Skip to main content
Hybrid search uses vector search, a beta feature available in versions 0.25.0 and above.
Hybrid search combines the strengths of text search and vector search. Text search is strongest when the query terms and indexed text overlap. Vector search is strongest when the right result is semantically related but uses different words. Because those signals fail in different ways, combining them often produces a better result set than either one alone.

​
Fusing by Rank

BM25 scores and vector distances operate on different scales, so they cannot be added together directly. ParadeDB’s hybrid examples use Reciprocal Rank Fusion, which discards raw scores and uses only each row’s position, or rank, in each result list. With rank fusion, a row that ranks first contributes the same amount whether its BM25 score was 1 or 100. That makes text and vector results comparable without forcing them onto a shared scoring scale.