Microsoft 365 Community Development, 28 Sep 2023

Previous Meeting

Next Meeting

⏯

youtube image

►

From YouTube: Microsoft 365 & Power Platform Development Community call September 28th, 2023

Description

Microsoft 365 & Power Platform Development Community call on September 28th. Recap on news and updates from Microsoft and community projects, followed by demos by the community on the art of possible.

✨ Call included three awesome demos:
• Kevin McDonnell (Avanade) | @kevmcdonk - Bringing third parties APIs to the Teams AI story
• April Dunnam (Microsoft) | @aprildunnam - Using Community Hits & Co-Pilot Tips with the Power Platform Prompts Repo
• Anoop Tatti (Content & Cloud) | @anooptells - Building Microsoft Teams message extension for Stock data with OpenAI function calling feature

Q&A takes place as always in chat throughout the call.

Screenshot summary for this call | https://warner.digital/pnpsummary09282023/

🧡 Join the next one, everyone is welcome! - https://aka.ms/community/calls

📃 Agenda
00:00 – Intro
06:15 – Latest updates
15:15 – Together mode picture
16:42 – Kevin McDonnell (Avanade) | @kevmcdonk - Bringing third parties APIs to the Teams AI story
34:49 – April Dunnam (Microsoft) | @aprildunnam - Using Community Hits & Co-Pilot Tips with the Power Platform Prompts Repo
45:33 – Anoop Tatti (Content & Cloud) | @anooptells - Building Microsoft Teams message extension for Stock data with OpenAI function calling feature
59:32 – Closing

📝 Learn more
• Teams AI Library - https://github.com/microsoft/teams-ai
• Get started with Teams AI library - https://learn.microsoft.com/en-us/microsoftteams/platform/bots/how-to/teams%20conversational%20ai/how-conversation-ai-get-started
• What is Semantic Kernel? - https://learn.microsoft.com/en-us/semantic-kernel/overview/
• https://www.deeplearning.ai/
• AI powered Chat Bot for Microsoft Teams using Teams Toolkit for Visual Studio Code, LangChain JS and OpenAI - https://github.com/garrytrinder/msteams-bot-langchain
• LangChain - https://docs.langchain.com/docs/
• Bringing natural language to the world of the Graph sample - https://github.com/kevmcdonk/TeamsAI-GraphSample
• Power Platform Prompt Library - https://aka.ms/powerplatform-prompts
• Prompt Management Solution - https://aka.ms/promptmanager
• Social Media Tracker for Analytics - https://github.com/pnp/powerplatform-prompts/tree/main/prompts/power-apps/social-media-tracker-for-analytics
• Expense Reimbursement Approval - https://github.com/pnp/powerplatform-prompts/tree/main/prompts/power-automate/expense-reimbursement
• Azure functions as Teams bot – https://www.youtube.com/watch?v=PtdOdYax84A
• Finnhub API ACE – https://t.co/1vtKN7lSb2
• OpenAI Function Calling – https://openai.com/blog/function-calling-and-other-api-updates
• Azure OpenAI Function Calling – https://learn.microsoft.com/en-us/azure/ai-services/openai/how-to/function-calling
• Sample – https://github.com/pnp/teams-dev-samples/tree/main/samples/msgext-openai-stock-price

💡 Learn more about the Microsoft 365 & Power Platform community
• Microsoft 365 & Power Platform Unified Sample gallery - https://aka.ms/community/samples
• Microsoft 365 & Power Platform Community in YouTube - https://aka.ms/community/videos
• Microsoft 365 & Power Platform - http://aka.ms/community/home

Sharing is caring! You are awesome! 🧡

A

B

We are tuned in and live with the community. This is the Microsoft 365 and Power Platform development. Bi-Weekly call today is September 28th. It is our last call for September in 2023. How did that happen fall is just around the corner. October October, you got it. We're excited.

B

Let's see we're going to talk about today, we're going to have updates on the community we're going to talk about pnp.net libraries, Powershell, yo teams, Microsoft teams, toolkit MGT, publisher, connectors, we're also going to talk about script, samples, team samples, Power, Platform samples, all the samples in all one place for you and, of course, everybody's favorite, Time picture time. Then we're going to get our All-Star presenters of the day. Kevin is going to cover bringing third-party apis to the team's AI story.

B

April is going to be here to talk about using Community hits and copilot tips with the Power Platform prompts repo, and the noop is back showing us how he can build Microsoft teams message: extensions for stock data with an open, AI function, calling feature excited for all of these, but first, let's talk about some resources that are available to each and every one of us. We've got community videos, LinkedIn groups for discussions, we've got open source tooling, and software that is available for you to make life easier. I mentioned those sample galleries.

B

We got plenty of them teams, spfx Power, Platform list formatting. All you got to do is remember: 7, 000, URLs, just kidding you just got to remember one AK dot, Ms, slash, Community, slash home, will give you access to all these resources and guess what they are absolutely free, so check out, AK, dot, Ms, slash, Community, Home and you'll be able to continue using all of them for free all the time now, you're on a community call because you like them. Well, we got plenty more for you.

B

Let's start off with the Microsoft 365 and Power Platform weekly call now what makes this unique is Microsoft speaker. So all those program managers.

A

B

Managers building the cool stuff that we all love to do things with they're gonna, be there showing stuff off it's on Tuesdays at 8, A.M Pacific, get your invite AKA dot, Ms, slash, Community, slash, Ms speakers call invite with a bunch of dashes in there. You see it on the screen, we'll put it in the links for you and you got it. Then we've got our Power Platform monthly call, which is going to be coming up right in the middle of the month. Next uh next month in October, it's the third Wednesday of every month.

B

It's gonna be right after the Power Platform Community uh conference, I should say so. We're gonna have some cool updates there and then, of course, our office add-ins and our sibling calls Microsoft 365 Power Platform Community, which you are on now, and our Viva connection SharePoint framework. You can get access to all the calls at AK, dot, Ms, slash, Community calls so go get them add them to your calendar with those ICS invites, and we will see you at all of them. Speaking of that Microsoft call. What's the agenda for Tuesday?

