Hybrid Search Now Runs on the Proprietary Opensolr Hybrid Query Parser

· Search · New Feature · All updates

Hybrid search now runs on the Opensolr {!hybrid} parser. Every vector-enabled index fuses keyword (BM25) and semantic (vector) relevance through a proprietary Solr query parser that runs only on Opensolr clusters.

01 One parser, one fused score

Keyword and vector relevance are computed on different scales, and combining them after the fact is where naive hybrid setups go wrong. Fusing them inside a single parser gives correct score fusion rather than two ranked lists stitched together.

Query Keyword (BM25) Semantic (vector) {!hybrid} One fused ranked list
Both legs are scored and fused inside the parser, not merged afterwards by the application.

02 Real search-mode semantics

The parser exposes genuine modes rather than a single blended weight, so you can decide how much each leg is allowed to decide.

ModeBehaviour
UnionEither leg can contribute a result
Keywords RequiredThe keyword leg must match
Meaning RequiredThe semantic leg must match
IntersectionBoth legs must match

03 What you get from it

  • Natural-language queries. Questions phrased as people speak, rather than as keyword lists, work without special handling.
  • Cross-lingual queries. A query in one language finds content in another.
  • Semantic search over documents. Not only over short fields.
View the full changelog