#sd-cpp

A multi-engine and multimodal local inference platform. One process routes GGUF, safetensors, quantized, and MLX models to the right engine lane (llama.cpp, mistral.rs, SGLang, MLX-lm, sd.cpp and whisper) and serves them through OpenAI- and ollama-compatible APIs, with durable jobs, federation, speech, vision, and image lanes behind the same CLI.