Self-hosted, GPU-pooling LLM inference run your own model fleet on your own hardware. One streaming, OpenAI-compatible API over transformers & llama.cpp. Inference you own.
Three chat models, live images, and the Hugpy console
Embeddable web console — mountConsole() + a PTY broker
The fleet's agent CLI — enroll any box into your central
Self-hosted, OpenAI-compatible inference you own