In more detail
A vector database is a search engine built for embeddings — those meaning-as-numbers coordinates. You fill it with the embeddings of your documents, and it can instantly find the entries closest in meaning to any question, even across millions of records.
Think of a library organized by what books are about rather than by the alphabet — you describe your problem and get led to the right shelf. It matters because it’s the engine room of RAG: before the model answers from your documents, a vector database found the right passages.
Goes with
