vector database
also: embedding database, semantic search database
A specialized database that stores and indexes high-dimensional numerical vectors (embeddings) to enable fast similarity searches, commonly used with machine learning models and large language applications.
A vector database is optimized for storing vectors—arrays of numbers that represent data in high-dimensional space—and performing efficient similarity searches across them. Unlike traditional relational databases that use exact matching, vector databases find vectors that are "close" to a query vector using distance metrics like Euclidean distance or cosine similarity.
Vector databases are essential in modern AI workflows. For example, when using a language model to answer questions, text is first converted into numerical vectors (embeddings). The vector database then quickly searches millions of stored embeddings to find the most relevant documents or context, rather than doing expensive text comparisons.
Popular vector databases include Pinecone, Weaviate, Milvus, and Chroma. They typically support indexing algorithms like HNSW (Hierarchical Navigable Small World) to make nearest-neighbor searches practical even with billions of vectors.