chatty

Why we included this project

Chatty runs open-source LLMs entirely in the browser through WebGPU, so your prompts and files never leave your machine. If privacy is the reason you have not warmed to hosted chatbots, this gives you a familiar ChatGPT-style interface without the third-party server in the middle. It goes beyond plain chat: you can load a PDF or code file and ask questions about it, add custom memory instructions, use voice input, and export conversations as JSON or Markdown. After the first model download it works offline, which makes it useful on low-connectivity or air-gapped setups. For teams curious about in-browser inference, it is also a complete, working example of what that stack looks like in practice.