Tiger Data Adds Native Search to Google Cloud PostgreSQL

Tiger Data has announced that its open-source pg_textsearch extension is now natively supported in Google Cloud’s AlloyDB and Cloud SQL for PostgreSQL. This integration introduces industry-standard BM25 ranking directly into two of the most widely used managed database services. The move eliminates the need for developers to maintain separate search infrastructure, which has traditionally been a complex and resource-intensive task.

According to HPCwire, this development marks a significant shift in how full-text search is handled within PostgreSQL ecosystems. By embedding the search engine directly into the database layer, teams can now achieve accurate relevance ranking without the operational overhead of managing external sidecar systems. The extension is currently in public preview, with general availability expected later this year.

Native integration simplifies database operations

Traditional full-text search in PostgreSQL often degrades in performance as data scales. The pg_textsearch extension addresses this by using BM25 scoring, which applies mathematical properties like inverse document frequency and term frequency saturation to improve ranking accuracy. Because the indexes are stored in standard PostgreSQL pages, they integrate seamlessly with existing database tools such as WAL logging, replication, and VACUUM.

This native approach removes the need for external storage or special backup procedures. Developers no longer have to manage the synchronization between a separate search cluster and their primary database. Mike Freedman, CTO of Tiger Data, noted that this architecture outperforms wrapping external libraries and reduces the complexity associated with maintaining disjointed systems.

Hybrid search capabilities accelerate vector operations

The integration also enables seamless hybrid search by combining keyword-based results with semantic vector data. AlloyDB and Cloud SQL customers can now use pg_textsearch alongside pgvector to merge different types of search results in a single query. This is achieved using Reciprocal Rank Fusion, a method that combines multiple ranking lists into a single, more relevant output.

Additionally, the platform supports ScaNN vector indexing, which accelerates vector search speeds by up to six times compared to standard HNSW methods. This performance boost is particularly useful for applications that rely heavily on semantic understanding. By handling both keyword and vector search within the same database engine, teams can streamline their architecture and reduce latency.

Trade-offs between preview and stability

While the performance benefits are significant, the feature is currently in public preview. This means that while it is available for testing and early adoption, it may still undergo changes before reaching general availability. Organizations planning to build critical applications on this stack should monitor for updates and potential adjustments in the coming months.

The trade-off for gaining this native efficiency is the reliance on a specific database version and cloud provider. pg_textsearch supports PostgreSQL 17 and 18, and its native integration is currently specific to Google Cloud’s managed services. Teams using other cloud providers or older PostgreSQL versions may still need to rely on external search solutions, creating a potential divergence in infrastructure strategies across different environments.