Why vector search matters
Traditional databases search by exact matches — find rows where category equals "plumbing". Vector databases search by meaning. Store a document about "fixing a leaky tap" and a query for "how to repair a dripping faucet" will find it, even though no words overlap.
This semantic search capability is what makes RAG work. When a customer asks a question in their own words, vector search finds the most relevant documentation regardless of phrasing.
Common vector databases
Pinecone is a fully managed cloud service — easy to start with, scales automatically. Weaviate and Qdrant can be self-hosted for more control. pgvector adds vector search to PostgreSQL, which is convenient if you're already using Postgres for your application data.
For most SMB use cases, pgvector or Supabase (which includes pgvector) provides enough capability without adding another service to manage.
Embeddings explained simply
An embedding is a list of numbers (typically 1,536 values) that represents the meaning of a piece of text. Similar concepts end up as similar number sequences. AI models generate embeddings, and vector databases store and search them efficiently.