Glossary

Vector Database

A vector database stores and searches not strings and numbers, but vectors (embeddings) — numerical representations of the semantic content of text, images, or audio. It finds content that is "similar in meaning", not just exact matches.

How it works

Text → through an embedding model (OpenAI, Sentence Transformers) → a vector of hundreds or thousands of numbers. Texts similar in meaning have similar vectors. Search: find the N nearest vectors to the query vector (ANN: Approximate Nearest Neighbors).

Popular databases

Use cases

Semantic search, recommendation systems, RAG, duplicate detection, classification.