B

Well, we've got wallet coming in to give us the latest on the Microsoft 365 developer. Proxy aicha is going to cover using Azure, open, Ai and GPT prompts to generate completions for email and SMS messages so that we can be cool with those mobile devices and then, of course, shabnam is going to cover getting started with adaptive card based Loop components.

B

So a wide spectrum of opportunity for us again 3rd of October October at 8, A.M, AK, dot, Ms, slash, Ms Community calls invite with all those dashes now you're going to see some really cool stuff today by our community and Microsoft presenters and you're, going to wonder, I want to show stuff off and we welcome you to all you got to do is fill out. This request form AK, dot, Ms, slash, Community, request, slash demo, and we will reach out to you and we'll schedule.

B

You and you may be thinking I, don't know my stuff's pretty basic guess what basic is fantastic, so don't overthink it everybody! You don't have to create a flux. Capacitor make time travel possible. Show us the cool feature that you've recently learned about. Maybe something that's newly released that you think is really cool and you're, using it a fun way or a great way, show it off and if you've never presented before, we'll buddy you up with somebody.

B

So just let us know in the notes that you're looking for a buddy and we will absolutely partner you up with another MVP you'll work on it together and you'll present together and it's a fantastic way to collaborate with more in the community. Now you also might want a little Hands-On Guidance with some things, maybe like contributing a sample and guess what we got you covered there too, Hugo Bernier. Tell me a little more. Tell me more. Tell me more.

A

So we have this guidance where we actually show you how to get started with things like GitHub and writing for the web, and things like that I know. Github can be very scary. I remember my first struggle request, I thought I was going to destroy the repository and that's what was actually holding me back from contributing to the community, but with sharing scaring. We have small Hands-On sessions that are never recorded. Where we show you exactly the steps to do your first pull requests. In fact, we will do April requests together.

A

So at the end of the session you have done your first real request: we've ripped the Band-Aid off and uh you know we because we never record the session and because they're small sessions, it's okay, to ask questions, you don't have to worry no one's going to make fun of me. In fact, most people just make fun of me during the call. But if those are great sessions you can actually find out more by going to aka.ms, sharing Dash is Dash gearing, and once you have control attributed to a sharing is getting sessions. Something happens.

A

David. Do you want to talk about that? Oh.

B

Boy do I something sure happens. You are eligible for our community recognition program and this is a program that provides recognition powered by credly for all the amazing work that you are doing now. It's free. uh There are plenty of opportunities for you to get involved and opt in for these wonderful Badges and because I don't know. Fallen October October is right around the corner. We got some fun thematic, badges, pumpkin, spice, a lot of commits and our hacktoberfest badge. So we are right and ready for that.

B

Oktoberfest event uh definitely go in there get some uh get some contributions made. All you need to do is one to get this badge from the community and it's a fantastic way to collaborate with others as well, so AK dot, Ms, slash, Community recognition.

B

Now, let's get to some Project Specific updates and we'll start with Bert on pnp.net libraries.

C

Thank you, David um philippinep.net World, some updates for PMP framework consumers did an update to enable a stream maximum options to load templates. So that's something that they're using in the Powershell side, so it'll be an equivalent Powershell update there and then Jack Stanger did some work on uh when you extract a site into a template and there's a library with the default values for folders.

C

Something went from and Jack fix that thanks for that Jack on across the case side, which is the newer Library, there is one change by David Apache, enabling our external authentication provider to call async. So it takes a multiplication provider enables you to to do whatever you want and give us access token. So you fully own a the authentication, and that was kind of synchronous. Stolen now is async as well, which is really nice, because we want everything to be async and with that, let's move to Powershell. Thank you all.

B

Awesome got him, take it away.

D

Thanks David um slide.

D

D

Feeling pretty good about that. uh What's going on with the latest nightly builds uh we added a new commandlet here, which is called get pnb 1094 uh it will get. It will help you retrieve tenant information such as the tenant, ID or default domains, and a couple of other more information that you might require in your further processing stuff.

D

uh Besides that, we have been not so lazy, we have also uh improved quite a few commandlets here, so we improved the search configuration the file upload as well as the PNP tenant template which what mentioned in the previous uh slide here. So, though, those pnb done templates now supports stream and XML templates as well, and we also extended the set PNP Microsoft 365 group 2 also support Dynamic group membership as well so yep. That was the update from us.

D

uh If you run into any issues, do reach out to us on discussion forum as well as on the GitHub issue list, would absolutely love to hear from you. What do you David thanks? Awesome.

B

Thank you, jaram, we'll move into your teams with Stefan.

E

Thank you David. uh We did. We do not have much to share as last time as we didn't release a new version of your teams. Yet uh the reason for that is, we wanted to fix some bugs which have been reported by users on GitHub, most recently before releasing the actual new version of 4.1.1.

E

Apart from that we're discussing the future of yo teams, and we want to make sure that Microsoft 36 developers have everything they need in order to develop awesome apps uh on top of Microsoft 365, so stay tuned and, as always, please feel free to contribute joint discussion on GitHub file issues. If you think there's something missing in your teams. We definitely like to hear from you. Thank you.

B

Awesome, thank you. uh Let's move into teams, toolkit John is out today, so I'll cover this uh version. 5.2 of teams toolkit for vs code is out. It's got some new uh samples and templates, as well as an AI chat, bot and Link unfurling template. uh The biggest note to take away here is that the co-pilot, plugins and simplified message. Extensions need your feedback. So if you're building Bots definitely go reach out to John, you can do that at AKA, dot, Ms, slash, tck, Dash chat so uh reach out to him.

B

He would love to collaborate with all of you in the community. This is a very, very cool project to get involved with and a useful tool as well. Next up we've got Microsoft graph. Toolkit uh 3.1.2 is out. The samples repository is live. These MGT components are the coolest and easiest way to integrate graph into your solutions. They are configurable and extensible, so definitely get involved. Seb would love and uh be eager to collaborate with everyone. If you have any questions, uh just let them know and of course, there's always help wanted in good.

B

First issues available in the issues list which you can get to at ak.ms MGT issues for all things: MGT check out, AKA, dot, Ms, slash, MGT, all right. Let's move into some independent publisher connectors with Jocelyn hello.

