Unanswered

Provide Relevant Image Based on Question and Answers

(0) Share

Report

Posted on by juvic

Hi Copilot Studio Community,

I'm playing with the Copilot Studio and I am trying to show the value of it within the organization by providing a simple use case.

My goal is for the copilot to show specific images when specific words are triggered. Pretty much all flows are starting as Unknow Intents however I want different images to be displayed based on the question the user asks.

For example suppose we have 5 car brands with 5 different colors. I want that a specific picture is displayed if the question contains Brand+ Color. If the question contains only Brand, follow up with color request question.

What is the best way to do it? Am i able to filter it by keywords? What about converting "Unknown Intents" to topics?

Thanks!

Categories:

Topic creation & management

I have the same question (0)

All responses (4)

Answers (0)

adilei on at

Like (0)

Report

You could create different topics with each topic sending a message with a specific image. If that does not answer your question, you should probably elaborate more on what you are trying to achieve.

Was this reply helpful? Yes No
lonewolf1 29 on at

Like (0)

Report

I am facing the same issue. My source document that I use for Document based QA includes some images that are sometimes part of an answer. My generative node only returns texts. I tried converting my Pdf to JSON and see if I can display the related images in case they show up in citation Via an adaptive card but unfortunately the generative node misses some of the texts and still dont return any reference to the images so that I can display them. any solutions you can recommend that would not require Azure services?

Was this reply helpful? Yes No
Umianta 290 on at

Like (1)

Report

Hi @lonewolf1 ,
If feasible, retrieve the complete URL of the image and incorporate it into the Adaptive card to showcase the image.
If this was helpful:
⭐Please mark it as a solution
⭐Give me a thumbs-up
⭐Connect on LinkedIn
⭐I consistently upload technical videos. If you're interested, please visit and subscribe to my YouTube Channel
⭐Always glad to help, Umesh Khandelwal!

Was this reply helpful? Yes No
lonewolf1 29 on at

Like (0)

Report

Hi @Umianta , Thank you very much for your reply!
do you have any suggestions for how to know which image to return?
My document is a software manual with lots of images (screenshots of the software environment) and the generative node's references are not that helpful for returning the correct image.
I even converted my Pdf to JSON file so that maybe I would get the image path pack as a part of references so that I can display the image using adaptive cards but unfortunately if I use JSON, the answer I get back is part text-part JSON structure code.

Was this reply helpful? Yes No