vectorize

pg_vectorize : The simplest way to do vector search on Postgres

Overview

Packages

Source

pig build pkg pg_vectorize;		# build rpm/deb

Install

Make sure PGDG and PIGSTY repo available:

pig repo add pgsql -u   # add both repo and update cache

Install this extension with pig:

pig install pg_vectorize;		# install via package name, for the active PG version
pig install vectorize;		# install by extension name, for the current active PG version

pig install vectorize -v 18;   # install for PG 18
pig install vectorize -v 17;   # install for PG 17
pig install vectorize -v 16;   # install for PG 16
pig install vectorize -v 15;   # install for PG 15
pig install vectorize -v 14;   # install for PG 14

Config this extension to shared_preload_libraries:

shared_preload_libraries = 'pg_cron, vectorize';

Create this extension with:

CREATE EXTENSION vectorize CASCADE; -- requires pg_cron, pgmq, vector

Usage

• Sources: repo README, extension README, v0.26.1 release

vectorize is the PostgreSQL extension from pg_vectorize. Upstream documents two modes: a standalone HTTP service and the in-database SQL extension. For the packaged extension here, the SQL workflow is the relevant one.

Enable The Extension

ALTER SYSTEM SET shared_preload_libraries = 'vectorize,pg_cron';
ALTER SYSTEM SET cron.database_name = 'postgres';

CREATE EXTENSION vectorize CASCADE;

The extension README lists pg_cron, pgmq, and pgvector as dependencies, plus vectorize.embedding_service_url for the embedding service.

Create A Search Job

The high-level SQL API starts with vectorize.table():

SELECT vectorize.table(
  job_name    => 'product_search_hf',
  relation    => 'products',
  primary_key => 'product_id',
  columns     => ARRAY['product_name', 'description'],
  transformer => 'sentence-transformers/all-MiniLM-L6-v2',
  schedule    => 'realtime'
);

The extension README says this creates and maintains an embeddings column for the source table.

Search, RAG, And Direct Model Calls

Search with:

SELECT * FROM vectorize.search(
  job_name       => 'product_search_hf',
  query          => 'accessories for mobile devices',
  return_columns => ARRAY['product_id', 'product_name'],
  num_results    => 3
);

Upstream also documents:

• vectorize.rag() for retrieval-augmented answers.
• vectorize.generate() for text generation.
• vectorize.encode() for direct embedding generation.
• vectorize.import_embeddings() for loading precomputed vectors.

Update Behavior And v0.26.1 Note

The extension README says schedule => '* * * * *' checks for updates every minute, while schedule => 'realtime' creates triggers for immediate refresh on inserts and updates.

The v0.26.1 release note only says “update dependencies”, so there is no upstream user-facing SQL/API delta to document beyond the existing README surface.