F

Everybody, as always, we have even more connectors in our platform, we're now at 1075, with 50, more made by independent Publishers community members, like yourself, then made by Microsoft connectors is going to have a very strong presence in mppc coming up just next week and I wanted to bring one specific session to your attention. Partners powering plugins is a special theater session, where we'll be talking about isv enabled connectors turned into plug-ins that can be leveraged in M365 scenarios.

F

We've been talking about this for a few months, so I'm super excited to finally be able to demo that, for you, I wanted to specifically bring it up because it's not listed as a session in Hoover, it's listed as a community Meetup, but that link will be in the chat afterwards during our next connectors. Labs call will be doing a specific what you miss at mppc session for those of you who weren't able to make it so looking forward to sharing. All of that with you see you all soon. Thank you. David.

B

Awesome thank you Josh and we're looking forward to seeing you next week. Let's move on to script samples with Paul Bullock script.

G

Samples is a place where you can share your scripts with the community, so we cover things like PMP powershots, peer management shelves, sea life at n365, and many many more and we've just crossed a uh a good Milestone. So we've now got over 300 uh scripts on the uh on the repository now. So thank you very much to all those that have contributed to this, and we've also got a bunch of new samples that have come in so uh we've got see it for CLI 365. We've got the export teams.

G

Channel conversation to HTML5 by Siddharth. We've got some new scenarios for PMP Powershell, so we've got report on SharePoint file. Incidents by Michelle we've got PMP modern search, V3 usage scanner by Casper um I've written a sample around bringing your own data with Azure open AI um into that. We've also got some updates to PMP Powershell, so we've got reset file permissions unication by Mark Anderson, and we've got SPO Management Shell, um which is now a new new sample from chandani to export and active sites based on dates to CFC.

G

So uh thank you very much to all these awesome contributions. If you need any help with getting established with a contribution, please feel free to reach out and don't remember and don't forget, it is hack table Fest. So uh the repos have been marked accordingly, so you can contribute and they will count towards those uh that that program.

B

Back to you, David awesome, thank you Paul and friends. This is one of the easiest ways to get involved. We all have those utility scripts so reach out to Paul. Take advantage of his availability, he's a great guy all right. Let's move on to Microsoft team samples, so there are plenty of Team samples. In fact, you can also see some of the hack fast solutions that were available just head on over to AK dot, Ms, slash, M365 sample tools or Community samples, we'll get that updated, and then you can access all of these wonderful samples.

B

Now we're open for more so reach out to Bob German. If you have any questions he's another great guy, that's excited to work with everybody and let's move into Power Platform samples with April Dunham.

H

Yeah so I wanted to highlight our Power Platform prompt Library. So this is your go-to place if you're using some of the Power Platform co-pilots or any of the AI tools in the Power Platform and outside of the Power Platform like Azure, open Ai, and this is a place you can go and get inspiration and ideas for what you can do and what you can put into these co-pilots and things. So we have all kinds of examples from how to automate customer process things and invoices and all that kind of stuff.

H

All these samples are out there. We just got some new ones from Ralph the other day which you're excited to see so definitely welcome your contributions in that repo as you're working more with all the AI Tools in and outside of the Power Platform there go to AKA dot Ms forward, slash Power Platform prompts now speaking of prompts, though we have something really timely, we're still in our Power Platform AI hackathon.

H

This is actually the last day of the hackathon, so hackathon submissions are due by 11 59 Pacific time today, and we have a closing session to wrap it all up nice nicely and we're going to be joined by Steven segiliano, the VP of power automate and he's going to be doing a little closing keynote in a live q. A actually pretty much right after this call at 9 00 a.m.

H

I just realized the time is wrong on that 9, A.M Pacific, and so, if you have any questions that you want to get answered from Stephen definitely encourage you to join that call in that live stream there, and we will get them answered in real time. So really looking forward to that and hope you can join us and if you haven't already make sure to get your hackathon submissions in there today.

B

Awesome, thank you. April we'll see you again in a few minutes on this very topic, so it's everybody's favorite time. It is picture time Sylvester's gonna take on Oprah, which he has done. I'm too.

I

Fast, you know you are yes, it's almost like you know. What's coming on my camera, yes, I kind of I knew, maybe maybe or maybe it was a guess. uh It's been pixelated today. Let's hope that it gets better. I will not enable my camera and I will getting better we're getting better. So we still have some seats in the room a few more. It is a bit pixelated today.

I

uh We cannot help on that, a few more States in a room, but not recording quite yet uh one more second, oh now, it's getting better excellent, excellent I think we're close I will put the recording on and let's do some hand framing everybody or Crap AK animation. Out of that, thank you for joining once again, awesome I recognizable faces a lot of recognizable faces in the room, a lot of new faces in the room. This is awesome, brilliant, brilliant, thank you for joining and let's get back on the scheduled program. David.

B

All right now, back to your regularly scheduled program of our community call next up on the episode of September 28th Kevin McDonald is going to be our presenter, bringing third parties apis to the team's AI story. Kevin. Take it away.

B

You're you're muted, though.

J

Thank you very much and when you go to speak and then you think I'm going to share my screen and suddenly that mute button moves, don't you just love it um right. Thank you very much. Everyone, as David said I'm going to be speaking about how you can bring your third party uh Solutions in um so bring in those apis that you've got into that teams.

J

Ai story um very quick thing about me: hello, Solutions, architect, avenard's been around the community MVP for about three and a bit years now, uh member of the Viva explorers and the great happy princess podcast, to uh hit me up on the socials with the links and details there. But that's enough about me because what I want to get excited about today is teams. Ai first had the chance to look at this Library. It came through, as was talked about as part of the MVP Summit before it's made, live and I was like.

J

Oh so, we've got a nice way of bringing that magic of AI into teams apps and surfacing that in Bots and I was like I I. Really like the sounds of this. So I'm going to talk about the kind of problems and where it's going to come in and help with things on there.

J

We talk about some of the challenges that you face in there and also one of those challenges that I've managed to overcome, and that's really what I'm sharing today um in terms of what is teams AI, if you think about it, you've got this div and you want to build that power of azure, open AI or the open, AI apis as well, and you want to bring those prompts to Define kind of how your app responds to different messages. So you can have that chat bot and you can give your Bot a bit of style.

