Vector Database: Definition and Examples

A vector database is a storage system designed specifically to manage multidimensional numerical representations called embeddings. These embeddings are produced by artificial intelligence models that transform text, images, audio, or any other type of data into vectors — lists of numbers that capture the semantic meaning of the original content.

Unlike traditional databases that search for exact matches (by keywords or identifiers), a vector database excels at similarity search. It uses algorithms like ANN (Approximate Nearest Neighbors) to quickly identify the vectors closest to a query vector, even among millions or billions of entries. This capability is at the heart of RAG (Retrieval-Augmented Generation) systems that feed LLMs with contextual knowledge.

The most popular vector databases include Pinecone, Weaviate, Qdrant, Milvus, and Chroma, as well as vector extensions for existing databases like pgvector for PostgreSQL. Each offers different trade-offs in terms of performance, scalability, and features (metadata filtering, multi-tenancy, hybrid search).

In the prompt engineering and generative AI ecosystem, vector databases play a fundamental role: they give memory and context to language models by retrieving the most relevant information to inject into a prompt, making responses more accurate and grounded in real data.