sqliteai/sqlite-vector

SQLite-Vector is a cross-platform, ultra-efficient SQLite extension that brings vector search capabilities to your embedded database.

What it solves

SQLite-Vector provides production-grade vector search capabilities directly within SQLite, eliminating the need for external vector databases (like FAISS or Weaviate) or complex, time-consuming preindexing processes. It allows developers to perform semantic search and build RAG (Retrieval-Augmented Generation) applications on the edge, in browsers, or on servers with a very low memory footprint.

How it works

It is implemented as a cross-platform SQLite extension that allows vectors to be stored as standard BLOBs in ordinary tables. It uses highly optimized C implementation with SIMD acceleration and supports various distance functions (L2, Cosine, Dot, etc.). To speed up searches, it employs quantization techniques—including TurboQuant (a low-bit scalar quantizer)—to create compact representations of vectors that can be scanned quickly without reconstructing the full vectors.

Who it’s for

It is designed for developers building Edge AI applications, mobile apps (iOS, Android), and privacy-preserving offline AI workloads who need an efficient, plug-and-play vector search solution that integrates with existing SQLite workflows.

Highlights

  • No Preindexing: Start searching immediately without long preprocessing or index-building phases.
  • Ordinary Tables: Store vectors as BLOBs in standard tables without requiring virtual tables.
  • TurboQuant Support: Uses SIMD 2, 3, and 4-bit quantization for ultra-compact scans.
  • Low Resource Usage: Defaults to just 30MB of RAM and works offline across all major operating systems.
  • Broad Type Support: Compatible with Float32, Float16, BFloat16, Int8, UInt8, and 1Bit formats.

Related

  • Project
    1yefuwang1/vectorlite

    A runtime-loadable SQLite extension that enables fast, SIMD-accelerated vector search using HNSW, providing a SQL interface for approximate nearest neighbor queries.

  • Project
    RyanCodrai/turbovec

    A Rust-based vector index with Python bindings that uses the TurboQuant algorithm to provide high-compression, fast SIMD search without requiring a separate training phase.

  • Dispatch
    Turbovec: High-Performance Vector Search using Google's TurboQuant

    Turbovec is a Rust-based vector index implementing the TurboQuant algorithm, offering up to 16x memory compression and faster search speeds than FAISS IndexPQFastScan.

  • Project
    weaviate/weaviate

    An open-source, cloud-native vector database that enables semantic search at scale by storing objects and vectors together.

  • Project
    VexDB-THU/VexDB-Lite

    A high-performance vector similarity search engine for PostgreSQL, DuckDB, and SQLite that uses a custom graph index and quantization to enable efficient ANN search.