Foundations

Vector Database

AMAdil Mektoub

Gepubliceerd op 13 July 2026Laatst herzien 13 July 2026

Definitie

Vector Database: is a system that stores information as numerical embeddings so an AI can retrieve it by semantic similarity — finding content by meaning rather than exact keywords — which is what makes Retrieval-Augmented Generation practical at scale.

Directiesamenvatting

Directiesamenvatting

A vector database converts text into embeddings — lists of numbers that capture meaning — and finds the passages most similar to a query. This lets an AI system locate the right paragraph even when the wording differs from the question.

It is the retrieval engine behind most enterprise Knowledge Base and RAG deployments. It is infrastructure, not intelligence: it finds candidate content, which the model then reads and reasons over.

Kernpunten

Kernpunten
  • A vector database retrieves content by meaning, not keywords.
  • It stores embeddings, the numerical representation of text.
  • It is the retrieval engine behind RAG.
  • It finds candidates; the model still does the reasoning.

Architectuur

Retrieval through a vector database follows a simple loop:

  1. 1EmbeddingText is converted into vectors capturing semantic meaning.
  2. 2StorageVectors are indexed for fast similarity search.
  3. 3QueryA question is embedded and matched against stored vectors.
  4. 4ReturnThe closest passages are passed to the model as grounded context.

Zakelijk voorbeeld

Voorbeeld van een implementatiescenario

A brokerage stores years of vessel descriptions and client notes as embeddings.

When a broker asks for boats matching a preference, the system retrieves semantically similar records even when the exact words never appear.

FAQ

Veelgestelde vragen

Do I need a vector database for enterprise AI?
If the AI must answer from your own documents and data, some form of semantic retrieval — typically a vector database — is usually required. Simpler use cases may not need one.
Is a vector database the same as a knowledge base?
No. The knowledge base is the curated content; the vector database is the retrieval engine that makes that content searchable by meaning.
AM

Auteur

Adil Mektoub

Medeoprichter · AI-engineering en -infrastructuur

DevOps-, platform- en AI-systeemengineer, gespecialiseerd in veilige en schaalbare agentische AI-infrastructuur.