J

So you say: I want my bot to be a super cool fin that is just laid back all the time, even when his machine blue screens on a regular basis, just stays calm, except when he's up on stage and wants to jump off on there. You can Define these prompts and let it fill out those style and reply in that very artistic fabulous way on there.

J

But you can also do a bit more with that in terms of how you get it to guide the responses that you get and one of the things I love is to connect into third-party apis and use that to bring back the data as well. Now I'm going to do a very quick demo here before I get into the main part of the demo, because I want to talk about how easy, with the team's toolkit that it is to actually bring up and get started with this.

J

So we've got a brand new uh bot here, I'm in vs code I've got to keep an eye on the chat. So hopefully that's big enough text for everyone, but give me a shout out if you want that any bigger and I have that teams toolkit installed and I can go in I can create a new app and from that app I've got a few options, especially maybe, if you're thinking a bit about co-pilot, there's those message extensions.

J

You know wink wink nudge, nudge, say no more, um but more importantly for today, if you bring the Bots there is now. As of the latest version of teams toolkit this AI chatbot- and this makes it even easier previously- you could create the bottom. There are a few steps, but now you click on this button and you get I'm not going to run through that uh now, but that will bring you this this up within. There is a couple of steps you need to do so.

J

You'll need to build out your um to get your key from your Azure open AI instance and put that into the m.local. Now you're thinking, Kev click on that, and then we can all see your key but I'm, not that Daft I hope so. uh Hopefully, I won't click on that any minute, but once you set up your keys within there, then you can start working on it.

J

There's a couple of key files here: index.tx that sets up all the details, but with this bot now you don't actually need to touch that one everything you need is here within this app.ts and you can see it sets up your Imports those details from the teams, AI Library itself, um if you're.

A

J

Because I do love a good console. Log probably shouldn't do that in a production app, but it's great for these play ones on there and then you can see it configures, your What's called the open, AI planner. One of the bits that you need to change in here is where you set up your model with Ensure open Ai, and let me just bring that window uh where I've got it in this.

J

Your open, AI, Studio you've got your different models that you can set up, whether you've got that gpt4 or you're still looking at 3.5 and then from there. You can deploy those and give them a name, and it's that name. Let me just make that a little larger for people, it's that name, that you want to pull out and use into that details there and what I love about this playground is that you can go in there and you can have a chat with the bot in here.

J

So you can Define your system message: Ace oops, a super cool laid back fin that hates shoes, um uh any advice for speaking at conferences, and you can play around to get an idea from the types of different um items on here. So you can see. This is what you get back in your box. That comes through, so you can start to try out those different prompts that you get through from it and then bring that model across the other bit to note and I've gone blinds as to where it is yeah.

J

So I've set that up in the Azure opener open, AI planner, because I'm using Azure, open AI. If you're using open Ai and using the the kind of chat, GPT models, then you've got this other section there. So just note these two different things by default, it will be set up with this open AI. You just need to comment out this line and re-comment that and I love what the team's toolkit's done. It's made it a lot easier to get started with those two different versions.

J

Just to show you what that looks like um I'm not going to run the actual thing, because I've got my other bot run nicely, but here's the example you can see if you've put the wrong bits into there. um You can put the different air there and what I love is it notices that you entered a random set um of characters in there and comments on that?

J

But once you are working you can say what sort of thing can you help with what books would you recommend on Viva, for example, you can just have that General chat, but that's um sorry getting distracted by that chat message about the old farts doing. Sharepoint I I feel particularly targeted on that one, but you can bring this chat through into your Bot very, very easily, just gonna hop back over. If you want to bring that personality through here, you can see the basic one is you are an AI assistant.

J

He give this to give more information and Power on there and you're right Ralph. Yes, you can bypass PVA on this and get a much cheaper alternative, but that's an argument for a different day right. Let's come back to these slides a sec, so that's a very quick way and uh keep an eye. Hopefully David is posting some of the links there on how to get started very easy on that to get started, especially with the team's toolkit on there.

J

So some of the challenges that you start to um some things to consider is that that, when you take that team's toolkit, it will cater for different llms. So as I say, you can use open AI, you could also use the hugging face ones as well. It does have support for open ai4. One thing I would say is that when you create those deployments of your model, make sure you prefix it with GPT Dash, there's a tiny little thing in the documentation in the code that mentions that.

J

But if you are, if you are lucky enough to have the gpt4 make sure you reference that to get the full power from it and I would recommend making sure you do that, one of the things if you've looked at some of the latest, you can now bring your own data into that into your open AI.

J

So you can provide us your cognitive search, and that brings citation at the moment the team's AI Library doesn't support that I would love to see more backing um of there to try and bring that through I've started looking at building that in but time uh hasn't helped in there. So that is something will be really useful that when you bring your own data that it provides the references to that. So if you wanted to build your own co-pilot for that limited scenario, it would really help with that as well.

J

I know, there's some some of us looking and chatting about how we can better help with that as well and then. Finally, one of the the big issue I found is that I wanted to do that. Connections to third party and I just couldn't find an easy way to do it. So I raised the discussions. I used the right way of engaging with um the open source communities and had a chat and found out. It was actually supported and actually built in to that teams.

J

Ai library is support for oauth and the second you've got oauth. You've got that way to connect to all sorts of different systems using your SSO and bringing those things in there and then there's capabilities which I'll show in a minute to make sure that when you, when you make certain phrases that it Roots through to that capability and then using those prompts to make the most of calling those libraries as well so I'm going to go into demo.

J

Talk about how to configure that oauth I've been desperately trying to write this up properly in a blog and failing miserably uh on that. But I promise at some point. It will, and it is on the library um that I've mentioned in there.

J

And skipping half I can see one there about open, AI being trained on more current information. Yes, it is, but it is still um within that realm. It won't expand and take on the latest. If you want to pull in latest information from other systems, then you need to find different ways to do that.

J

Okay, let's jump back now, um I actually created this bot from the teams toolkit, but before the AI one was in there. So, whereas previously you saw I was look at the app.ts all the same functionality. If I come up to the top here, you can see that same planner and I'm still doing that. Log of those so open, AI, endpoints and keys must keep an eye on that and change those after this.

