aicoolies logo
pgvector PostgreSQL parent mark
pgvector PostgreSQL parent mark

pgvector

Vector similarity search for PostgreSQL

open sourceverified Jul 6, 2026
Visit Website →
ShareShare on XShare on LinkedInShare on Reddit

pgvector is an open-source PostgreSQL extension with 22K+ GitHub stars adding vector similarity search to your existing Postgres database. Store embeddings alongside relational data, perform exact and approximate nearest neighbor search using L2, inner product, cosine, and L1 metrics. Supports HNSW and IVFFlat indexes for fast similarity queries at scale. Eliminates the need for a separate vector database by bringing vector capabilities into existing PostgreSQL infrastructure.

Read our pgvector review

A detailed review by the aicoolies team — click to read

pgvector adds vector similarity search to PostgreSQL. With 22K+ stars, it is the default choice for teams wanting vector search without adding a separate database.

Store embeddings as a native column type alongside relational data. Combine SQL filters with vector similarity search in a single query.

HNSW indexes for high-recall approximate search, IVFFlat for faster builds. L2, inner product, cosine, and L1 distance metrics.

Works with Supabase, Neon, AWS RDS, Google Cloud SQL, and self-hosted. Integrates with LangChain, LlamaIndex, and all AI frameworks supporting PostgreSQL.

Pricing

Free and open-source

Platforms

PostgreSQL extension

Categories

PostgreSQL ToolsVector DatabasesDatabases & ORMs

Tags

Multi-ModelAgenticBYOK

Use Cases

Database Schema DesignData Engineering

Alternatives

All pgvector alternatives →
TigerData logo

pg_textsearch

BM25 full-text search extension for PostgreSQL

pg_textsearch is a PostgreSQL extension from Timescale that adds BM25 relevance-ranked full-text search directly inside Postgres. Using the same ranking algorithm as Elasticsearch and Lucene, it provides search-engine quality results without requiring a separate search cluster — particularly valuable for developers building RAG pipelines on PostgreSQL who want semantic-quality ranking alongside pgvector.

Open Source
Tembo logo

Tembo

Managed Postgres platform with 200+ extensions as pre-built stacks

Tembo is a managed PostgreSQL platform that packages 200+ Postgres extensions into purpose-built stacks for specific workloads. Stacks include OLAP analytics, vector search, message queues, geospatial, and machine learning, turning PostgreSQL into a specialized database for each use case. Eliminates the need for separate Redis, Elasticsearch, or Kafka instances alongside Postgres.

freemiumOpen Source
Marqo logo

Marqo

Embedding-first search and discovery engine for AI-powered product experiences.

Marqo is an open-source tensor search engine that combines embedding generation and vector search in a single API, removing the need to manage separate embedding pipelines and vector databases. Built for product discovery and multi-modal search, it lets teams index text, images, and structured data together, returning ranked results based on semantic similarity rather than keyword overlap.

freemium

Related Tools

computed discovery: shared active categories · kept separate from editor-verified Alternatives

VexDB-Lite VexDB parent mark

VexDB-Lite

One vector-search extension across PostgreSQL, DuckDB and SQLite

MIT-licensed vector-search extension for PostgreSQL, DuckDB and SQLite that shares one graph-index core with PQ/RaBitQ quantization, persistent indexes and metadata filtering; SQLite packages cover Linux, macOS, iOS, Android and WASM, so it runs inside existing databases instead of as a separate vector service.

Open Source
Cloudflare logo

Cloudflare Vectorize

Edge-native vector database for Workers and AI applications

Cloudflare Vectorize is Cloudflare’s managed vector database for Workers and edge AI applications. It is distinct from the existing Cloudflare Workers tool page: Workers is the compute runtime, while Vectorize is the embedding index and vector-query layer used to add semantic retrieval to Cloudflare-hosted apps.

freemium
Upstash Vector logo

Upstash Vector

Serverless vector database with pay-as-you-go API pricing

Upstash Vector is a managed serverless vector database for RAG, semantic search, and embedding lookup. It is separate from the existing Upstash platform record in the aicoolies catalog: this slug covers the Vector product line, not the broader Redis, Kafka, or QStash platform.

freemium
OpenSearch logo

OpenSearch

Open-source search engine with vector and hybrid retrieval

OpenSearch is an Apache-2.0 distributed search engine with native vector-search support for teams that want BM25, filters, aggregations, and k-NN retrieval in the same search stack. It is distinct from Elasticsearch in the aicoolies catalog: OpenSearch is the AWS-backed open fork with its own docs, plugin path, and serverless deployment options.

Open Source
Supabase MCP logo

Supabase MCP

MCP server for connecting AI assistants to Supabase projects

Supabase MCP is Supabase's Apache-2.0 server for connecting AI assistants to Supabase projects. It can expose database, configuration, and project-management workflows to MCP clients such as Cursor, Claude, and Windsurf, while the official docs emphasize permission and security review before production use, SQL changes, or high-privilege database access.

Open SourceTelemetry
Deep Lake logo

Deep Lake

AI data runtime for multimodal datasets and vector search

Deep Lake is an open-source AI data runtime from Activeloop for storing, versioning, and querying multimodal data and embeddings. It fits teams building RAG, training, evaluation, or dataset-heavy agent workflows that need a bridge between vector search, structured metadata, and large image, text, audio, or video collections.

