ParadeDB

ParadeDB is an open source Postgres extension that adds full-text search and analytics directly to your database — an Elasticsearch and Algolia alternative that gives you BM25 ranking, vector retrieval, and OLAP aggregations without running a second search system alongside Postgres. It ships as pg_search, a pure extension you install in your primary Postgres, so the search index stays in sync in the same transaction as your writes.

What is ParadeDB?

ParadeDB is a PostgreSQL extension that turns your existing Postgres into a full-text search and analytics engine. It embeds Tantivy — a Rust-based Lucene alternative — for a BM25 inverted index, and Apache DataFusion for columnar OLAP queries, so application data, search, vectors, and aggregations all live in one database. Because it is an extension rather than a fork or separate server, there is no ETL pipeline to babysit between Postgres and an external index.

What is ParadeDB best for?

ParadeDB is best for teams already running Postgres who want Elasticsearch-quality search without the operational cost of a second data store and the ETL that keeps it in sync. It fits insert- and update-heavy workloads where search results must be transactionally consistent and current, and datasets from hundreds of gigabytes up to tens of terabytes where native Postgres full-text search runs out of headroom.

What can ParadeDB do?

• Full-text search with BM25 relevance scoring, result highlighting, and multiple configurable tokenizers.
• Vector and hybrid search for semantic retrieval and RAG, alongside keyword search in the same query.
• Fast analytics over columnar storage powered by Apache DataFusion for aggregations and dashboards.
• Index that updates in the same transaction as your writes, so search is always consistent with your data — no sync lag.
• Standard SQL, JOINs, filtering, and faceting against indexed fields — no separate query DSL to learn.
• Deploys as a Postgres extension via Docker, Kubernetes, or hosts like Railway, Render, Fly.io, and DigitalOcean.

Where does ParadeDB fall short?

Running search and analytics inside your primary Postgres puts extra CPU, memory, and I/O pressure on the same instance that serves transactions — at large scale you may still want to isolate the search workload onto a replica, which requires the Enterprise edition.

Extension availability varies across managed Postgres providers, so you cannot always drop pg_search into an existing managed database; some hosts don’t allow custom extensions, pushing you toward self-hosting or a provider that supports it.

The community edition is single-node only — read replicas, high availability, and an SLA are Enterprise features — and a fully managed ParadeDB Cloud is still in beta behind a waitlist, so today it is primarily a self-hosted tool.

Is ParadeDB free?

Yes — the community edition is free and open source under AGPL-3.0 for single-node self-hosted deployments, with community support. ParadeDB follows an open-core model: an Enterprise license (custom pricing) adds read replicas, high availability, and an SLA, and a managed ParadeDB Cloud is on a waitlist. For a single Postgres instance, the free edition is fully functional.

What does ParadeDB replace?

ParadeDB is positioned as a transactional replacement for a dedicated search system bolted onto Postgres. It stands in for Elasticsearch and Elastic Cloud as a search-and-analytics engine, and for hosted search-as-a-service like Algolia — removing the per-record and per-search pricing and the ETL sync those setups require. If you’d rather run search as a standalone service, self-hosted options like Typesense and OpenSearch cover similar ground outside the database.

FAQ

Is ParadeDB open source? Yes. The core (pg_search) is released under AGPL-3.0 on GitHub, with an optional commercial Enterprise license for advanced features.

Can I self-host ParadeDB for free? Yes. The community edition is free forever for single-node deployments — install the extension in your own Postgres via Docker, Kubernetes, or a host like Railway or Fly.io.

Is ParadeDB a good Elasticsearch alternative? For teams already on Postgres, yes — it delivers BM25 search and analytics with transactional consistency and no ETL, which is exactly the sync pain Elasticsearch introduces. If you need a distributed, standalone search cluster independent of your database, a dedicated engine like OpenSearch may fit better.

What do I need to run ParadeDB? A PostgreSQL instance where you can install the pg_search extension — self-hosted, in Docker or Kubernetes, or on a managed host that permits custom extensions. It runs on standard Postgres, so no separate search infrastructure is required.