infrastructure
Vector Database
Vector Database
A database that searches by meaning, not exact words
Reading level
PRACTITIONER — Technical context
Vector databases store high-dimensional embeddings and support approximate nearest-neighbor (ANN) search for semantic retrieval. Common options: Pinecone, Weaviate, Qdrant, Chroma, Milvus. They use indexing structures like HNSW (graphs) or IVF (inverted files with quantization) to achieve sub-linear search time. Metadata filtering, hybrid search (dense+sparse BM25), and multi-tenancy are key enterprise features.
Real-world example
Notion AI uses a vector database to search your notes by meaning. When you ask 'What did I write about my Q3 goals?', it finds relevant notes even if they never contain the exact phrase 'Q3 goals'.
infrastructureretrievaldatabases