prometheus-inference-platform

Production-grade self-hosted LLM inference platform — FastAPI gateway with JWT auth, fine-grained per-model authorization scopes, rate limiting & circuit breakers; multi-backend model manager (llama.cpp/MLX/vLLM/SGLang) across distributed hosts; React admin dashboard; full OTel tracing. Podman-native, 100% self-hosted — no vendor lock-in.

GitHub

Official source

Start locally with the CLI guide