How to Use ParadeDB in PostgreSQL

How do I install and query ParadeDB in PostgreSQL?

ParadeDB extends PostgreSQL with vector, hybrid, and full-text search so you can build AI-ready queries inside the database.

Welcome to the Galaxy, Guardian!

Oops! Something went wrong while submitting the form.

Description

What is ParadeDB?

ParadeDB is an open-source PostgreSQL extension that bundles pgvector, HNSW indexing, and full-text search. It lets developers run lightning-fast semantic, keyword, and hybrid searches without leaving Postgres.

How do I install ParadeDB?

Install the extension with your package manager or Docker image, then run CREATE EXTENSION paradeb; in each target database. This registers custom index types and search operators.

How do I add vector search to my tables?

Create a vector column to store embeddings.Build an HNSW index for ANN search. ParadeDB adds <-> and <#> operators for distance calculations.

How can I combine full-text and vector search?

Store text in a tsvector column and embeddings in a vector column.ParadeDB’s hybrid_distance() function blends lexical and semantic scores for ranked results.

What are common ParadeDB use cases?

Popular scenarios include product recommendations, semantic customer support search, personalized marketing content, and AI chat retrieval over Orders and Products tables.

Which best practices speed up ParadeDB queries?

Use the same embedding dimension across tables, choose appropriate ef_search and m index parameters, and filter by metadata (e.g., category) before vector ranking to reduce candidates.

Can ParadeDB work with pgBouncer or read replicas?

Yes—extensions load once per session.Confirm the extension exists on every replica and use PRELOAD to avoid start-up overhead in pooled connections.

Why How to Use ParadeDB in PostgreSQL is important

How to Use ParadeDB in PostgreSQL Example Usage


-- Find 5 similar products to “wireless headphones” under $150
WITH params AS (
    SELECT 
        'wireless headphones'                 AS query_text,
        embed('wireless headphones')::vector AS query_vec  -- ParadeDB helper
)
SELECT p.id, p.name, p.price
FROM products p, params t
WHERE p.price < 150
ORDER BY p.embedding <#> t.query_vec   -- cosine distance
LIMIT 5;

How to Use ParadeDB in PostgreSQL Syntax


-- Enable extension
CREATE EXTENSION IF NOT EXISTS paradeb;

-- Example: Products table with text search & vector search
CREATE TABLE products (
    id           serial PRIMARY KEY,
    name         text,
    description  text,
    price        numeric,
    stock        int,
    embedding    vector(768),      -- OpenAI/CoHere size
    tsv          tsvector GENERATED ALWAYS AS (to_tsvector('english', name || ' ' || description)) STORED
);

-- HNSW index for vectors
CREATE INDEX idx_products_embedding ON products USING hnsw (embedding vector_l2_ops);

-- GIN index for full-text
CREATE INDEX idx_products_tsv ON products USING gin (tsv);

-- Hybrid search query skeleton
SELECT *, hybrid_distance(embedding, :query_vec, tsv, :query_text) AS score
FROM products
WHERE price < 100
ORDER BY score
LIMIT 10;

Common Mistakes

Mistake: Forgetting to run CREATE EXTENSION in each database. ParadeDB functions will be missing and queries will fail. Fix: connect to every database that needs vector search and execute CREATE EXTENSION paradeb.
Mistake: Using embeddings with mismatched dimensions across rows. ParadeDB’s vector type enforces fixed length, so INSERTs error out. Fix: always generate vectors with the same model and dimension (e.g., 768).