Metallama is a web UI for managing llama.cpp instances

Metallama is an open-source Python FastAPI tool that provides a web interface to manage llama.cpp server instances. It allows users to search and download GGUF models from Hugging Face, configure and monitor servers, and manage model weights directly from the UI.

• Search and download GGUFs from the Hugging Face API
• Configure, spawn, and monitor llama.cpp servers
• Manage model weights library via the web interface
• Provide an Ollama compatible proxy gateway at /ollama
• Monitor host RAM and VRAM usage
• Plug remote llama.cpp servers under the Ollama proxy at /models
• Estimate total memory footprint of an instance (WIP)

The project aims to simplify the process of handling inference parameters and model management without requiring manual terminal juggling.