Infrastructure

Vector Database

A database optimized for storing and searching high-dimensional embeddings by similarity.

A vector database indexes embeddings — arrays of hundreds or thousands of floats that represent the meaning of text, images, or audio. Instead of exact-match queries, it returns the k nearest neighbors to a query vector, ranked by cosine similarity or dot product.

Popular options include Pinecone, Weaviate, Qdrant, Milvus, and pgvector (a Postgres extension). For most enterprise workloads under 10M vectors, pgvector on managed Postgres is the simplest and cheapest choice.

Vector DBs are the backbone of RAG, semantic search, recommendation systems, and anomaly detection.

Key points

  • HNSW is the dominant index algorithm — fast queries, higher memory
  • Hybrid search (vector + keyword) is essential for enterprise document retrieval
  • Metadata filters must be indexed separately for good performance
  • Reindexing on embedding-model upgrades can take hours — plan for it

Common use cases

RAG knowledge bases
Semantic product search
Duplicate and near-duplicate detection
Recommendation and personalization

Frequently asked

Related terms

Embedding

A numerical vector that captures the semantic meaning of a piece of content.

Retrieval-Augmented Generation (RAG)

A pattern that grounds LLM answers in your private documents by retrieving relevant snippets at query time.

Applying Vector Database in your business?

We help enterprises design, ship, and govern AI systems end-to-end.

Start AI Readiness Check Talk to an expert
Back to glossary