exo-explore/exo ? reverse-engineered prompt

I want to build an app that lets me run the really big, frontier AI models locally on my own hardware. The main idea is to create an AI cluster by connecting all my devices, like a few Macs. It should automatically discover other devices on the network so there's no manual configuration needed.

The system needs to be smart enough to split a huge model across all the available devices, figuring out the best way to do it based on each machine's power and the network connection between them. It should definitely take advantage of high-speed connections like Thunderbolt for low latency. Let's use Python and the MLX framework for the AI backend, since this is geared towards Apple Silicon.

It also needs a built-in web dashboard. From there, I should be able to see the status of my cluster, manage the models, and have a chat interface. And most importantly, it has to offer an API that's compatible with the standard OpenAI and Claude APIs so I can easily use it with my existing clients and tools.