Skip to main content
The ability to chat with your documents and create a private knowledge base is powered by a sophisticated, on-device vector store. How It Works: When you import a document, the app breaks it down into chunks of text and generates a “vector embedding” for each chunk. This vector, along with the original text, is saved to the on-device vector store. When you ask a question, the app generates a vector for your question and then uses a high-speed search algorithm to find the text chunks in the database with the most similar vectors. Key Architectural Features:
  • Private and Secure: The vector store is powered by an embedded, on-device database.
  • Multiple, Isolated Stores: You can create multiple, named vector stores.
  • Built for Speed: The vector store is built on a high-performance database and uses a state-of-the-art HNSW index for its vector search.