pg_search

pg_search is a PostgreSQL extension developed by ParadeDB that provides BM25-based full-text search capabilities directly within PostgreSQL databases.

Reviewed by 7wData

On this page

Publisher review

pg_search is a PostgreSQL extension developed by ParadeDB that provides BM25-based full-text search capabilities directly within PostgreSQL databases. As a component of the broader ParadeDB platform rather than a standalone product, pg_search enables developers to implement search functionality without external search infrastructure like Elasticsearch. The extension uses the Tantivy search engine (a Rust-based alternative to Apache Lucene) and supports features including relevance ranking via BM25 scoring, phrase matching, fuzzy matching, faceted search, and hybrid search capabilities when combined with pgvector embeddings.

Since it operates as a native PostgreSQL extension using the @@@ operator, search queries remain transactionally consistent with the underlying data. However, adoption is declining: as of March 19, 2026, Neon deprecated pg_search for newly created projects, recommending alternatives like tsvector, pg_trgm, pgvector, or lakebase_text instead. The extension trades operational simplicity for increased resource pressure on the database itself—running search and analytics workloads inside PostgreSQL consumes additional CPU, memory, and I/O that previously would have been isolated in a dedicated search engine. For organizations already managing PostgreSQL infrastructure, pg_search eliminates external dependencies but may not be suitable for write-heavy or performance-sensitive environments.

Get the AI & data signal, daily.

335k+ subscribers read this every morning. One email, both newsletters. Unsubscribe anytime.

How it works

  1. BM25 relevance ranking

    Scores and ranks results using the Best Matching 25 algorithm, similar to Elasticsearch, without requiring external infrastructure.

  2. Real-time index updates

    Automatically updates search indexes as rows are inserted, updated, or deleted—no manual reindexing required.

  3. Flexible tokenization and multi-language support

    Supports multiple tokenizers and language-aware token filters for improved accuracy across different languages.

  4. Hybrid search with pgvector integration

    Combines BM25 full-text scores with vector embeddings for semantic search in a single query.

  5. Faceted search and aggregations

    Enables categorization, filtering, and aggregation of search results for navigation and analytics.

  6. Elasticsearch-like JSON query DSL

    Provides expressive query syntax alongside standard SQL, familiar to developers experienced with Elasticsearch.

Strengths and trade-offs

Strengths

  • No external infrastructure required—search stays within PostgreSQL, reducing operational complexity.
  • ACID-compliant transactions ensure data consistency between search indexes and transactional data.
  • 20x faster indexing than PostgreSQL's native tsvector on million-row datasets.

Trade-offs

  • Neon deprecated pg_search for new projects in March 2026, signaling vendor deprioritization and uncertain long-term viability.
  • Increased CPU, memory, and I/O pressure on the database; not suitable for write-heavy or performance-sensitive workloads.
  • AGPL licensing requires careful evaluation for proprietary software; Enterprise tier available only through custom sales channels.

Pricing context

pg_search is part of ParadeDB, which operates on a freemium model. ParadeDB Community (including pg_search) is free and open-source under AGPL-3.0 license for self-managed single-node deployments with community support. ParadeDB Enterprise adds read replica support, high availability, and dedicated SLA with custom pricing available through sales@paradedb.com. ParadeDB Cloud (fully managed), currently in private beta, will include all Enterprise features with managed infrastructure and one-click deployment.

Alternatives

User reviews

No user reviews yet. Be the first to write one.

Sources

Reporting on this tool draws on these publicly available sources.

  1. github.com — pg_search is a component of ParadeDB; technical requirements (PostgreSQL 15+) and development context.
  2. neon.com — Neon's March 19, 2026 deprecation of pg_search for new projects; recommended alternatives (tsvector, pg_trgm, pgvector, lakebase_text).
  3. docs.paradedb.com — pg_search installation, system requirements, production recommendation for Enterprise tier, and deployment models.
  4. www.paradedb.com — pg_search capabilities: 20x faster indexing than tsvector, BM25 relevance, fuzzy matching, real-time indexing, performance comparisons with Elasticsearch.
  5. www.paradedb.com — ParadeDB freemium model: Community (free, open-source), Enterprise (custom pricing), Cloud (private beta); GitHub adoption metrics (1M+ Docker deployments, 200k+ extensions installed).