J

But it shouldn't come up on here so here you can see that same planner, all in that same index within there, and it's got the other configuration you can even now bring in the moderator uh onto that. So if you wanted to stop um someone swearing- and you know talking about the wrong thing, then you can add that moderator built into that to either open AI or is your open AI where things get different to the others?

J

Is this block here around authentication, and this is the part that will hopefully take your SSO to the Bots, but you define this connection name here and that's set up in the config so in the environment variables and that will use your Bot config, so I'm going to jump over if I can find my window here, you can see that I'm in a resource, Group I've got here if I jump back to the overview I'm here in a bot service.

J

So when you create your Bot in Azure by publishing it, you have this resource in there by default and I've got this in the notes on there. It goes to I, think dev.botframework.com and you can transfer that over. So even if you're just doing local, you don't have to provision this up. You can take the local version and make that an Azure one and that's the step you need to do because within this configuration and I'm hoping there we go within this configuration.

J

You add those oauth settings down the bottom and if I click on there, you can Define your provider. So obviously, in most cases it's the Azure active directory, you can give it a name. You can give it out a d secret, but you've got other providers here. So one of the samples I'm working on at the moment that I can't see listed there uh is serviced now, but you could connect up to Trello. You can connect up to Salesforce.

J

This same method will work for anything with oauth that you can bring it back to and that name there is what you will put in whoops.

J

That's what you put in this connection, name in the config and then it will use that configuration for the oauth to go and Connect into there.

J

Now, when you do that in this case, with the Microsoft takes five community, what's the most used API, we have obviously it's the Microsoft graph, so I've built this sample based on that and if I jump up here, I've just created in here a graph service. So any of the calls I'm making to the graph call this graph service and in the Constructor I'm.

J

Just passing it a token, and this part here that will grab it and if it's not set will bring from that auth provider to set the token as a variable here and then, whenever I'm calling the graph client it's set up that token.

J

So if I come back, you can see here when I'm instantiating, that graph service to call within this function, I'm just passing this property, and this is a fixed property as part of the teams AI library, to send that auth token. So, by setting up this configuration here, setting up that oauth in years your portal, you can then retrieve that token and you can pass it on it. At that point, the world is your oyster.

J

So, to give you an example, you can see here that I've got these actions, so in this case I've said read: mail and what I want it to do is call the graph service that will go in there return. Some mails return that using the create mail cards, which is just below I've kind of I've got in my code here. Where have they gone some adaptive cards?

J

So it would look nice pretty using adaptive cards to IO, build out that retrieve the function load that blah blah blah blah blah blah load that into there and I can render those results and that's a really nice way of doing it, and let's just bring up I'm going to cheat slightly and not do it live right now, because I need to make sure that that no one emails me with anything under NDA that I get in trouble.

J

But here you can see I've returned those cards as different bits of news on there lovely news from rencore on their blogs sooner safer happier, that's bringing my mouse through now. Would you really want a bot to do that?

J

Probably not, but it gives you an indication I'm going to try and speed up to make sure I've got lots of other time for other people, because what's more interesting and also you don't always want to put adaptive cards hard-coded in there if I look at these search results, so I've got in here another query: search for files about Viva and again I've got an Adaptive card. But if I go back to where this is, let me just scroll down a bit you can see.

J

I did have that hard coded, but now I'm, actually chaining together. So I'm calling summarize search, which is another prompt on there, uh make sure I get the right one there. It is I think it's that one and you can see if I hide there return the Json for an Adaptive card with a carousel I have not hard coded any adaptive card in there, I've asked the Azure open AI to return the Adaptive, and here is another function. If I go back to index just below that is search files for the query.

J

I can pull the conversation out, and here you can see I'm pulling that search query that has been defined in the line above I'm calling my graph service to bring back that query and load that result so effectively.

J

My uh sorry, this one, the prompt, is loading those results in there and that becomes my prompt and the reason I'm showing. That is because you can build this prompt up in that playground to get those results. So I haven't hard coded any uh adaptive card. It's pulling that automatically from Azure open Ai, and this is really what I'd love you all to think about with your prompts is taking a bit in a different way. Another example here is I've, taken the results from my.

C

B

Sorry interrupt just we're almost into the next demo time, so if we could just wrap it up real quick.

J

Very quickly, absolutely on there. This is taking the kind of looking at the well-being and productivity and, if I, just very quickly go back to that apologies to uh I think April who's. Next, you can see here it's looking that I've got quite a busy calendar with back-to-back meetings, it's essential to have breaks it's taking those results and bringing them back as a nice easy engaging way. So it's not just about putting things in cards. It's about bringing these text-based ways through to show for other people.

J

um If you do have questions, there's a few links for people, the sample, I've got I. Think David's put into the chat there as well. I would definitely recommend looking at some of the semantic kernel and also Gary Trin has been doing some lovely work in an alternative way. Instead of teams AI about using a different um framework called Lang chain so well worth taking a look at that.

J

That's awesome. I will finish up and hand over there.

H

Hey everyone, I am April Dunham and I'm. A power platformer advocate of Microsoft happy to be here talking about the Power Platform prompts repo and doing a quick demo of that. This is my YouTube channel. If you want to reach me there, that's also my handle on all the socials. So that's how you get in touch with me all right, so let's Dive Right In so who all has been there. Generative AI gives us endless possibilities and we're excited about all the stuff that we can do with it. And then you start thinking about it.

H

More and you're like generative AI gives us endless possibilities. Then, where do I start? It can almost be overwhelming right when you have this amazing tool that you can do anything with, and you just don't know where to start with that blank canvas sometimes, and maybe you could use a little bit of help, so, let's kind of dive into that a little bit and then we'll talk about where the prompt repo the library fits in. So obviously it all starts with a prompt.

H

That's how we interface with generative AI, whether it be any of the Power Platform tools and the co-pilots that we have across powerapps and automate, and all that Azure open, AI chat, GPT, Dolly mid-journey. All of those uh we interface with those tools with props, so they're, really just text based queries or instructions that guide the AI model into generating content for us, whether that be text, content images, whatever tool that we're using there now there's a few.

