Vector database (vector store)

Definition

Vector databases (also called vector stores) are specialized for similarity search in embedding spaces. They rely on index structures such as HNSW or IVF to make approximate nearest-neighbor search (ANN) efficient. Examples include pgvector (a PostgreSQL extension), Pinecone (managed service), LanceDB (embedded), and Chroma (open source). They are the standard infrastructure layer behind RAG, semantic search, and recommender systems.

How Swiss Knowledge Hub uses this term

Swiss Knowledge Hub supports LanceDB, Pinecone, or ChromaDB as configurable vector stores per workspace or tenant. PostgreSQL with pgvector is used internally but is not exposed as its own vector-store provider.

Related terms

• Embedding
• RAG (Retrieval-Augmented Generation)
• Semantic search

Sources

1. pgvector — GitHub — https://github.com/pgvector/pgvector
2. Wikipedia: Vector database — https://en.wikipedia.org/wiki/Vector_database