Re-embedding
Also called Reindexing
Recomputing all your embeddings — usually because you switched to a better embedding model.
Think of it like
Re-translating an entire library when you hire a better translator; the old translations no longer match.
Example
Upgrading from an old embedding model to a new one forces re-embedding every chunk, since vectors from different models aren’t comparable.
How it actually works
Embeddings from different models live in different vector spaces, so you cannot mix them — upgrading the model means re-embedding the whole corpus and rebuilding the index. It is a real operational cost and a reason to choose an embedding model carefully up front. Planning for it — versioning vectors, running old and new side by side during migration — avoids a painful big-bang cutover.
For product teams
A hidden switching cost when adopting a new embedding model; budget time for it.
For engineers
Full corpus re-vectorization on model change; spaces aren’t compatible, so plan versioned, side-by-side migration.
Related
- Embedding Model — Often triggered by a new embedding model.
- Ingestion — Repeats much of the ingestion work.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome