Home / Hybrid search
Hybrid search guides
15 published guides. Start with What is hybrid search.
Hybrid search combines keyword and vector retrieval in one ranked list. See how fusion works, where it helps, when it fails, and how to test it.
Read the guide- Why keyword and vector search fail on different queries
- The four steps from query to results
- Three ways to merge the two result lists
- When hybrid search is worth the extra index
Learning path
All guides
Understand it4 guides
Start hereWhat is hybrid searchHybrid search combines keyword and vector retrieval in one ranked list. See how fusion works, where it helps, when it fails, and how to test it.ExplainerHow hybrid search worksThe query path through a hybrid system: parallel retrieval, score normalisation, fusion and reranking.Decision guideWhen to use hybrid searchThe query and data patterns where hybrid beats pure vector search, and the ones where it adds cost for nothing.ComparisonHybrid search vs vector searchMeasured differences between hybrid and vector-only retrieval and how to test which one your data needs.
Merge the results5 guides
ExplainerReciprocal rank fusionHow RRF merges ranked lists using positions only, the formula, the k constant and when it underperforms.ExplainerRelative score fusionHow relative score fusion normalises and blends raw scores, and when it beats rank-based fusion.ExplainerScore normalization in hybrid searchWhy keyword and vector scores must be put on one scale before blending, and how each normalization method changes the ranking.ExplainerScore fusion methodsRRF, relative score, convex combination and learned fusion compared with their failure cases.TutorialBalancing keyword and vector scoresChoosing the alpha weight from evaluation data instead of guessing, with normalisation pitfalls.
Build and run it4 guides
TutorialHybrid search tutorialA runnable hybrid search build: BM25 plus vectors, fusion, and an evaluation to prove it helped.TutorialHybrid search in SQL databasesHow to run keyword and vector search in one SQL query, with Postgres and other relational examples.ComparisonDatabases that support hybrid searchDatabases and search engines with native hybrid search compared on fusion methods, filters and limits.ExplainerHybrid search on distributed databasesHow distributed databases shard, fuse and rank hybrid queries, and what breaks as data grows.