How to Install and Use ParadeDB in PostgreSQL

How do I install ParadeDB and run vector search in PostgreSQL?

ParadeDB is an open-source PostgreSQL extension that adds high-performance vector similarity search and hybrid retrieval to your database.

Welcome to the Galaxy, Guardian!
You'll be receiving a confirmation email

Follow us on twitter :)

Oops! Something went wrong while submitting the form.

Description

Example H2

Example H3

What does ParadeDB add to PostgreSQL?

ParadeDB integrates ANN vector search, hybrid BM25 ranking, and RAG-ready functions directly into Postgres, letting you build AI & search features without extra infra.

How do I install ParadeDB?

Run CREATE EXTENSION paradedb; after copying the compiled paradedb files into $PGHOME/lib and updating shared_preload_libraries. Restart the server to load the extension.

How do I create a vector index?

Use CREATE INDEX ... USING paradedb_ivfflat on a vector column.Choose dim, metric, and lists parameters for speed-accuracy trade-offs.

Example

CREATE INDEX idx_products_embeddings ON products USING paradedb_ivfflat (embedding vector_l2(1536)) WITH (lists = 100);

How do I perform a similarity search?

Use the <> operator with ORDER BY and LIMIT for nearest neighbors, or call paradedb.knn() for extra options like filtering.

Example search

SELECT id, name, price FROM products ORDER BY embedding <> paradedb.vector('\[0.12, ...]') LIMIT 5;

Can I mix keyword and vector ranking?

Yes.paradedb.hybrid_rank() combines BM25 scores from tsvector columns with vector distance, returning a single relevance score.

Hybrid query

SELECT p.id, p.name FROM products p CROSS JOIN LATERAL paradedb.hybrid_rank(p.embedding, p.search_tsv, 'wireless headphones') r ORDER BY r.score DESC LIMIT 10;

Best practices for ParadeDB

Start with 1–2K vectors per list for IVFFlat. Always VACUUM ANALYZE after bulk loads. Store vectors as FLOAT4[] for smaller disk usage.

Common use cases

Product recommendation, semantic customer support search, AI-powered dashboards, and real-time personalization are typical ParadeDB workloads inside ecommerce stacks.

Why How to Install and Use ParadeDB in PostgreSQL is important

How to Install and Use ParadeDB in PostgreSQL Example Usage


-- Recommend similar products to items in an order
SELECT oi.product_id AS source_product,
       p2.id           AS recommended_product,
       p2.name,
       p2.price
FROM orderitems oi
JOIN products p1 ON p1.id = oi.product_id
JOIN LATERAL (
    SELECT id, name, price
    FROM products
    WHERE id <> p1.id
    ORDER BY p1.embedding <&> embedding
    LIMIT 3
) p2 ON true
WHERE oi.order_id = 42;

How to Install and Use ParadeDB in PostgreSQL Syntax


-- Install extension (superuser)
CREATE EXTENSION IF NOT EXISTS paradedb;

-- Add vector column
ALTER TABLE products ADD COLUMN embedding vector(1536);

-- Create IVFFlat index
CREATE INDEX idx_products_embeddings
ON products USING paradedb_ivfflat (embedding vector_l2(1536))
WITH (lists = 100);

-- KNN search
SELECT id, name, price
FROM products
ORDER BY embedding <&> paradedb.vector('[0.11,0.9,...]')
LIMIT 5;

-- Hybrid rank with keyword filter
SELECT o.id, o.total_amount
FROM orders o
JOIN customers c ON c.id = o.customer_id
WHERE c.name ILIKE '%smith%'
ORDER BY paradedb.hybrid_rank(o.order_vector, o.order_tsv, 'recent high value') DESC
LIMIT 10;

Common Mistakes

Ignoring <code>shared_preload_libraries</code>. ParadeDB needs to be listed there; otherwise Postgres starts without the extension hooks and queries error. Add the library and restart.
Creating an index before bulk loading. IVFFlat indexes perform poorly if created before inserting most vectors. Load data first, then create the index and run <code>ANALYZE</code>.