Vector Database
A specialized database optimized for storing and searching embeddings by similarity.
A vector database (like Pinecone, Weaviate, Qdrant, or Milvus) stores embeddings and retrieves them by semantic similarity rather than keyword matching. Traditional databases search by exact match or keywords; vector databases find the closest neighbors by distance in embedding space.
Vector databases are essential for RAG workflows: embed your documents, store the embeddings, and when a user asks a question, embed their query and retrieve the k-most-similar documents. They're also used for recommendation systems, anomaly detection, and personalization.
Vector databases handle the computational overhead of similarity search efficiently using index structures like HNSW (Hierarchical Navigable Small World) so that finding similar embeddings stays fast even with millions of vectors.
Example
Store 10,000 support tickets as embeddings in a vector DB. When a customer asks a question, embed it and retrieve the 5 most similar tickets as context for an AI assistant.