Vector Database
Also known as: Vector Store · Embedding Database
A vector database stores content by meaning-similarity rather than exact matching, unlike a traditional database.
A traditional database finds a record when a value matches exactly, or with a wildcard. A vector database instead stores, for each piece of text, a so-called vector — a series of numbers describing its meaning relative to other texts — and uses that to find content that is similar in meaning even when no word matches. It exists because language is rarely repeated exactly, yet a search still needs to find the right passage.
In practice, every document, paragraph or product description is converted into such a vector when it's loaded in. A search query gets converted the same way, and the database returns the vectors closest to it — the passages most similar in meaning. That search step is exactly what a RAG system relies on to find matching documents before answering.
A common misunderstanding is treating a vector database as a replacement for a normal database. It's built for similarity search, not exact lookups like customer numbers or stock levels, and it will return the nearest matches even when none of them truly fit. Similarity is also no guarantee of correctness, so the results still need to be checked.
What it means in practice
For a business, a vector database isn't a product bought on its own — it's the technical building block behind a RAG project, once a company's own documents need to become searchable. What matters is loading in only the content that should be findable. NDVDL sets up such search systems as part of RAG projects.
Is this handled properly at your site?
We look at how it actually stands with you — and say honestly whether anything needs doing.
Related terms
All termsA term from your quote missing here?
Send us the passage you do not follow. We will explain it — with no obligation to order anything.