Hybrid search
Hybrid search combines keyword and semantic search, merging both rankings into one result list instead of running one method alone.
Category
Industry concept
See all terms
What is hybrid search?
Hybrid search runs a keyword search and a semantic search against the same query at the same time, then merges both result lists into one ranked list instead of relying on either method alone. Keyword search catches exact terms, a product code, a name, an acronym, that carry little meaning for a semantic model to match on, while semantic search catches a rephrased question that shares no words with the right answer.
This solves a real gap: neither method alone covers every kind of query well, keyword search misses paraphrases and semantic search misses exact identifiers, so a system built on just one of them reliably fails the queries the other one was built for.
How does hybrid search differ from running two searches separately?
Running a keyword search and a semantic search as two separate steps still leaves the merging problem unsolved: something has to decide which result from which list goes first, and doing that by hand, or not doing it at all, usually means one method's results get ignored. Hybrid search treats the merge as part of the search itself, combining both rankings into a single result list automatically, so an exact match and a conceptual match get compared.
How does hybrid search work?
A common way to combine two ranked lists is reciprocal rank fusion: each result gets a score based on its position in each list and scores get summed across lists and re-sorted. A document ranked first by keyword search and eighth by semantic search still scores well overall, so a strong result from either method surfaces near the top instead of only the ones both methods agreed on.
Qontext exposes hybrid search as one of the context repository's core read operations, alongside grep and find, so an agent gets one merged, ranked result instead of running two separate searches and deciding for itself how to combine them.