Vector database
Specialized storage for semantic similarity search, not exact word match. A pillar of RAG.
A vector database stores documents transformed into numeric vectors (embeddings) that capture their meaning. It enables semantic similarity search: finding "how to recover my password" even when the docs use "credential reset". Examples: Pinecone, Weaviate, pgvector (a PostgreSQL extension), Qdrant. It's the piece that makes RAG and intelligent search systems work. The choice depends on volume and whether you already use PostgreSQL — pgvector is usually the simplest option for mid-sized companies.