Secret Llama is a fully private chatbot supporting Llama 3, Mistral, and other open source models, all running entirely in your browser with no need for a server or installation. It has an easy-to-use interface comparable to ChatGPT, thanks to the inference engine by webllm. System requirements include a modern browser with WebGPU support. Supported models range from TinyLlama-1.1B to Mistral-7B. The project welcomes contributions to enhance the interface, add support for more models, improve loading times, and address bugs. Other projects by the same author include screenshot to code and Pico, an AI-powered app builder.
https://github.com/abi/secret-llama