Vector database: complete definition in AI for SMEs

A vector database is a database specialized in storing and searching vectors — numerical representations of the meaning of texts, images or data. It answers "what are the elements closest to this query?" in milliseconds, even over millions of vectors. Examples: Pinecone, Weaviate, Qdrant, pgvector (PostgreSQL extension).

What it changes for an SME

The vector database is the smart search engine behind RAG:

• an internal assistant that finds the right procedure by asking a natural language question, without knowing the exact document name;
• a support agent that finds the 3 most relevant answers among hundreds of FAQ articles;
• a monitoring tool that identifies documents similar to a specific contract.

The choice for an SME

Cloud options (Pinecone, Weaviate Cloud) are fast to deploy but bill by usage. Self-hosted open-source options (pgvector, Qdrant) are more economical at volume but require a server. For an SME, pgvector is often the right compromise: it is a PostgreSQL extension, so no extra database to maintain — Supabase includes it natively.