Are KM documents equivalent to vector db collections?

[akordowski]

Kernel Memory has the function ImportDocumentAsync and with the required DocumentUploadRequest a files collection can be passed by the Files property. As vector databases like Chroma or Qdrant are using collections to store data, I was wondering if the KM documents are equivalent to those db collections?

If yes, wouldn't it make more sense if the function ImportDocumentAsync would be renamed to ImportCollectionAsync. IMO the term Document is somewhat misleading, because you usually think of a document as one file (and not multiple files). And this would also bring more clarity regarding the vector db functionality.

[dluc]

KM Documents are information units with a unique ID for updating or deleting them. They typically contain one file but can include multiple files. Documents are processed to extract multiple Memory Records. You can think of Documents like OpenXML documents, that under the cover are a ZIP archive containing multiple files.

KM Indexes are collections of records, similar to Qdrant collections or Postgres tables. They enable horizontal scaling and physical separation of records.

[akordowski]

Thank you for the explanation.