Phase 1 · Building with LLMs · Context & retrieval
Vector databases
Store and search millions of embeddings quickly.
Open this lesson Create a free account
Why it matters
Once a knowledge base grows past a few thousand chunks, or several apps need the same index, or documents change daily, a matrix in memory stops being enough. Vector databases store embeddings with their metadata, search millions of them in milliseconds, and let you filter and update. Knowing what they do (and when you don't need one) saves both money and architecture time.
What you'll be able to do
- Explain what a vector database adds beyond a NumPy array
- Build a small vector store with add, search, filtering and delete
- Describe approximate nearest-neighbor search and its trade-off
- Choose between an in-memory index, pgvector and a hosted vector database
Before this lesson
Next in this phase: Retrieval-augmented generation