Language models & chatbots
A vector database is a storage system built to hold embeddings and quickly find the ones most similar to a query, which powers search by meaning.
Say you have turned millions of documents into embedding vectors. You need somewhere to keep them. You also need a fast way to ask 'which of these are closest to this new vector?' A normal database matches exact values. A vector database is built for this closest-match search, across huge collections, in milliseconds.
It does this with clever indexing. Instead of comparing the query to every stored vector, it checks only the promising ones. The results are approximate: it may occasionally miss the true closest match, in exchange for speed. Each stored vector usually carries extra data too, such as the original text and its source.
Vector databases are a key part of RAG systems and semantic search. As of 2026, options include dedicated products like Pinecone, Weaviate, and Milvus. Familiar databases such as PostgreSQL and Elasticsearch have added vector features too. Many small projects need nothing more than a simple in-memory library.
A company stores embeddings of every page in its internal wiki, so an employee typing 'how do I claim travel costs' instantly finds the expenses policy.