AI glossary · Data and knowledge
Vector database
In one sentence
A vector database stores embeddings, lists of numbers that represent the meaning of content, and finds the ones closest to a query very quickly. It's the search component in most RAG systems.
Also known as: vector DB · vector store · vector search database · embedding database
What is a vector database?
An ordinary database searches for exact values: the customer with this account number, the orders from March. A vector database searches by similarity: the ten documentation extracts closest in meaning to this question.
Put it to work
UTIO's AI, through an API
From €29/mo + VAT
For developers and agencies building their own products.
See the AI APIWhat it stores are embeddings: vectors of hundreds or thousands of numbers, one for each extract of text, image or audio. Alongside each vector it keeps the original content and useful fields to filter on, such as language, date or which customer it belongs to.
Related to this term
AI API
- Transcription, voice, text, image, OCR and embeddings
- Per-client keys with a spending cap
- Prepaid wallet: €50 minimum top-up
From €29/mo + VAT Panel account, plus usage
See the AI APIHow it works
- Index. Your documents are split into chunks, the embedding of each chunk is calculated and stored with its metadata.
- Query. A question comes in, its embedding is calculated and the database is asked for the nearest matches.
- Filter. Similarity is combined with ordinary conditions: only documents in English, only for this customer, only current ones.
Comparing a question with millions of vectors one by one would be slow. So these databases use approximate search indexes: they give up perfect precision in exchange for answering in milliseconds. In practice you won’t notice the difference.
In a RAG system, the vector database is the search engine. What it returns is passed to the language model, which writes the answer. If the search brings back the wrong extract, the answer will be wrong however good the model is. That’s why grounding starts here.
Example
A lift maintenance company in Leeds holds technical manuals from twenty manufacturers, hundreds of PDFs. Its engineers waste time hunting for the right error code. The company indexes the manuals in a vector database, with manufacturer and model as filters.
From their phone, an engineer types “door won’t close, error E17, model X” and gets the three paragraphs from that model’s manual that deal with the fault. The manufacturer filter stops another brand’s E17, which means something else entirely, from turning up.
What it means for your business
If all you want is an agent that answers from your documentation, you don’t need to choose or run a vector database. At UTIO, the agent answers from your knowledge base, and the agent’s memory and the CRM are stored in Spain.
If you’re building your own search or your own RAG, UTIO’s AI API creates the embeddings to index and search your content, and turns scanned documents into text with OCR before indexing. The vector database is your choice.
Frequently asked questions
Do I need a vector database to use RAG?
Not always, if you have few documents: you can compare vectors in memory or with an extension to the database you already use. With tens of thousands of extracts, lots of queries or per-customer filters, a vector database pays off.
Which vector databases are there?
There are dedicated services such as Pinecone, Qdrant, Weaviate or Milvus, and extensions to well-known databases, such as pgvector for PostgreSQL. The choice depends on volume, where you want the data hosted and what your team already uses.
Does a vector database replace a normal database?
No. It's for searching by similarity. Orders, invoices and customers stay in your usual database, where exact values are what count.
Related terms
- Embeddings
An embedding is a list of numbers that represents the meaning of a piece of text, an image or audio. Items with similar meaning have embeddings that sit close together, which lets you search by meaning rather than by exact words.
- RAG
RAG (retrieval-augmented generation) is a technique in which a language model first retrieves the relevant passages from your own documents and then writes its answer from them, instead of relying on what it memorised in training.
- Knowledge base
A knowledge base is a business's organised store of information (FAQs, terms, opening hours, product sheets, procedures) that an AI agent or a team draws on for answers.
- Grounding
Grounding is the practice of making an AI model base its answers on specific, verifiable sources, such as your documents, a database or a search, rather than on what it remembers from training.
- API
An API (application programming interface) is a set of rules that lets one program ask another for data or an action: you send a request in an agreed format and get a response back, without needing to know how the other system works inside.
Updated on