serge

• Serge is a self-hosted chat interface built around llama.cpp for running LLM models without requiring API keys
• Features a SvelteKit frontend and uses Redis for storing chat history & parameters
• Backend powered by FastAPI + LangChain, wrapping llama.cpp calls using python bindings
• Designed for users who want to run language models locally with full control over their data
• Requires significant memory resources to run LLaMA models properly
• Offers Docker-based deployment options for easy setup
• Includes authentication features with configurable session expiry
• Provides API documentation at /api/docs endpoint when running locally
• Community support available through Discord fa-solid fa-up-right-from-square
• Open-source under MIT/Apache-2.0 licenses with active contribution opportunities

The project is particularly notable for its Svelte-based frontend implementation and self-contained architecture that eliminates dependency on external API services.