Glossary
Vector Database
A database optimized for semantic similarity search over embeddings.
Definition
A vector database stores high-dimensional numerical embeddings — typically produced by an embedding model from text, images, or other data — and provides fast nearest-neighbor search by semantic similarity. They are the storage layer underneath most RAG systems.
Context
Vector databases became a category around 2022–2023 with dedicated products like Pinecone, Weaviate, Qdrant, and pgvector. By 2026, the line has blurred — most managed databases support vector indexes, and many RAG stacks combine vector and keyword search (hybrid retrieval) rather than relying on pure semantic similarity.