Stage 2 · Lesson 6 of 7
Versioning documents, chunks and indexes
In a retrieval-augmented generation () app, the model answers from documents fetched at request time, so the documents are part of the model. Treat the document snapshot, the settings (how documents are split into passages), the model (which turns text into vectors) and the resulting as one versioned .
Record the index version on every answer. Then a change in answer quality can be traced to a new index, compared with the old one and rolled back. Keep old indexes until a new one has passed your evals.
Common trapre-embedding only part of the corpus with a new embedding model, so vectors from two different models sit in one index and retrieval quality quietly drops.
A question to check yourself, an AI practice session and 1 cheat sheet entry come with the full roadmap. Unlock
Resources for this stage
Start with
- CourseFreeMade With ML
Goku Mohandas · Advanced
Free: the versioning and data testing lessons, built around one real project