ChatVault – Local-first semantic search for WhatsApp (Rust and WASM) (github.com)

ChatVault has emerged as a revolutionary local-first semantic search engine for WhatsApp, enabling users to index and search their chat histories privately. Utilizing Rust and WebAssembly, ChatVault runs a BERT neural network entirely in-browser, eliminating the need for external servers and ensuring complete data confidentiality. Unlike WhatsApp's traditional keyword-based search, ChatVault leverages vector embeddings to understand semantic meaning, allowing users to retrieve contextually relevant messages. The significance of ChatVault lies in its innovative engineering and the seamless user experience it provides. By integrating a hybrid search algorithm that balances cosine similarity with keyword matching, it maximizes search accuracy while offloading heavy computing tasks to a Web Worker. This ensures smooth UI performance at 60fps, even with intensive AI tasks. ChatVault is also fully offline-capable, processing WhatsApp exports directly in the user's RAM without any data egress. This project not only showcases high-performance systems programming in the browser but also emphasizes the growing trend towards privacy-focused AI applications.