pgvectorscale
activeby TigerData (formerly Timescale) · since 2024 · also known as pgai vectorscale
TigerData's Postgres extension adding StreamingDiskANN indexes on top of pgvector.
Overview
pgvectorscale is an open-source PostgreSQL extension from TigerData (formerly Timescale) that complements pgvector with a StreamingDiskANN index, based on Microsoft's DiskANN research, plus Statistical Binary Quantization and label-based filtered search. Indexes are designed to live on SSD rather than RAM, giving high recall and throughput at a lower cost than in-memory HNSW for large tables. It keeps pgvector's vector type and query syntax, so adopting it is mostly a matter of creating a different index. It is PostgreSQL-licensed and preinstalled on Tiger Cloud.
Strengths & trade-offs
Strengths
- +DiskANN-style index keeps vectors on SSD instead of RAM
- +Statistical Binary Quantization improves accuracy over plain binary quantization
- +Label-based filtering integrated into the index
- +Drop-in on top of pgvector types and operators
- +Permissive PostgreSQL licence
Trade-offs
- −Not available on most hyperscaler managed Postgres
- −Index builds are still slower than IVF-based approaches
- −Feature set narrower than dedicated vector databases
- −Release cadence slowed between 0.9.0 and 0.9.1
Great for
- +Large pgvector tables that no longer fit HNSW in RAM
- +Cost-efficient RAG on Postgres
- +Filtered semantic search with labels
Think twice for
- −RDS/Cloud SQL users who cannot install extensions
- −Tiny datasets
- −Workloads needing BM25 hybrid ranking out of the box
Capabilities
Native or official first-party support · 15 of 25 tracked
Scale & consistency
- Comfortable data size
- Large · ≤ 1 PB
- Horizontal scaling
- None
- Multi-region
- Read replicas
- Transactions
- Full ACID
- Consistency
- Strong
- Schema
- Strict
- Architecture
- Single node, Primary replica, Client server
- Query languages
- SQL
Single Postgres primary; SSD-resident DiskANN indexes handle tens to hundreds of millions of vectors per node.
Deployment
available on · AWS · Azure
Ecosystem
Alternatives to pgvectorscale
The standard Postgres extension for vector similarity search with HNSW and IVFFlat indexes.
pgvectorscale vs pgvector →Disk-friendly Postgres vector extension using IVF + RaBitQ, the successor to pgvecto.rs.
pgvectorscale vs vectorchord →PostgreSQL extension adding hypertables, columnar compression and continuous aggregates for time series.
pgvectorscale vs timescaledb →Elasticsearch-quality BM25 search and fast aggregations inside Postgres via the pg_search extension.
pgvectorscale vs paradedb →Postgres vector extension built on USearch HNSW; development has stalled since late 2024.
pgvectorscale vs lantern →Rust vector search engine with rich payload filtering, quantization and a managed cloud.
pgvectorscale vs qdrant →pgvectorscale FAQ
What is pgvectorscale?
pgvectorscale is a vector database by TigerData (formerly Timescale), first released in 2024. pgvectorscale is an open-source PostgreSQL extension from TigerData (formerly Timescale) that complements pgvector with a StreamingDiskANN index, based on Microsoft's DiskANN research, plus Statistical Binary Quantization and label-based filtered search. Indexes are designed to live on SSD rather than RAM, giving high recall and throughput at a lower cost than in-memory HNSW for large tables. It keeps pgvector's vector type and query syntax, so adopting it is mostly a matter of creating a different index. It is PostgreSQL-licensed and preinstalled on Tiger Cloud.
Is pgvectorscale open source?
Yes. pgvectorscale is open source. License: PostgreSQL License.
Does pgvectorscale support ACID transactions?
Yes. pgvectorscale supports full ACID transactions, with strong consistency.
Does pgvectorscale scale horizontally?
pgvectorscale does not scale horizontally; it runs on a single node, so you scale it up rather than out. It is comfortable at large scale (≤ 1 PB). It supports cross-region read replicas.
How can I run pgvectorscale?
You can self-host it. Managed offerings include Tiger Cloud. It is available on AWS and Azure.
What is pgvectorscale best for?
pgvectorscale is a strong fit for large pgvector tables that no longer fit HNSW in RAM, cost-efficient RAG on Postgres and filtered semantic search with labels. Think twice for RDS/Cloud SQL users who cannot install extensions.
What are the best alternatives to pgvectorscale?
Common alternatives to pgvectorscale include pgvector, VectorChord, TimescaleDB, ParadeDB and Lantern. The right choice depends on your workload, scale and hosting; WhichDB's smart match ranks them against your requirements.
How popular is pgvectorscale?
pgvectorscale ranks #243 of 844 databases on the WhichDB Score (13.9/100), and #17 among vector databases. The score combines public adoption signals such as package downloads, GitHub activity, job posts and developer discussion.