Get the contents of a document by its documentId or filename

[hansmbakker]

Summary

This is a feature request to directly fetch an indexed document based on its documentId or its filename.

Background

I like the KernelMemory document import pipeline, but only having the AskAsync method to work with the indexed content is too limiting for me. It forces me to always use LLM calls, and the LLM calls have less flexibility than semantic kernel.

It would help to be able to work with the content of a document directly. Either to do a custom LLM call based on that content (e.g. using additional semantic kernel features like requesting a specific ResponseFormat), or to show a citation to the referenced content.

Related feature requests

Expose loaded documents list #623

[hansmbakker]

Workaround: is SearchAsync with a ByDocumentId filter meant for this?

[KurtP20]

+1 It would also help checking if the content was extracted correctly. Also, one could pass the entire document to the LLM to ask questions like "Give me a summary" - in my present understanding this is not possible thus far.