Sanity Context and content APIs (opens in a new tab)
The official reference for the tools used in this lesson.
Link
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.
Lesson 1.24m 23sAdvanced8.7k students
Chapters
4:23What 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
Resources
Personal notes are coming soon.