starter.wikiSearch⌘K

PostgreSQL vector extension

pgvector

An open-source PostgreSQL extension for storing vectors and running exact or approximate nearest-neighbor queries.

Visit official site

Editorial verdict

Choose pgvector when keeping vectors, metadata, joins, transactions, and application records together removes more complexity than a dedicated vector service would, and representative recall and latency tests pass.1

Best for

  • PostgreSQL applications with moderate vector workloads and strong data co-location value
  • Teams that want SQL joins, transactions, and metadata filters beside vectors
  • Organizations prepared to tune indexes, memory, queries, replication, and database capacity
1

Not ideal for

  • Teams expecting a separately scaling managed vector service from the extension itself
  • Workloads whose vector indexing or query load causes unacceptable database contention
  • Systems requiring dedicated-service operations, regions, or vendor support not supplied by the PostgreSQL provider
1
Main trade-off

Co-location avoids a second vector system and synchronization path, but vector indexing, recall, performance, and resource contention become PostgreSQL responsibilities.1

Product boundary

Whether co-locating vectors with PostgreSQL records is more valuable than adopting a dedicated vector service.

pgvector is an extension installed into PostgreSQL, not a hosted database vendor or complete managed service. Hosting, backups, availability, replication, scaling, monitoring, and support come from the chosen PostgreSQL operating model.1

For: PostgreSQL applications with vector workloads that fit the database's capacity, index, latency, filtering, and operating constraints

  • Vector load competes materially with transactional PostgreSQL workloads.
  • Required recall, filtering, update, and latency targets fail representative tests.
  • The PostgreSQL provider does not support the required extension version or operating controls.
  • Independent vector scaling and a vector-specific managed SLA matter more than co-location.

Why teams consider pgvector

  • Data co-locationVectors can live with relational records and metadata in PostgreSQL.1
  • Exact and approximate searchThe extension documents exact search plus HNSW and IVFFlat approximate indexes.1
  • PostgreSQL semanticsQueries can use SQL filters, joins, transactions, and the surrounding PostgreSQL operating model.1

Pricing

pgvector is open-source extension software with no separate software license fee; total cost belongs to the selected PostgreSQL hosting or self-operated infrastructure, storage, memory, replicas, backups, transfer, monitoring, and engineering work.12

PostgreSQL estimate

Extension plus chosen database

Estimate database compute, memory, storage, replicas, backups, transfer, monitoring, index build, maintenance, and support; do not invent a pgvector subscription.12

Software boundary
No separate pgvector vendor subscription12
Cost owner
The chosen PostgreSQL operating model and vector workload12
Pricing checked View official pricing

pgvector vs alternatives

Pinecone

Choose when
Choose Pinecone when a fully managed vector-specific service and independent scaling matter more than SQL co-location.
Compared with pgvector
Vendor-specific APIs and usage economics add a separate data system.35

Qdrant

Choose when
Choose Qdrant for a vector-native engine with self-hosted and managed Cloud routes.
Compared with pgvector
The team must operate or purchase a separate vector system.68

Weaviate

Choose when
Choose Weaviate when its object-and-vector model, hybrid search, integrations, and deployment options fit the product.
Compared with pgvector
The platform and billing surface are broader than a PostgreSQL extension.1012

Resources and sources

Official product, deployment, and pricing

  • pgvector source and documentation
    Open
  • pgvector license
    Open

Related tasks

  1. 1
    pgvector source and documentation

    pgvector · Accessed Official

  2. 2
    pgvector license

    pgvector · Accessed Official

  3. 3
    Pinecone overview

    Pinecone · Accessed Official

  4. 4
    Understanding Pinecone cost

    Pinecone · Accessed Official

  5. 5
    Pinecone pricing

    Pinecone · Accessed Official

  6. 6
    Qdrant documentation

    Qdrant · Accessed Official

  7. 7
    Qdrant Cloud billing

    Qdrant · Accessed Official

  8. 8
    Qdrant pricing

    Qdrant · Accessed Official

  9. 9
    Qdrant license

    Qdrant · Accessed Official

  10. 10
    Weaviate Database documentation

    Weaviate · Accessed Official

  11. 11
    Weaviate Cloud billing

    Weaviate · Accessed Official

  12. 12
    Weaviate pricing

    Weaviate · Accessed Official

  13. 13
    Weaviate license

    Weaviate · Accessed Official