Canvas App Chat Bot via Mistral Large

[@hokemspitter]

As usual excellent video and demonstration. I would also love to see the Power App integration with the Chat. I could see how having this with one added feature would be really useful for my needs. To have a "Save Thread" button that saves the thread to an SP List with a specified "thread name" so that one could continue along various threads with a "Load Thread" button.

[@djust270]

Will you do a future video on building the PowerApp? That is the part I am most interested in. Thanks!

[@DamoBird365]

Im definitely thinking about it. It would also combine nicely with 👉 Power Apps Dynamic Text Strings | Documents | Prompts | Paragraphs #PowerApps kzbin.info/www/bejne/kJbbXp2lpLRon9E

[@gustavomonne4921]

Really good stuff. Will you demo the PowerApps as well? It looks really interesting.

[@DamoBird365]

If there is interest I will go for it. This requirement could be orchestrated in a couple of ways. Using flow to maintain the history or using the app and collections. If you’ve got a specific use case let me know 👍

13:00 Noticed in system prompt that can use emojis 📝🍋 What are the limits of dialog history, token limit of the Api Call?

[@DamoBird365]

32,000 tokens mistral.ai/news/mistral-large/

@@DamoBird365 thanks I don't recall that power app you built. Have I missed a video?

[@DamoBird365]

@ I’ve never showcased that one. It was the first time onscreen. I didn’t think it would be interesting. It was integrated with Azure OpenAI GPT but could now easily switch between models and have history and have prompts. Lots of ideas just not enough hours 😂

[@geralddahl9159]

Thanks for this. Two questions: (1) Is it correct that Co-Pilot in the edge browser works in the same manner as the deployment you have demonstrated? (ie, it interprets one’s most recent question in view of the history established within the context of questions previously asked from within the current session); and, (2) How much money will it cost to ask and answer the few items seen in your demo? What I really would like to know is whether or not one can configure things so as to guard against asking a very expensive question that somehow or other generates an even more expensive answer or series of unwelcome charges?

[@DamoBird365]

Hi Gerald, at a high level, yes. Chat Completion. The more advanced skills are based on an assistant and I’m thinking about that idea as a demo too. The cost is based on tokens used. This increases based on the length of a conversation, exponentially. Token use can be capped by the parameters in the body. I’m learning here too though.

[@flightbug5296]

Awesome stuff...even if it is french 😅

[@DamoBird365]

I drove Citroen cars for 20 years 😂👍🤷‍♂️ they were good to me.

[@johnfromireland7551]

Boris Johnson? Who's he?

[@DamoBird365]

Interestingly you could still pull that demo on GPT until a couple weeks back when it got an upgrade to ground on the web 😭