# RigStack

An open-source command-line tool that unifies Ollama, vLLM and ComfyUI behind one endpoint on NVIDIA hardware, for running a private local inference stack without per-token cloud costs.

RigStack turns a machine with an NVIDIA GPU into a self-managed private AI server. It unifies NInfer, Ollama, vLLM and ComfyUI behind a single Traefik-routed endpoint, and its command-line interface can switch inference presets, move a model between CPU and GPU, and check the status of every service from one command.

It supports a unified model registry across the backing services, multiple named inference presets per model for different quantization or context-length trade-offs, a built-in retrieval-augmented generation endpoint backed by Qdrant, and self-hosted prompt observability through Langfuse, with everything running as containerized, reproducible Docker services and no data leaving the operator's own network.

Website: https://github.com/gogluejf/rig-stack