Lesson 1.24mAdvanced8.7k students
Vector databases and indexes
Exact nearest-neighbour search does not scale, so vector indexes trade a little recall for a lot of speed. Knowing the trade-off keeps you out of trouble.
This lesson sits in Embeddings and Vector Search, part of Retrieval-Augmented Generation from Scratch. It assumes what came before it and leads directly into the next lesson in the module.
In this lesson you will
- Understand approximate nearest neighbour search
- Trade recall against latency deliberately
- Store metadata alongside vectors for filtering
Resources
Your notes for this lesson will appear here.