ParadeDB

// Overview

ParadeDB adds full-text search, vector retrieval and aggregations to Postgres through the pg_search extension. Instead of running a separate search engine next to your database and keeping the two in sync, your application data and your search index live in the same Postgres instance, and the index is updated in the same transaction as the write.

The search engine underneath is Rust: Tantivy powers full-text search and BM25 scoring, Apache DataFusion handles the analytical side, and pgrx bridges the whole thing into Postgres. Vectors are indexed through the pgvector extension today, which is why the install docs tell you to install pgvector first and to create the extension with CASCADE.

You query it with ordinary SQL — a search predicate and a score function in a normal SELECT — so filtering, joins, ACID guarantees and your existing Postgres tooling all still apply. ParadeDB Community is licensed AGPL-3.0; a separately-licensed ParadeDB Enterprise edition and hosted deployment options exist alongside it.

// What it does

• BM25-scored full-text search inside Postgres, with tokenizers, token filters, highlighting and top-K ordering
• Vector retrieval via pgvector, so keyword and embedding search sit behind the same SQL interface
• Aggregations, facets and bucket/metric queries backed by columnar storage
• Search indexes update transactionally with your writes — no ETL job or sync pipeline between two systems
• Built on Tantivy, Apache DataFusion and pgrx, with fixes contributed back upstream
• ORM integrations for Drizzle, Django, SQLAlchemy, Rails and EF Core, plus an MCP integration for AI agents

// Getting started

The fastest path is the install script, which pulls a ParadeDB Docker image and drops you into psql. To add pg_search to a Postgres server you already run, install the extension package and enable it.

curl -fsSL https://paradedb.com/install.sh | sh

curl -L "https://github.com/paradedb/paradedb/releases/download/v0.25.3/postgresql-18-pg-search_0.25.3-1PARADEDB-resolute_amd64.deb" -o /tmp/pg_search.deb
sudo apt-get install -y /tmp/*.deb

# in postgresql.conf
shared_preload_libraries = 'pg_search'

CREATE EXTENSION pg_search CASCADE;

CREATE INDEX search_idx ON mock_items
USING paradedb (id, description, rating)
WITH (key_field='id');

SELECT description, pdb.score(id)
FROM mock_items
WHERE description ||| 'running shoes' AND rating > 2
ORDER BY score DESC
LIMIT 5;

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

// When to use it

• Add relevance-ranked product or document search to an app whose data already lives in Postgres, without standing up Elasticsearch alongside it
• Retire a search cluster and the ETL pipeline that feeds it, so search results can never lag behind a write
• Build hybrid retrieval for a RAG pipeline where keyword matching, vector similarity and SQL filters are all one query
• Run faceted, aggregated search over a large table while keeping transactional guarantees on the same rows

// How ParadeDB compares

ParadeDB alongside other open-source vector databases tools AI/TLDR tracks, ranked by GitHub stars.