Reference · Glossary
Vector database
A database optimized to **store embeddings and run similarity search** fast — Pinecone, pgvector, Chroma, Weaviate, and others.
When to use
Production RAG, large document sets, semantic search over thousands+ chunks.
When not to
Five static docs in a demo — an in-memory list of embeddings may be enough.
Example
1. Embed 10,000 support articles → store vectors + metadata
2. User asks a question → embed query → return top 5 nearest chunks → feed to LLM