Vector database
What is Vector database?
A database built to store embeddings and find the closest matches fast, instead of looking for exact keyword hits. Ask it for the items most similar in meaning to your query and it returns them from millions in milliseconds. It is the retrieval layer under most semantic search and RAG systems.
Why it matters
The pitch for a vector database is speed at scale, but the more useful question for most companies is whether they need a separate one at all. For a few thousand documents, similarity search can run inside a database you already operate, and adding a specialist system just gives you another thing to host, secure and keep in sync. The case for a dedicated one grows with volume: millions of items, heavy query traffic, the need to filter by metadata and search by meaning at once. Get that judgement wrong in either direction and you either hit a wall as you grow or pay to run infrastructure your data never justified.
In practice
A support team wires a chatbot to its help articles and starts with a few hundred of them, small enough that a plugin on the existing database handles the search. As the archive grows into the tens of thousands and answers slow down, they move to a dedicated vector store built for that load. The trigger to switch was measured volume, not the assumption that serious AI demands a specialist database from day one.
