Pull a model, load it, mint a key, watch the cards — all from a browser. A self-hosted OpenAI-compatible API on your own NVIDIA GPUs, running two mainline engines: vLLM for throughput, llama.cpp for GGUF models vLLM cannot serve.
About
LLM Warden is your own OpenAI-compatible API on your own NVIDIA GPUs: two mainline inference engines — vLLM and llama.cpp — behind one control plane, one published port and a browser UI. It pulls weights from HuggingFace, loads a model onto the cards you choose, puts /v1/* in fro…
