What is
Hybrid Search?
A search technique combining keyword (BM25) and semantic (vector) retrieval.
Definition
Hybrid search combines two retrievers — a sparse keyword retriever (typically BM25) and a dense vector retriever — and fuses their results. Vector search captures meaning; keyword search captures exact identifiers, codes, names, and rare terms that embeddings sometimes miss. Reciprocal Rank Fusion (RRF) is the most common fusion method.
Example
A search for "ERR_CONN_RESET in Stripe webhook" benefits from hybrid retrieval — vector search finds related debugging articles, keyword search finds exact-match code references.
How Vedwix uses Hybrid Search in client work
Default in our RAG stack. We rarely ship pure vector search to production.
We ship this.
If you're building with Hybrid Search in production, we can help — from architecture review to full implementation.
Brief us