VectorChord
activeby TensorChord Inc. · since 2024 · also known as vchord, pgvecto.rs successor
Disk-friendly Postgres vector extension using IVF + RaBitQ, the successor to pgvecto.rs.
Overview
VectorChord (vchord) is a PostgreSQL extension from TensorChord that replaces its earlier pgvecto.rs. It builds on pgvector's data types but uses IVF partitioning with RaBitQ quantization instead of in-memory HNSW, which makes index builds much faster and lets large indexes live mostly on disk with good recall. The companion VectorChord-bm25 extension adds BM25 ranking for hybrid search. It suits teams who started on pgvector and hit build-time or memory walls in the 50M to 1B vector range but want to stay in Postgres.
Strengths & trade-offs
Strengths
- +Index builds far faster than pgvector HNSW at large scale
- +RaBitQ quantization keeps recall high with low memory
- +Uses pgvector types, so migration is mostly an index swap
- +Companion BM25 extension for hybrid search
- +Full Postgres transactions and joins
Trade-offs
- −AGPL/ELv2 licensing is restrictive for some vendors
- −Not available on most hyperscaler managed Postgres
- −Smaller community than pgvector
- −Tuning IVF lists and probes adds a learning curve
Great for
- +Scaling Postgres vector search past pgvector's comfort zone
- +Cost-efficient disk-based RAG retrieval
- +Hybrid BM25 plus vector search in Postgres
Think twice for
- −Teams restricted to RDS or Cloud SQL
- −Tiny datasets where pgvector is enough
- −Licence-sensitive SaaS redistribution
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; IVF+RaBitQ indexes keep most data on disk, making 100M+ vectors on one node practical.
Deployment
Ecosystem
Alternatives to VectorChord
The standard Postgres extension for vector similarity search with HNSW and IVFFlat indexes.
vectorchord vs pgvector →TigerData's Postgres extension adding StreamingDiskANN indexes on top of pgvector.
vectorchord vs pgvectorscale →Elasticsearch-quality BM25 search and fast aggregations inside Postgres via the pg_search extension.
vectorchord vs paradedb →Postgres vector extension built on USearch HNSW; development has stalled since late 2024.
vectorchord vs lantern →Rust vector search engine with rich payload filtering, quantization and a managed cloud.
vectorchord vs qdrant →Cloud-native, distributed open-source vector database built for billions of vectors.
vectorchord vs milvus →VectorChord FAQ
What is VectorChord?
VectorChord is a vector database by TensorChord Inc., first released in 2024. VectorChord (vchord) is a PostgreSQL extension from TensorChord that replaces its earlier pgvecto.rs. It builds on pgvector's data types but uses IVF partitioning with RaBitQ quantization instead of in-memory HNSW, which makes index builds much faster and lets large indexes live mostly on disk with good recall. The companion VectorChord-bm25 extension adds BM25 ranking for hybrid search. It suits teams who started on pgvector and hit build-time or memory walls in the 50M to 1B vector range but want to stay in Postgres.
Is VectorChord open source?
Yes. VectorChord is open source. License: AGPL-3.0 or Elastic License 2.0 (dual).
Does VectorChord support ACID transactions?
Yes. VectorChord supports full ACID transactions, with strong consistency.
Does VectorChord scale horizontally?
VectorChord 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 VectorChord?
You can self-host it and use the vendor's managed cloud. Managed offerings include VectorChord Cloud.
What is VectorChord best for?
VectorChord is a strong fit for scaling Postgres vector search past pgvector's comfort zone, cost-efficient disk-based RAG retrieval and hybrid BM25 plus vector search in Postgres. Think twice for teams restricted to RDS or Cloud SQL.
What are the best alternatives to VectorChord?
Common alternatives to VectorChord include pgvector, pgvectorscale, ParadeDB, Lantern and Qdrant. The right choice depends on your workload, scale and hosting; WhichDB's smart match ranks them against your requirements.
How popular is VectorChord?
VectorChord ranks #250 of 844 databases on the WhichDB Score (13.7/100), and #18 among vector databases. The score combines public adoption signals such as package downloads, GitHub activity, job posts and developer discussion.