ParadeDB Releases pg_search v0.21.15 with Platform-Specific Postgres Packages

ParadeDB shipped pg_search v0.21.15 on March 14, 2026, and the headline change is straightforward but meaningful: the release introduced platform-specific Postgres packages, making it easier to pull the extension directly without building from source.

pg_search is a Postgres-native full-text search extension that runs Tantivy under the hood. Tantivy is a Rust-based search engine library, and pg_search bridges it into Postgres via pgrx, the framework that lets Rust code compile into a proper Postgres extension. The result is that you get Tantivy's indexing performance and BM25 scoring inside a standard Postgres database, without routing queries through a separate search service.

The v0.21.15 release assets on ParadeDB's GitHub repository and the corresponding PGXN entry both reflect the platform-specific packaging. For anyone who has wrestled with building pgrx-based extensions from source on different targets, pre-built platform packages cut down a non-trivial amount of friction. The pgrx build pipeline is not complicated once you know it, but it does require matching the right Rust toolchain, Postgres major version, and system libraries, and doing that repeatably across multiple deployment targets is exactly the kind of toil that pre-built packages eliminate.

ParadeDB has been positioning pg_search as a replacement for external search stacks, the argument being that if your data already lives in Postgres, adding a Tantivy-backed index should be cheaper operationally than standing up Elasticsearch or OpenSearch alongside it. Platform-specific packages make that pitch more credible, since the barrier to actually installing and testing the extension drops considerably when you can grab a binary matched to your Postgres version and architecture rather than compiling it yourself.

The practical implication for anyone running Postgres on multiple environments is that v0.21.15 is the version worth testing if you have been waiting for the packaging story to mature before committing to pg_search in a production pipeline.