# Vector Database

> A vector database stores embeddings alongside their source content and metadata and answers queries by finding the stored vectors nearest to a query vector, usually with approximate nearest-neighbor search.

- Identifier: PTL-0124
- Category: Retrieval & Tool Use
- Canonical URL: https://protologue.com/t/vector-database/
- Also known as: vector store, vector index, vector search

## Description

Exact nearest-neighbor search becomes slow at millions of vectors, so vector databases use index structures such as HNSW graphs or quantized inverted indexes that trade a little recall for large speedups. Most also support metadata filtering and, increasingly, keyword search for hybrid retrieval. A vector index library such as FAISS provides the search without the database features.

## Related terms

- [Embedding](https://protologue.com/t/embedding/)
- [Dense Retrieval](https://protologue.com/t/dense-retrieval/)
- [Hybrid Search](https://protologue.com/t/hybrid-search/)
- [Retrieval-Augmented Generation](https://protologue.com/t/retrieval-augmented-generation/)

## Sources

- Malkov and Yashunin (2016). Efficient and robust approximate nearest neighbor search using Hierarchical Navigable Small World graphs. https://arxiv.org/abs/1603.09320
- Johnson et al. (2017). Billion-scale similarity search with GPUs. https://arxiv.org/abs/1702.08734

## Cite this entry

Protologue. (2026). Vector Database. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.4.0, PTL-0124). https://protologue.com/t/vector-database/

License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)