What it is: MongoDB now supports native hybrid search through two aggregation stages, $rankFusion and $scoreFusion, which combine the results of full-text and vector search queries into a single ranked result set. $rankFusion merges results using reciprocal rank fusion, which ranks documents by their position across each result set; $scoreFusion combines normalized relevance scores using configurable per-pipeline weights.

Who it's for: Developers building retrieval-augmented generation (RAG), semantic search, and recommendation systems who need both the keyword precision of full-text search and the semantic relevance of vector search within a single query, and who previously had to author and maintain complex aggregation logic to combine the two.

Why it matters: Hybrid search runs natively against existing MongoDB Atlas Search and Vector Search indexes, requiring no new index type and no reindexing. Result sets that previously required chained $unionWith and $group stages with manual score normalization now resolve to a single, composable pipeline stage.

How to get started: Define $vectorSearch and $search queries as named sub-pipelines within a $rankFusion or $scoreFusion stage, then set per-pipeline weights. Review the documentation below for a step-by-step example.