A database optimized for storing and searching embeddings (vectors). Instead of matching exact keywords like a traditional database, vector databases find the most semantically similar items. You ask "how to fix a memory leak" and it returns documents about "debugging RAM consumption" because the embeddings are close.
Why it matters
Vector databases are the storage layer that makes RAG work. Without them, you'd need to embed your entire knowledge base on every query. They're also the backbone of recommendation systems and semantic search.