[FEAT]: Store source URL; Link to the source document #592

Closed
opened 2026-02-28 04:51:30 -05:00 by deekerman · 1 comment
Owner

Originally created by @Tiberius1313 on GitHub (Mar 24, 2024).

What would you like to see?

After uploading a document a parsed JSON file is stored in “custom-documents”. That file contains the information:

  • "url": “file://C:\Users\[USER]\AppData\Roaming\anythingllm-desktop\storage\hotdir\[NAME_OF_THE_SOURCE_FILE.EXTENSION]"
  • "docSource": "pdf file uploaded by the user."

To my understanding this information is useless.

Would it instead be possible to store the path URL of the originally uploaded source document?
Would it also be possible to show that link in the LLM reply as a reference and also to allow the user to open that document and maybe even to see the refereed text highlighted?

Originally created by @Tiberius1313 on GitHub (Mar 24, 2024). ### What would you like to see? After uploading a document a parsed JSON file is stored in “custom-documents”. That file contains the information: - "url": “file://C:\\Users\\[USER]\\AppData\\Roaming\\anythingllm-desktop\\storage\\hotdir\\[NAME_OF_THE_SOURCE_FILE.EXTENSION]" - "docSource": "pdf file uploaded by the user." To my understanding this information is useless. Would it instead be possible to store the path URL of the originally uploaded source document? Would it also be possible to show that link in the LLM reply as a reference and also to allow the user to open that document and maybe even to see the refereed text highlighted?
Author
Owner

@timothycarambat commented on GitHub (Mar 25, 2024):

The reason why this data is omitted and/or useless (it's not even used in inferencing) is because

  1. It would be impossible to replicate this for hosted/self-hosted ANythingLLM. Unable to open a local file from a remote system.
  2. Security implications of the above and should the file move, its reference would be broken anyway.

There is some ideas for "living" files but this would be handled in a different way for local files and would be unavailable for hosted instances regardless since the #1 point still applies.

Currently, the functionality described only works for URLS

@timothycarambat commented on GitHub (Mar 25, 2024): The reason why this data is omitted and/or useless (it's not even used in inferencing) is because 1. It would be impossible to replicate this for hosted/self-hosted ANythingLLM. Unable to open a local file from a remote system. 2. Security implications of the above and should the file move, its reference would be broken anyway. There is some ideas for "living" files but this would be handled in a different way for local files and would be unavailable for hosted instances regardless since the #1 point still applies. Currently, the functionality described only works for URLS
Sign in to join this conversation.
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
starred/anything-llm#592
No description provided.