Definition
Vector search retrieves documents based on embedding similarity rather than exact keyword matches. A query is embedded and compared against pre-computed document embeddings; the closest matches are returned. Powers modern RAG systems and semantic search. Requires a vector database (Pinecone, Weaviate, Qdrant, Postgres pgvector) for production scale.
Example
Query 'cancel plan' returns docs containing 'end subscription', 'stop service', 'unsubscribe' even without keyword overlap.
When to use
RAG retrieval, semantic search, recommendation engines, similarity matching at scale.