Lesson 1.2
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.
Taught byPriya RamanLLMs · RAG · Prompt engineering · Evaluation
Overview
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.
What this lesson covers
- Understand approximate nearest neighbour search
- Trade recall against latency deliberately
- Store metadata alongside vectors for filtering
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