What is avector database?
The vector database is the search memory behind RAG: it finds the passages whose meaning best matches a question.
Get in touch
Let's talk about your project.
First we check whether the project fits your business model. Then you get a proposal with phases and effort.
Talk about a RAG project or call: +49 151 1576 5566A vector database stores chunks of text as long strings of numbers that roughly capture their meaning, and for a given question it finds the chunks closest to it in content. In RAG it is the place where the AI looks things up before answering.
- An embedding model turns every chunk into a string of numbers, and the question gets the same treatment.
- Many systems combine search by meaning with an ordinary keyword search.
- Without rights and roles, the AI may surface content the person asking was never meant to see.
How does a vector database find the right passage in your documents?
Preparation happens in two steps. First the documents are cut into chunks of a few paragraphs (chunking, in the jargon), then an embedding model turns each chunk into a string of numbers, and those numbers go into the vector database. When a question comes in later, it is converted the same way, and the database returns the chunks whose numbers sit closest to it.
That lets the search find passages that match the sense of a question without sharing its words. A part number like 4711-B, though, is often found more reliably by keywords than by meaning, which is why many systems combine the two; vendors call that hybrid search.
Keeping your vector database useful with up-to-date answers
What sits in a vector database is a snapshot. When a document changes it has to be re-indexed, otherwise the search keeps finding the old version, and if two versions sit side by side the answer may rest on the wrong one. Anyone running such a store should therefore also decide who keeps the collection maintained over time.
Why a vector database deserves as much protection as your documents
The OWASP project, which catalogues security risks in AI applications, lists weak access controls on vector databases as LLM08:2025, because otherwise the AI may reveal content the person asking was never meant to see. IBM raises a further point: under some circumstances the stored numbers can be traced back to the original text.
There is good news too. In its October 2025 guidance on RAG, Germany's Data Protection Conference notes that a rights and roles concept can be applied to the vector database and the stored documents, which is not possible inside the language model itself.
Frequently asked questions
Why does RAG need a vector database?
It is where the system searches your documents for the right passages before every answer. Without that store, the language model would only have what it learned in training.
Does ChatGPT use a vector database?
For developers, OpenAI documents a file search tool built on vector stores that works on exactly this principle. When the service searches uploaded files, it follows the same pattern.
Terms you should know in the same context
RAG (retrieval augmented generation) · Fine-tuning · LLM (large language model) · Local AI · Back to the AI glossary A to Z
A vector database is only as good as the collection you give it and only as safe as the permissions you set on it.