AI coding glossary
Vector Search (Semantic Search)
Also known as: semantic search, embedding search, vector retrieval
In one sentence
Search that ranks results by semantic similarity using embeddings, finding documents that mean the same thing as the query, not just documents that contain the same words.
Full definition
Vector search ranks documents by semantic similarity computed against embeddings, typically high-dimensional vectors produced by an embedding model. Query and document each get embedded; cosine similarity ranks the matches. It complements (and often replaces) classic BM25 keyword search because it handles paraphrase, synonyms, and intent. The 2026 best practice is hybrid search: combine BM25 (lexical) and vector (semantic) into a single re-ranked result list. Common 2026 stacks: Postgres + pgvector, Qdrant, Pinecone, Weaviate, ParadeDB (BM25 + vector in Postgres). Vector search underpins RAG, recommendation, dedup, and semantic clustering.