Document Store
Also called Doc Store
Where the raw text and chunks actually live, so retrieval can pull them back on demand.
Think of it like
The filing cabinet behind the front-desk index cards — the cards point, the cabinet holds the real papers.
Example
The vector index returns chunk IDs; the document store hands back the actual paragraph text to stuff into the prompt.
How it actually works
A retrieval system usually separates the index (vectors for finding things) from the store (the text itself). The store keeps original chunks, metadata, and IDs so that once search identifies a match, you can fetch the exact content to show or feed the model. Some databases fuse the two; conceptually they play different roles, and keeping the source text intact matters for citations.
For product teams
The system of record for what your assistant actually quotes; keep it clean and traceable.
For engineers
Persistent chunk+metadata storage keyed by ID, decoupled from the ANN index that references it.
Related
- Vector database — Paired with a vector database index.
- Chunking — Holds the chunks produced by chunking.
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