# ScaNN Library

> Scalable Nearest Neighbors library by Google Research that provides efficient vector similarity search at scale. Uses anisotropic vector quantization and advanced compression techniques to handle twice as many queries per second compared to alternatives.

- **URL:** https://github.com/google-research/google-research/tree/master/scann
- **Category:** Developer Tools & Libraries
- **Tags:** Ann, Google, Quantization
- **Updated:** 2026-04-22 04:16
- **Canonical page:** https://vectordb.ever.works/items/scann-library

## Details

## Overview

ScaNN (Scalable Nearest Neighbors) is a method for efficient vector similarity search at scale, developed by Google Research. It introduces innovative compression techniques and quantization methods to significantly improve search performance.

## Features

- **Anisotropic Vector Quantization**: Unlike traditional quantization, ScaNN prioritizes preserving parallel components between vectors, ideal for Maximum Inner Product Search (MIPS)
- **Three-Phase Search**: Partitioning, scoring with quantized vectors, and rescoring
- **High Performance**: Handles roughly twice as many queries per second for a given accuracy compared to the next-fastest library
- **Open Source**: Available on GitHub with comprehensive documentation
- **Easy Installation**: Install via `pip install scann`

## Algorithms

- Product Quantization with anisotropic loss
- Tree-based partitioning (optional)
- Asymmetric hashing
- SOAR (Spilling with Orthogonality-Amplified Residuals) for improved efficiency

## Use Cases

- Large-scale similarity search
- Recommendation systems
- Image retrieval
- Document search
- Neural search applications

## Pricing

Free and open-source.

---

_This Markdown mirror is generated by Ever Works for AI agents. The canonical HTML page is at https://vectordb.ever.works/items/scann-library._