H

You know a few generative General tips here for kind of crafting the the perfect prompt, I would say: there's I mean the perfect prompt is maybe a far-fetched word, but you get where I'm going out here. First Clarity is key, so make sure that your prompt is clear, concise, specific and you're, really conveying the intent to avoid any misrepresentation, provide context. That's really important as well so give context about to help the generative AI, understand the scope and the purpose of what you're wanting it to help you with specify that is desired output.

H

This is another really good one. If you're wanting it in a very specific format like I want to return. This and you know Json format or whatever it might be, or this should be code whatever that output. It should be a poem, a story specify that in your prompt and use examples like here's an example of I want this, and this is really good, where you can kind of specify the tone or the style that you're looking for. So it can better grasp your expectations and, of course, iterate and experiment.

H

So we're going to be looking at the prompts Library here and a lot of the prompts you're going to be seeing are made from people iterating an experiment, they're trying something and they're figuring out what works best to get the best result from the tools. So that is really where the prompt library is to the rescue. So that's kind of a lot to learn just to know how to write a good prompt to get the most out of the tools and again it could be overwhelming to know what you can even put in there.

H

So with the prompt libraries here. That's really the purpose of it is giving you a really good starting point that you can start with here to explore generative AI use cases you can browse prompts that have been tested, so the ones that we have in there right now. People have tested these in the tools. They know that they work and get the output that they're expecting. So it helps you boost productivity and then, of course, as we always talk about on these calls, we encourage you to share your prompts with others.

H

So it's going to help others. If you share the prompts that you have, because students are, if you're trying to use a co-pilot or an AI tool for something someone else could benefit from that same use case. So that's why we're doing this now. There's also another thing that I'm going to be showing you a few different demos, but the other thing I want to show is a prompt manager app. So this is a sample on our Power Platform samples, repo that I actually built.

H

That shows, if you want to kind of do something more internally for your organization, that you can have this internal application built in power apps to kind of categorize. Your different prompts that you have, and then everyone in your org can benefit from that, and you can create a list of favorites so with all that, let's dive into the demos that we want to show so we're going to be highlighting a few prompts from The Prompt repo and that prompt to Manager application. So you can see how it all works together.

H

So the prompts library and David, of course, is putting all these links in the chat here is available at aka.ms4. Slash, Power Platform prompts so the first kind of stage that you'll get here is where you can browse the different prompts that we have so. This is kind of just an easy way to see what we have. So we can click on any one of these prompts like this plagiarism detector and we can easily copy and paste directly from the screen, and we can filter by the different services that it's related to like.

H

If there's Dolly prompts that we're looking for Azure open, AI, whatever it might be, but if we look at this, the base of it is on GitHub. So if we go over to GitHub here, there's a link just by the way in the upper right hand, corner that will take you to the GitHub repo where all of this is stored. So if we look at our prompt repo, we have a prompts folder and then in that prompts folder.

H

We have different folders for all other related Technologies, so whether you're wanting problems related to powerapps automate, even power Pages, we just got a new sample from Ralph on power Pages, which is really cool, Power Virtual agents and AI Builder. So you would just dive into the different kind of folder that you want so I happen to know that maybe I want some power. Apps prompts so again. Some of these are new prompts that we just have from from Ralph the other day.

H

So I can see, there's one for a prompt, collector and then one that looks pretty interesting to me for the social media tracker for analytics. So we go into the prompt and then we see a description of kind of what it does for us. So this particular one is helping to create a starting point where we can kind of track metrics for social media posts and then, of course, supported languages. So this is another thing I wanted to highlight.

H

If you're looking to contribute to the repo, we do eventually want to support multiple languages, for these prompts as some of the co-pilots and things expand and support broader languages. So that's why? If you look at the structure, you'll see this en US folder. So if you do speak other languages and you're able to put your prompts in different languages, eventually, we could have that multi-language support, but the prompt itself is stored in those language folders and if we want to use this, it's extremely simple.

H

So here is a prompt that Ralph has already tested, and he knows that it gets him. The result that he wants to build this social media tracking power apps. So now all I have to do is copy that and go over to powerapps and we'll use our copilot, so I make.powerapps.com right now and we'll go into the powerapps copilot paste in Ralph's prompt now. Another thing you'll notice in these prompts is we have these brackets and in between the brackets. We have text.

H

So this is the way kind of telling you that you want to customize the output here so start in text. You want to remove that and then you'll just replace it with your values.

H

So, whether you want you know phone Cloud apps or what you're wanting to track there, so I'll just um I'll keep what Ralph has that sounds good I'll just remove these start in text, because that's the value that I want there and let's go ahead and click enter so we'll see what powerapps copilot does we're doing this all in real time here, we'll see what happens, I will say you know sometimes with the copilot.

H

You might get slightly different results, so Ralph might have got something a little bit different in his co-pilot than what I'm going to get here. But generally it's pretty pretty close! Oh awesome, you know you gotta love co-pilot there. It was just working the other day. That's what I said with doing uh doing all this real time. Sometimes, if you do it twice, um it'll it'll behave and we'll see what happens if not we'll move on to the next one, because I know this works there, you go okay there.

H

It is so the there's the uh the app there the dataverse table in this case that they go back created for us. So that's pretty much everything that we need there to customize um to have the social media tracking application.

H

So it's automatically getting columns for reach and Impressions and engagement and clicks and all that great stuff, so you just click, create app and then now we have a fully functioning power, app, that's responsive and all the great stuff that copilot gives us with powerapps co-pilot, thanks to that prompt for mouth I didn't have to go and think about exactly what I needed to put in to make that happen in the fields. I know that this works and I'm good to go. So that's just a really great way of how this can help you.

H

So this is a power apps example, but we have the same thing as we saw in the repo there for power automate and all the other different Power Platform tools and Ralph has relieved that that it worked. Of course, it worked, I think that was user error on my part, of course, your prompt works so I'll try to if powerapps will be really fast here. I just want to show you the outputted app.

H

So here we go, here's an app to be able to manage our social media, which is great, and then we have the same thing for power automate. So these are all the power automate examples that we have here again. All of these links are going to be in the chat, but if we want to do like an expense, reimbursement, workflow automation, there is a prompt for that which same exact thing.

H

We would just copy this particular prompt, go into Power, automate, Copilot and have it generate the suggested flow structure for us and we could customize the inputs there of where we want to. You know send the request to and all that and get pre-tested great prompts and then finally, the last thing here.

