Unanswered

Copilot stop responding with Generative answer based on uploaded files on the "Generative AI" page

(0) Share

Report

Posted on by IM-19081311-0

I created copilots earlier that knowledge base are files uploaded on the "Generative AI" page and it works correctly, but several days ago I created new copilot and uploaded some documents as a data source, but from generative answer I was redirected to fallback topic. Is there any changes on Copilot Studio and advice how can I have working Generative AI based on uploaded documents (Dataverse).

Generative AI page has this warning:

Screenshot 2024-05-20 at 16.10.31.png

Thanks in advance

Categories:

General topics

I have the same question (0)

All responses (3)

Answers (0)

citron-truc 394 on at

Like (0)

Report
Copy link

Link copied!

Good morning,

I hope you are doing well. Warning is unrelated to your problem (when you do any changes to the "conversational boosting topic", you can't change the generative answer node from the generative AI window).

The problem could come either from the documents or from the chatbot :
- Did you check that the text from your pdfs could be extracted ?
- Did you check that your question gets to your AI before it goes to fallback ?

Have a great day, hope it helps ?

Was this reply helpful? Yes No
IM-19081311-0 13 on at

Like (0)

Report
Copy link

Link copied!

Thank you @citron-truc
Regarding this:
"- Did you check that your question gets to your AI before it goes to fallback ?"

yes, I checked it, it goes to Conversational boosting before Fallback
I will appreciate if you can advice how to check this - " Did you check that the text from your pdfs could be extracted ?" (I use txt)

Was this reply helpful? Yes No
citron-truc 394 on at

Like (0)

Report
Copy link

Link copied!

Good morning,

If you are using txt files, then my remark doesn't apply (copilot studio is said to accept txt files). https://learn.microsoft.com/en-us/microsoft-copilot-studio/nlu-documents

What I meant by "Did you check that the text from your pdfs could be extracted ?" is that in Pdfs, there are often images (Copilot studio does not read images). Sometimes, pdfs keep the text and sometimes there is just one big image. In order to check if the text can be extracted, open your pdf and check that you can highlight the text.

Sometimes if you upload a document, it takes time to be taken into account, have you retried later ?

Was this reply helpful? Yes No