AI glossary
Hybrid search
Hybrid search runs a keyword search and a vector search on the same query and merges the two result lists, so each covers the other's blind spot.
Keyword search, such as BM25, is good at exact terms like "ERR_42"; vector search is good at paraphrase like "the app will not start". Running both and merging finds documents either one alone would miss.
A common way to merge is reciprocal rank fusion: each result scores by its position in each list, so you do not have to compare two very different kinds of score.
Measure before and after. Hybrid search usually helps, but the gain depends on your questions, and a test set of real queries is the only way to know.
Learn it hands-on
These free lessons teach it with commands to run in a lab and checked tasks. Read them whole with no account.