Vector Database

A vector database stores and retrieves embedding vectors efficiently, enabling similarity search for RAG and recommendation workflows.

It is a specialized store for finding nearest-neighbor vectors at scale.

Vector databases index embedding vectors for approximate nearest-neighbor search, often alongside metadata filters for permissions and hybrid retrieval.

They are infrastructure for semantic search and RAG at scale.

• RAG corpora
• Similar-item search
• Multimodal retrieval stores

• Fast similarity search
• Scales dense retrieval

• Index quality depends on embedding model choice
• Not a substitute for access control design

• A vector DB alone is not an AI product strategy

Do I always need a specialized vector DB?

Not always—some teams start with existing search stacks that add vector fields. Choice depends on scale and ops constraints.

• NIST AI RMF

• ANN search literature (HNSW and related methods)