H

So this is again check out that prompt library and get some inspiration and ideas and if you have prompts to share, please do share that on there, but that's more everything that we put on there is externally facing everyone can benefit for it, which is great, but maybe you have some prompts that are really specific to your business.

H

So maybe you don't want to share those widely, but you do want a system for your organization or maybe just for your personal use, to keep track of your favorite props and that's where this prompts manage your application fits in again. This is all on the repo, so you can download and install that in your tenant and all that, but it just gives you a way to kind of arrange your prompts by products.

H

So if you want to see all your you know, your Dolly prompts or Azure openai and by category so maybe you're getting prompts like uh marketing related I.T related and then you can just browse those and then it has an easy add to favorites. So you can create a list of your favorite prompts there, and then it has a built-in copy and paste. So it's an easy copy button.

H

I don't have to do what I was doing to the prompt repo do there and I can just copy that paste that into my copilot of choice or tool, choice and I'm good to go, and you can add notes and all that. So it's just a really great tool to check out if you want to have an internal repository as well. So that is all I had definitely encourage you to check out the links.

H

David Post in the chat to check out the repo check out that application and share your prompts as you're starting to use some easy guide tools.

B

Awesome, thank you April by the way, a full rights to just call me Hugo, since I'm just calling everybody by different names today, apparently so. Thank you so much. This is super cool everybody check it out prompts are fantastic super easy way to contribute and that tool is super powerful, so did I, say super enough I think so it's super all right. Let's move on to a new building, Microsoft teams, message extension for stock data with openai function, calling feature Anup, take it on over.

K

All right Kristoff, thank you very much. David So in this demo we'll be talking about the Microsoft teams, message extension, which will help us forward the current uh stock price, and we do that using open, AI function calling and any API cordless uh finhub, which we'll be looking at in a minute right. So my name is Anup I'm, an MVP in the Microsoft 365 development, category, I work at a company called content and cloud based in the UK. There are links to my profile, Blog, Twitter and GitHub.

K

You can follow me on those platforms if needed right. So the agenda for today is we'll start with the demo first and then we'll take a look at the open, AI function, calling feature uh then we'll have a look at the code and we'll conclude so before starting the demo. I would like to thank Lee, Ford and kunj for their samples uh on Microsoft teams using openai and an adaptic card extension which uses finop API. So I've used the combination of these two to create a message extension in themes.

K

So let's take a look at the message extension. So what we have here is a chat in themes, and so what I'm going to do now is in this chat, as you can see, Megan has asked me about the stock price of Microsoft. So if I click on this and then open the stock price message extension, which is the custom one, what happens here is we can go ahead and ask a question saying: how is back so you can do a typo if needed, uh foreign.

K

This sensor request to open AI, along with uh with the details that it gets from finhub uh API and then open AI looks at that information, uh and then uh you know gives us a nice message over here saying just a funny funny message and then it also has all these stock details which are provided by finhub, and then we can go ahead and send that to Megan who I guess will like it as you can see from the chat history. So that's what uh this particular message. Extension does now.

K

I know that there is a stocks message extension provided by Microsoft, but this one over here is just for demonstration purposes, because I just wanted to show how we can use open, AI or how open AI comes into this picture into the picture in order to show this for functionality right now, if I go back to the presentation, uh so the scope of this particular message extension is applicable to both open Ai and Azure. Open Ai and the models with which it can be used is our GP gbt, 3.5, turbo and gpt4.

K

um So in this particular sample we are using open AI, but it can be changed to Azure open AI very easily. Now before looking into the code, let's understand what function calling is uh and then once we've got that understanding we can. We can come to know how we are using that in the code so function calling we can think of that, as uh you know, extending open AI models uh to give more functionality.

K

So the way it works is whenever we ask a question, uh it understands the intent of the user and based on that it tells us which function to call. So to give you an example, uh if we ask openai to get the current weather uh open AI or a comes back to respond, saying that it doesn't have access to the current weather because it's been trained until September 2021..

K

However, if we tell openai that our code has got some functions which can give the information about the current weather, then open air tells us okay, you can call that function in your code with certain parameters and we go ahead and call that function and then send the response back to open AI. So, for example, if you ask openai to get the current weather in a particular City, open AI tells us your.

K

You need to call the get current weather function in your code and then we call that we call that function and then send the data back to open AI right. So let's have a look at this function, calling in a bit more detail. So imagine uh this particular message. Extension that I showed you did not have a function calling feature and then, if we go ahead and ask stock price of Microsoft, uh then what the what open AI does is. It says it doesn't have access to stock data.

K

So now, let's say we introduce function, calling feature into the message extension. So what we do is we take that question stock price of Microsoft and send that to open AI. Along with that, we also send a couple of functions that we have in our code. So we we tell open AI that we've got a function called as getcode in in our code and also a function called display code. So we we pass these two functions.

K

So basically, the name of the function and the nice description of what those functions do along with this particular question to open, AI, open AI, gets this information and then it tries to understand the intent of the user. So in this case we are asking for the stock price of Microsoft. So when I understand that understands that and tells us, we need to call the function called as getcode and then the argument symbol. The value of that must be msft, so open, AI computes all this itself.

K

Based on the question we have asked and tells us to call this function and then what we do is in our code. We go ahead and call go ahead and call this function get code and then that get code function might call some external API in this case finhub and then gets the data which, which can be raw raw Json or anything else, so that data will include at the stock price details.

K

What's the current price, uh what was the high low Etc and then at that particular Json is sent back to open AI openai gets this Json interprets that and then gives a nice message based on whatever the uh you know, the values are. We then take that message and then convert that into a nice adaptive card with the stock values that we had got earlier and then show that to the user. So that's how function calling works now, let's take a look at this solution, so this particular message extension is a team spot.

K

So if we look at this particular slide from right to left so we've got the teams client, where the message extension resides and then that communicates with an Azure bot. The Azure Port then passes the request to an Azure function.

K

The Azure function then passes the request to open Ai and then the computer response is ultimately passed back to the team's client now I'm, using an Azure function here as the backend in order to run the bot, but we can use any other backend as well if needed right now, let's take a look at the code, so all this code is available on the GitHub teams, samples repository.

