VectorChord

概要

VectorChord is a PostgreSQL extension engineered for highly scalable, high-performance, and disk-efficient vector similarity search. Serving as the successor to pgvecto.rs, it delivers enhanced stability and performance. This tool empowers users to host massive numbers of high-dimensional vectors within PostgreSQL, significantly reducing infrastructure costs compared to alternative solutions, while maintaining high search quality and ensuring seamless integration with existing pgvector setups.

主な機能

• Enhanced Performance: Up to 5x faster queries and 16x higher insert/index build throughput.
• Seamless Integration: Fully compatible with pgvector data types and syntax, offering optimal defaults.
• Long Vector Support: Capable of storing and searching vectors up to 60,000 dimensions.
• Accelerated Index Build: Leverages IVF and RaBitQ compression for efficient vector storage and faster index creation.
• Affordable Vector Search: Stores millions of vectors with low memory footprint, significantly cutting infrastructure costs.
• 831 GitHub stars

ユースケース

• Migrating from pgvector or pgvecto.rs to a more performant and memory-efficient vector database solution.
• Implementing scalable and cost-effective vector search for AI/LLM applications within PostgreSQL.
• Deploying large-scale vector databases handling billions of vectors in production environments.