llmserve

llmserve is a terminal app for people who keep local LLM models across several folders or inference tools. It brings model discovery and server management into one interface so you can choose an existing model and an available backend.

• Model discovery: Find GGUF and MLX models in LM Studio folders, llama.cpp and Hugging Face caches, or custom directories, then search, sort and filter them by source.
• Backend selection: Serve GGUF files through llama-server, KoboldCpp or LocalAI, and MLX models through MLX on Apple Silicon, with separate settings for each backend.
• Multiple servers: Run different models and backends at the same time, see which models are serving, and stop servers from the interface.
• Logs and diagnostics: Watch backend output with highlighted warnings and errors, and inspect retained logs after a server exits.
• Vision models: Automatically detect companion mmproj projector files for use with llama-server.

The app runs locally on macOS and Linux and is open source under the MIT license. It relies on installed inference backends to run models. It also detects Ollama, vLLM and LM Studio, but can't launch local model files through those backends because they manage their own model registries or servers.