Vector Database

A database specifically designed to store and query high-dimensional vectors, often used for embeddings in RAG systems.

How it works

A vector database stores dense embedding vectors (typically 768 to 4096 dimensions) alongside metadata. When a query arrives, its embedding is computed, and the database performs approximate nearest-neighbour (ANN) search — using algorithms like HNSW or IVF-PQ — to find the stored vectors with the highest cosine or dot-product similarity. Unlike exact search (O(n) per query), ANN search returns results in sub-millisecond time for millions of vectors by building index structures.

Why it matters

Vector databases are the backbone of production RAG systems. They enable semantic search: finding documents that mean the same thing rather than just containing the same keywords. This is transformative for enterprise knowledge management, customer support systems, and legal/medical research tools. Popular options — Pinecone, Weaviate, Qdrant, ChromaDB, pgvector — have proliferated rapidly as LLM applications have scaled, making vector database proficiency a critical backend engineering skill.