Notum-Robotics/ClusterFlock ? reverse-engineered prompt

Build me a distributed AI orchestrator called ClusterFlock that can run on a mixed set of machines and make them work together as one cluster.

I want a main dashboard and API that run on a central host, then separate agents I can install on different computers like Macs, Linux machines with NVIDIA GPUs, DGX style boxes, or LM Studio setups. The system should automatically check the hardware on each machine, pick the best model it can handle, load it, and keep everything connected. It should also expose an OpenAI compatible chat endpoint so I can send requests to the cluster like a normal AI API.

Please include support for autonomous missions too, where I can send a task and a coordinator model breaks it up and routes work to other agents. A simple web dashboard to see the cluster, agents, and jobs would be great. Use Python and keep the main orchestrator lightweight, and if you need current setup details or docs for model or Docker behavior, look them up online.