Vector Database
Database specialized in storing and searching vectors (embeddings) for large-scale semantic search.
Database specialized in storing and searching vectors (embeddings) for large-scale semantic search.
Définition détaillée
A vector database is a storage system optimized for embedding vectors. Unlike traditional relational databases, it indexes vectors using approximation algorithms (HNSW, IVF) enabling similarity search in milliseconds, even across millions of documents. Popular solutions include Qdrant, Milvus, Weaviate, Pinecone, and pgvector for PostgreSQL. Vector databases are the core technical foundation of RAG systems.
Cas d'usage
RAG knowledge base: 500,000 pages of technical documentation are vectorized and queryable in real time. Employees get sourced answers in 2 seconds instead of 15 minutes of manual search.
Termes associés
Questions fréquentes
Do I need a separate vector database or can I use PostgreSQL?
pgvector is an excellent starting option, right inside PostgreSQL. For large volumes (>1M vectors), a dedicated solution like Qdrant or Milvus performs better.
What does a vector database cost?
Open-source solutions (Qdrant, Milvus) are free but require infrastructure. Cloud solutions (Pinecone) charge per storage and queries, starting from tens of euros per month.