Skip to main content
PProfileMatch
Read Play
← Path

Retrieval-Augmented Generation · RAG Chunking and Embeddings: Build a Toy Vector Search in Python

0 XP
  1. 1 current stop: learning
  2. 2 upcoming
  3. 3 upcoming
  4. 4 upcoming
  5. 5 upcoming

Finish after stop 5

Stop 1 of 5 · Cutting the long ribbon at the finish arch

Trick to keep in mind+3 XP

Overlap is a seam allowance: cut a little extra so edges still join.

When a document is cut into pieces, a fact may sit right on the cut, with half in each piece, so neither piece answers a question alone. Sharing some text between neighbouring pieces means the whole fact appears together somewhere. The price is some duplication and a slightly larger index.

Read it once, then a question tests whether the trick stuck.