Open Source

Comparisons

Milvus vs pgvector: Which Vector Database Wins in 2026?

For most teams building RAG apps or MVPs, pgvector is the stronger default: it adds vector search to the Postgres you already run, keeping embeddings beside relational data with no extra cluster to operate. Milvus is a purpose-built distributed vector database that pulls ahead at massive scale, hundreds of millions of vectors, very high QPS, and GPU-accelerated indexes. This comparison shows where each fits.

Milvuspgvector

Weaviate vs pgvector: AI-Native Hybrid Search or Postgres Simplicity?

Weaviate and pgvector can both support production RAG, yet their product boundaries are fundamentally different. Weaviate is an AI-native vector database with object and vector storage, BM25 and vector hybrid search, model-provider integrations, reranking, multi-tenancy, replication, and access-control features. pgvector is a PostgreSQL extension that adds vector similarity to the relational database many applications already use. For teams explicitly comparing the two to build a search or RAG platform, **Weaviate is the winner**. Its integrated hybrid retrieval, tenant-aware data model, modular vectorization, and production search controls reduce the amount of application glue required for a sophisticated retrieval service. pgvector remains the better minimalist option when vectors should stay beside existing relational data, but Weaviate wins the dominant search-platform intent.

Weaviatepgvector

Qdrant vs pgvector: Dedicated Vector Engine or Postgres-Native Search?

Qdrant and pgvector solve vector retrieval from opposite directions. Qdrant is a dedicated vector database with payload-aware filtering, dense and sparse retrieval, hybrid query composition, quantization, and a service API. pgvector extends PostgreSQL so embeddings live beside relational data and participate in SQL, transactions, joins, backups, access controls, and the rest of an existing Postgres operating model. For the broadest buyer group—application teams that already trust PostgreSQL—**pgvector is the winner**. It avoids a second data system, keeps transactional data and embeddings together, and turns vector search into an incremental database capability. Qdrant is the stronger specialist for greenfield retrieval services, complex payload filtering, or workloads that need a purpose-built vector engine, but most teams should exhaust the simpler Postgres-native path before adding another distributed service.

Qdrantpgvector

Chroma vs pgvector: AI Retrieval Database or Postgres-Native Vectors?

Chroma and pgvector solve the vector-search problem from opposite directions. Chroma is the better fit when AI retrieval should live in a specialized collection API with documents, embeddings, metadata, filters, and hosted vector or hybrid search options. pgvector is the better fit when vectors should live beside application data in Postgres with SQL, JOINs, ACID semantics, backups, point-in-time recovery, and familiar database operations. For the primary buyer intent, Chroma is our pick because it offers a focused retrieval layer; pgvector remains the better fit when PostgreSQL operations are the governing constraint.

Chromapgvector

pgvectorscale vs pgvector — Scaling PostgreSQL Vector Search

pgvectorscale and pgvector are not simple substitutes: pgvector is the standard PostgreSQL vector extension, while pgvectorscale builds on pgvector data with Timescale's StreamingDiskANN and filtered-search focus. For teams already committed to Postgres, the real choice is whether pgvector alone is enough or whether production RAG workloads need an additional scaling layer. This comparison separates default adoption, index performance, managed-Postgres constraints, and operational risk.

pgvectorscalepgvector

pgvector vs Pinecone — Postgres-Native RAG or Managed Vector Database?

pgvector and Pinecone answer the same RAG question from opposite directions: should your vectors live inside Postgres with the rest of your application data, or should you use a managed vector database built for search at scale? pgvector is simpler when your data model already belongs in Postgres. Pinecone is the stronger default when vector search becomes its own production workload with scaling, latency, and operations requirements.

pgvectorPinecone

VectorChord vs pgvector — Postgres Vector Search at Two Different Scales

VectorChord and pgvector are both Postgres extensions for vector search, but they answer different questions. pgvector is the simple, ubiquitous choice for adding vectors to Postgres at small to medium scale. VectorChord is the engineered answer for teams that need pgvector-style operations at billion-vector scale — the spiritual successor that picks up where pgvector hits its limits.

VectorChordpgvector

FAQ

What is pgvector?

pgvector is an open-source PostgreSQL extension with 22K+ GitHub stars adding vector similarity search to your existing Postgres database. Store embeddings alongside relational data, perform exact and approximate nearest neighbor search using L2, inner product, cosine, and L1 metrics. Supports HNSW and IVFFlat indexes for fast similarity queries at scale. Eliminates the need for a separate vector database by bringing vector capabilities into existing PostgreSQL infrastructure.

Is pgvector free?

Yes — pgvector is open source and free to use. Free and open-source

Is pgvector open source?

Yes — pgvector is open source.

Is pgvector still maintained?

Yes — pgvector is active. Its listing was last verified on July 6, 2026.

What are the best pgvector alternatives?

The top editor-verified pgvector alternatives are pg_textsearch, Tembo, Marqo.

How does pgvector score in our review?

Our hands-on review scores pgvector 86/100 overall, based on speed, privacy, and developer-experience testing.