veqlite
A lightweight TypeScript vector database library that stores text chunks with metadata and embeddings in SQLite using the sqlite-vec extension, with built-in Hugging Face Transformers embedding support for cosine similarity search.
veqlite is a vector database library for building retrieval-augmented generation and similarity search, storing text chunks, metadata and embeddings in SQLite with the sqlite-vec extension, and integrating with the huggingface/transformers package for embedding generation.
It offers fast cosine-distance similarity search, support for custom metadata types, bulk insertion for faster loading, and a choice of in-memory or persistent databases. It runs on multiple JavaScript runtimes through separate adapters: a built-in adapter for Bun, a built-in adapter for Node.js 24 and later, and an optional better-sqlite3 adapter with native bindings for faster bulk operations on Node.js.
Built with TypeScript, distributed as the veqlite npm package.