Embedding Cache
Also called Vector Cache
Saving embeddings you’ve already computed so you don’t pay to re-embed the same text.
Think of it like
Keeping receipts so you never re-do a calculation you already finished.
Example
When the same paragraph appears in ten documents, the cache returns its stored vector instead of calling the embedding model ten times.
How it actually works
Embedding calls cost money and time, and lots of text repeats or reappears unchanged. A cache keyed by a hash of the input (and model version) returns prior vectors instantly, which speeds ingestion and lowers bills. The subtlety is invalidation: if you change embedding models, the cache must be keyed by model too, or you’ll silently mix incompatible vectors.
For product teams
A quiet cost-saver on large or frequently re-ingested corpora.
For engineers
Content+model-hash keyed vector store; must invalidate on model change to avoid mixing embedding spaces.
Related
- Ingestion — Cuts cost during ingestion.
- Re-embedding — Avoids needless re-embedding.
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