Vector Database
A database optimized for storing and querying high-dimensional vectors (embeddings).
A vector database lets you find the N items most similar to a query vector quickly, even with millions or billions of items. Under the hood they use approximate nearest-neighbor algorithms like HNSW or IVF.
The options have multiplied. Dedicated vector DBs include Pinecone, Weaviate, Qdrant, Milvus, and Chroma. Postgres can be a strong vector DB via the pgvector extension (used by Supabase and Neon). Search platforms like Elasticsearch and OpenSearch also have vector support.
For most startups, pgvector is the right starting point it removes the need for a separate piece of infrastructure. Move to a dedicated vector DB once latency or scale demands it.