SPTAG

Source Code

Description

SPTAG (Space Partition Tree And Graph) is an open-source, distributed library for approximate nearest neighbor (ANN) search, designed for building and searching large-scale vector indexes efficiently and scalably.

Features

• Approximate Nearest Neighbor Search: Efficiently finds vectors in large datasets that are nearest to a query vector, using L2 or cosine distance metrics.
• Large-Scale Vector Indexing: Designed to handle billions of vectors, suitable for big data scenarios.
• Multiple Indexing Methods:
  • SPTAG-KDT: Uses kd-tree and relative neighborhood graph; advantageous for faster index building.
  • SPTAG-BKT: Uses balanced k-means tree and relative neighborhood graph; advantageous for higher search accuracy in high-dimensional data.
• Distributed Online Serving: Supports searching across multiple machines for scalability.
• Online Vector Updates: Fresh update support allows for online vector insertion and deletion.
• Index Builder and Searcher Modules: Modular architecture for index construction and search.
• Docker Support: Can be built and run in Docker containers for easy deployment.
• Language Support: Core in C++ with Python wrappers available.
• Open Source and Extensible: Released under the MIT license, welcoming community contributions.
• Research-Backed: Features and algorithms are based on recent research, including support for relaxed monotonicity and incremental in-place updates.

Requirements

• swig >= 4.0.2
• cmake >= 3.12.0
• boost >= 1.67.0

Installation

• Supports installation on Linux and Windows, with instructions for both.
• Dockerfile provided for containerized builds.

Tags

open-source, ann, distributed, scalable

License

MIT License

Pricing

SPTAG is open-source and free to use under the MIT license.