pgvector
Open-source PostgreSQL extension for vector similarity search that keeps embeddings and scraped data in the same database.
pgvector is an open-source PostgreSQL extension for vector similarity search. It helps teams store embeddings alongside scraped content, so retrieval stays close to the data they already manage.
Key Features:
- PostgreSQL Extension - Add vector similarity search without moving data out of PostgreSQL.
- Unified Retrieval Store - Keep embeddings and scraped content in one database for search and retrieval.
- Built in C - Implemented in C as a native PostgreSQL extension.
- Database-First Use - Use it through PostgreSQL rather than a separate SDK package.
Web scraping workflow
How to use pgvector in a web scraping pipeline
pgvector adds vector storage and similarity search to PostgreSQL. It can make scraped content searchable by meaning while keeping source metadata and ordinary application tables in the same database.
A practical workflow
- A separate crawler extracts clean text and splits it into chunks identified by page and content version.
- An embedding stage supplies vectors, which your loader stores alongside chunk text and source metadata in PostgreSQL tables using pgvector.
- Query by vector similarity and join the matches to page records so the application can display source URLs and collection dates.
When it fits
Use it when PostgreSQL already stores the collected records and adding semantic retrieval there avoids a separate retrieval system. Evaluate exact versus approximate search against your corpus and relevance requirements.
Boundaries
pgvector is a database extension, not an embedding model, crawler or hosted search product. You must manage model consistency, indexes and content refreshes. Updating a page also requires a policy for replacing its old chunks; similarity scores alone do not certify that a retrieved passage answers the question.
Growing a website? Track its keywords, Google rankings, Search Console insights, and share of voice with bisibility. Free during beta.
Our productRepository
pgvector/pgvector
Stars
23.4KForks
1.4KLast commit
2026-10-01License
PostgreSQLLanguage
C
Metrics recorded 2026-10-16
Repository details
Last push: 2026-10-15
License source: Repository license file
Frequently asked questions
What is pgvector?
pgvector adds vector storage and similarity search to PostgreSQL. It supports exact nearest-neighbor search and approximate indexes such as HNSW and IVFFlat, while queries can use PostgreSQL tables and SQL.
Sources (checked ): pgvector documentation
How do I install pgvector?
Install the extension for your PostgreSQL server and platform using a supported package or the documented build procedure. Then enable it in each target database with CREATE EXTENSION vector. Extension installation and enabling it in a database are separate steps.
Sources (checked ): pgvector documentation
Is pgvector free or paid?
pgvector is available under the PostgreSQL License without an extension subscription fee. Hosting and operating the PostgreSQL server still cost money, and a managed provider can charge for its service.
Sources (checked ): pgvector license; pgvector documentation
What are some alternatives to pgvector?
Qdrant and Chroma also provide vector retrieval, with database or library interfaces separate from a PostgreSQL extension. Compare deployment, filtering, indexing and how vector data relates to your existing application tables.
Sources (checked ): pgvector documentation; Qdrant documentation; Chroma documentation
What are the key differences between pgvector and Pinecone?
pgvector runs inside a PostgreSQL database you operate or obtain from a hosting provider. Pinecone provides a managed vector-database service with its own APIs. Choose based on whether retrieval should share your PostgreSQL data and operations or use a separate managed service.
Sources (checked ): pgvector documentation; Pinecone documentation
Sources
Verified against the pgvector/pgvector repository.