Vector databases let AI tools search large collections of documents by meaning at speed. Options include Pinecone, Weaviate and Qdrant, as well as vector search features in PostgreSQL through pgvector, Azure AI Search and Amazon OpenSearch Service.
For example, a knowledge assistant might store embeddings of policy documents and contracts, then retrieve the most relevant passages for each question. The misconception is that every AI project needs a dedicated vector database. For modest volumes, the vector features in a database you already run are often enough, which means one less system to secure, pay for and maintain.