Chatty is a client-side AI chatbot that leverages WebGPU to run LLMs entirely in the browser without server-side processing. It supports popular open-source models including Gemma, Llama 2 and 3, and Mistral, enabling offline usage after initial model download. The application includes file-based question answering using local embeddings with MemoryVectorStore, markdown rendering, code highlighting, conversation history management, and custom memory instructions.