Hybrid search inside SurrealDB: one query, vector + keyword + RRF

By Archie Marshall - Applied AI Lead at Squad AI. Built during the LangChain × SurrealDB London Hackathon. TL;DR RAG systems fail when vector search returns semantically similar results instead of the exactly named function you asked for. The fix is hybrid search: vector + keyword in parallel, fused by Reciprocal Rank Fusion. SurrealDB's search::rrf() runs the whole thing inside the database — one query, no middleware, no score normalisation. With graph enrichment on top, a 3B local model handle