timescale/pgvectorscale

Health v2 · maintenance only

Adoption not part of the score

3136 stars · 156 forks observed · 2026-10-01

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

pgvectorscale is a PostgreSQL extension written in Rust that complements pgvector with a StreamingDiskANN index, statistical binary quantization, and label-based filtered vector search. It delivers significantly higher performance and cost-efficient embedding search for AI applications on self-hosted PostgreSQL.

Use cases

• speed up pgvector similarity search in postgres
• run vector search at scale without a dedicated vector database
• replace pinecone with self-hosted postgres embeddings
• filtered vector search with metadata labels
• store and query millions of embeddings cost-efficiently
• build RAG applications on postgres

When to choose

• you already use PostgreSQL and pgvector but need lower latency and higher throughput
• you want to self-host vector search instead of paying for managed vector databases
• you need label-filtered approximate nearest neighbor search
• you prefer a Rust-based extension with active development

When to avoid

• you need a database other than PostgreSQL
• your vector workload is small enough that plain pgvector suffices
• you cannot install custom Postgres extensions (e.g., some managed hosts)
• you need turnkey managed infrastructure without self-hosting

Facets

library · maturity active

vector-database search-engine database databases machine-learning developer-tools self-hosted cloud rust postgres-extension pgvector diskann vector-search embeddings ann pgrx quantization search retrieval-augmented-generation linux docker

Member repositories

For agents

markdown · JSON · MCP: product_card(name="timescale/pgvectorscale")