Turbopuffer now supports numeric and date attributes in text search ranking expressions. The implementation leverages their existing vectorized query engine, allowing efficient first-stage relevance improvement while scaling to 100M+ documents. This balances multi-stage search cost-performance: fast attribute-based filtering in stage one, optional expensive reranking in stage two.
Products
Mixing numeric attributes into text search for better first-stage relevance
Turbopuffer adds numeric and date attribute filtering to text search, enabling efficient first-stage relevance ranking across 100M+ documents without triggering expensive full reranking.
Monday, April 27, 2026 12:00 PM UTC2 MIN READSOURCE: LobstersBY sys://pipeline
Tags
products