K

So what we, if you look at the readme and then follow the instructions, then you'll know how to set up the bot and once the bot is set up. uh What we have to do is in the local settings.json file. We need to go ahead and change some of the values. So, basically you need to add an open, AI, API key and then add a model and then also add the finhub API key, which ultimately goes and gets the stock price. Once all the settings are done.

K

You know the app will start running so in in this particular app. We've got a messages endpoint, which is an Azure function and that particular Azure function runs using this index.ts file. So here we do. The initial setup of the bot- this is all common code which, if, if you're, if you have created team sports before uh then in the bot file, which is an extension of the teams activity Handler, uh what we do is we need to start communicating with openai.

K

So, firstly, we need to create this array called as open AI messages, because these are the messages which we will send to open here. Firstly, we go ahead and get a system message now that particular system message is, you know it has got the rule of system and then we describe how open AI should behave. So this was similar to the prompts that you saw earlier in Kevin's demo and in April's demo as well.

K

So you can just tell what uh how you want the system to behave so in this case I've just said, it has to be a funny stock broker so that it can respond with some funny messages, and once we have the system message, we go into submitting the message. So what we do here is we go ahead and get the the query of the user and then push that to the open, AI messages.

K

So in this case the query was: what's the stock price of Microsoft, so we put that in the messages array and then we go ahead and call openai with these messages. So there's a method called, let's call openai and that's what we're going to hit now. That will be present that is present in in the helper file openingi.ts.

K

So this one uses open, AI library in order to call open AI. So it's quite straightforward. There's a method called as create chat, completion which X, which expects some parameters like model uh the messages and tokens Etc. So we go ahead and call that method, and then, in that particular method, like I said it expects messages and functions. So messages are the ones that we saw earlier and functions are: are the ones that we need to pass in this case, get code or display code, and things like that.

K

So if I just show you what gets passed to open AI. So if you've got this system message that we saw earlier, we've got the user message which is uh asking about the stock price of Microsoft, and then we send these functions, which will have a look at in a minute. So we send the name of the function and a nice description so that openai can understand. Based on the user's question which function it needs to call.

K

So all these details are sent to open AI. So if we take a look at that get code function, so we can see that it's got a name, a nice description and then it has got some parameters. So in this case we are asking open AI to tell us what symbol we need to use. What's the name of the company and what's the primary exchange now open, AI uses is uses its intelligence to give us the symbol, the company name and the primary exchange based on the question that was asked.

K

So in this case for Microsoft the symbol is msft, company name is Microsoft and primary exchange is NASDAQ, I, think so that that's the function. So we can Define several functions like this. Based on our requirement and pass that to open AI now, uh the response that comes back from open AI is adjacent, so we go ahead and process that response, so open AI tells us.

K

We need to call a function and the name of the function is getcode and you can see that it has passed in the symbol, the name of the company and primary exchange, so in this case msft, Microsoft and NASDAQ. So what we do is we go ahead and extract that Json and get the name of the function and we go ahead and call that function. So in this case the function is called as getcode and then here we use finhub API and then pass the name of the symbol along with the API key.

K

This then goes ahead and Returns the data which is raw Json. We pass the Json to open, AI and then open AI response saying we need to stop the execution and it tells us a nice message based on the system prompt and then what we do is we go ahead and take that message and then put that in an Adaptive card and then show that adaptive card to the end user, so in summary, function calling is available in opening and Azure open AI. It helps us provide structured output.

K

This sample uses open AI, but there will be a PR coming soon to use Azure open AI instead, and here are some of the resources that were used for this demo and with that I hand it back to David.

B

Thank you, a new, very, very cool stuff. uh Let me get these slides shared back out here, uh really really like this. It's very functional, useful uh and I'm sure that it could be applied to different markets. The only feature request, maybe if you could build it in next time- is if it could tell me what the price will be tomorrow, that would be super cool. I would pay for that. Actually, so maybe sometime in the future awesome all right.

B

Well, we've had a perfect see what I did there perfect a number of demos today, Kevin April and anoop? Thank you for sharing with us all of our presentations today. You all may have questions or you may just want to. Let them know that there did off awesome, which we welcome. Our community call conversations is here for just that. We have forums set up in the power apps Forum Pages, the power user site. Where you can collaborate and ask questions and, like I said just go, let them know they did an awesome job.

B

That is a contribution to the community. So we welcome anyone and everyone uh to like Kevin April and know they did great. These QR codes will take you directly to each Forum. I will put the links in the chat directly after the call as well but go. Let them know everybody, they did a great job and we want to celebrate and encourage them and let all of our presenters know each week that they did a good job. So please go.

B

Let them know that if you want to be one of the cool kids in the new Discord server for uh the community, you can do that AK dot, Ms, slash, Community Discord will get you access. We've got over 500 I believe now that have joined up it's a great single One-Stop shop to communicate and collaborate on different topics about the community and it's growing each day. So we welcome you to join in no cost at all and we'd love to get some feedback from you. So let us know how the calls are going.

B

Let us know what you'd like to see. More of uh we've been looking at the feedback and looking to make some tweaks here in the coming weeks. So we appreciate everyone's feedback, but do keep letting us know if there are things you do like or want to see changed it lets. Our management know that you are appreciating the calls and you find Value in them and we can continue to invest in them.

B

So thank you for those who have and for those that will the recording will be available within the next 24 hours of the Microsoft 365 and Power Platform Community YouTube channel. You can access that at ak.mslash, Community, slash videos subscribe, you'll, get alerted as soon as that video drops, uh so don't try to download it from teams. Teams is going to tell you that it's available for download it's done. It's a little liar unless you're in Microsoft you're not going to have access.

B

So if you go subscribe on YouTube, you will get alerted as soon as it drops. You can follow us on X or whatever. It's called these days at M365, PNP and, of course, the LinkedIn Channel ak.mslash Community LinkedIn, our next Power Platform and M365 bi-weekly call two weeks from today: April 12th, 7 A.M next week, Viva connections, SharePoint framework October 5th, so definitely go get signed up for all those calls and for those that are going to the conference next week.

B

We look forward to seeing you have a great rest of your day and a wonderful rest of your week and a wonderful weekend. Thank you, everybody for all that you do in this community.