A simple web interface for running llama.cpp models locally, built using:
- React
- Material UI
- Express,js
- NodeJS
- MongoDB (not implemented yet)
- Socket.IO and socket.io-client
This is a work in progress and is being continually updated! I am not an expert in the MERN stack and web dev, so code and UX design may not always be of the highest quality!
- Unit tests
- Might change backend from Express.js to Flask (this is my first time using Javascript and I'm more comfortable with Python)
- Easier setup and installation (paths to models and settings are currently hardcoded)
- Configuring model parameters like prompts, temperature etc. from web UI
- Storing model settings, chat history, logs etc. in MongoDB
- Cleaner UI
- Support for multiple users
example.mp4
Tiny usage example (sped up slightly because my CPU is pretty slow, even when running running the smallest LLaMa model with 4-bit quantization)