Understanding Attachments vs. Knowledge Folder Data in Conversations

Hi. I'm not sure if there is short answer to this but i was wondering how attachement to a specific conversation differs from data added to a knowledge folder . Also, is it possible to require attachement for a specific assistant / specific conversation and mention it in the prompt assistant something like "as an assitant you first analyse the {uploaded_file} sent by the user in the file ..." Thank you for your answer 🙂

6 comments

Remi
·
·
Thanks for your question Benoit M.!
how attachement to a specific conversation differs from data added to a knowledge folder
The difference is in how you can use it once it's loaded.
If pushed in the convo, the whole doc will be sent as context.
If in a data source, it is more likely to be divided in "chunks" then how they will be used will depend on the tool you plug it too (search, etc)

require attachement for a specific assistant / specific conversation and mention it in the prompt assistant something like "as an assitant you first analyse the {uploaded_file} sent by the user in the file ..."
Yes, you can add things like "ONLY answer if user sent an attachement, if not say "X"" Hope this helps !
✨1
Benoit M.
·
·
Great, and is there any "memory/traces" left by uploads in convos ? Mentionned later or part of the knowledge for instance ?
Remi
·
·
It's stays linked to the convo. So won't be usable once you start a new convo. I am not sure what's the data retention policy of the file attached but I can clarify with the team if needed 😊
👍1
Benoit M.
·
·
Thank you for your response. Regarding data retention, my concern was more about whether users could 'pollute' the knowledge of a particular assistant by submitting documents. Wouldn't that risk overloading the knowledge base they are supposed to rely on as a priority?
Remi
·
·
If the doc is shared in a discussion, it will only be used in the discussion so no risk here 👍
Benoit M.
·
·
Awesome