Glossary / vector-store

Vector Store

A database optimised for similarity search over high-dimensional embedding vectors; the canonical storage layer for RAG.

Component

← All terms

A vector store (also called a vector database or vector index) is a database optimised for similarity search over high-dimensional embedding vectors. RAG pipelines store the embeddings of indexed document chunks in a vector store and query it by embedding the user prompt and asking for the top-k nearest neighbours.

Security considerations: tenant isolation (one vector store backing multiple customers without strict namespace separation leaks data via similarity), metadata filter integrity (attacker-controlled metadata bending who sees what), and embedding inversion (where the attacker recovers original document content from the stored embeddings).

Managed and self-hosted variants are equally in scope for AI-SPM asset inventory. The asset row tracks the engine, region, namespace count, and the RAG systems that consume it.

Related terms

Other entries in this neighbourhood.

Retrieval-Augmented Generation (RAG) An LLM pattern where the prompt is augmented with documents retrieved from a vector store at query time; the retriever and the corpus are the new attack surfaces. Embedding A dense numeric vector representation of text, image, or audio produced by an embedding model and used for similarity search and clustering. Embedding Inversion A privacy attack that reconstructs the original input text from a stored embedding vector.

See Vector Store in production.

The Penaxtra platform implements the controls and assessments described above as part of its AI-SPM programme.

AI-SPM platform overview →