A vector database stores embeddings and answers nearest-neighbour queries: “which stored vectors are closest to this one?” Since closeness means similar meaning, this is semantic search, matching by concept rather than keyword.
To stay fast over millions of vectors it uses approximate-nearest-neighbour indexes (e.g. HNSW) that trade a little accuracy for large speed-ups. It’s the retrieval engine behind RAG, recommendation, and deduplication.