the one that brings everything together. Most sub-components have their own repositories and can be used independently. Extensive documentation is provided within all packages. It’s a versatile collection of components designed to handle everything from deterministic filtering to semantic AI-driven tasks.
Please note that the stack is under continuous development and is evolving nearly daily.
You can also see it live on Twitch (link below). Ask it about itself, tell it to do real-time coding to solve your task, or even let it evolve. (Don’t panic, it’s moderated! :D)
RQT Graph
LLM Node (a ROS driver around the well known OpenAI compatible chat API)
Qwen3 TTS (a STT component is also available but not deployed in the example images)
Audio / Video Streamer (developed for Twitch but can also be used local with ffplay/ffmpeg)
See it live at Twitch