---
title: "Vector database — Glossary | LLMind"
description: "A database optimized for nearest-neighbor search over high-dimensional embedding vectors — common in RAG stacks."
url: https://llmind.org/glossary/vector-database/
source_format: html
---
[← Glossary](https://llmind.org/glossary/)

# Vector database

**A database optimized for nearest-neighbor search over high-dimensional embedding vectors — common in RAG stacks.**

A vector database is optimized for searching high-dimensional data. Instead of traditional SQL queries (WHERE status = 'active'), vector DBs answer nearest-neighbor queries: "find the k vectors closest to this query vector." They use approximate nearest-neighbor algorithms (HNSW, IVF, quantization) to achieve sub-linear search time over millions or billions of vectors.

## What it does

Vector databases store embeddings as vectors (often 768 to 1536 dimensions) and answer similarity queries in milliseconds. Common players include Pinecone, Qdrant, Weaviate, Chroma, Milvus, and pgvector (PostgreSQL extension). They handle the operational complexity of indexing, scaling, and filtering that would be expensive in a traditional relational database.

## When you need one

Use a vector DB for large corpora where similarity search is the bottleneck. If you're building a semantic search engine, LLM-powered Q&A system, or recommendation engine, a vector DB is the standard choice. It's designed for this problem and has been optimized by many teams.

## When you don't

If your corpus is small or easily browsable by filename, filesystem access via an MCP server can replace the vector DB entirely for many agent use cases. LLMind enriches files with semantic metadata so downstream agents can access rich context directly without needing similarity search.

## Related terms

-   [RAG](https://llmind.org/glossary/rag/)
-   [Embedding](https://llmind.org/glossary/embedding/)
-   [MCP](https://llmind.org/glossary/mcp/)

## See also

-   [Agent file access patterns](https://llmind.org/learn/agent-file-access-patterns/)
-   [Spec](https://llmind.org/spec/)