Microsoft 365 Community Microsoft Teams, 29 Sep 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: How to build a good vibes generator bot for Microsoft Teams

Description

In this 12-minute developer-focused demo, Luise Freese and Lee Ford enable Teams owners to send positive affirmations to Teams users. Design goals: Secret-less (uses managed identity), server-less (Runs in Azure cloud shell, no package installed locally), and open source. Architecture – uses Azure Functions, a bot, assembles vibe from phrases stored in Cosmos DB, and sends as Adaptive Card into a Teams conversation. Two types of interactions (entry points) for bot to get vibes (listening to messages (@mention) or based on a schedule (durable function). This PnP Community demo is taken from the Microsoft 365 & Power Platform Development Community call recorded on September 29, 2022.

Demo Presenters
• Lee Ford (Symity) | @lee_ford
• Luise Freese | @LuiseFreese

Supporting materials
• Repo - Good Vibes generator | https://github.com/working-on-it/good-vibes-generator
• Guidance - Deployment guide | https://github.com/working-on-it/good-vibes-generator/blob/main/docs/deploymentGuide.md
• Documentation - What are Durable Functions? | https://learn.microsoft.com/azure/azure-functions/durable/durable-functions-overview

Learn more
• Microsoft 365 Unified Sample gallery - https://aka.ms/m365/samples
• Microsoft 365 Platform Community in YouTube - https://aka.ms/m365/videos
• Microsoft 365 Platform Community - http://aka.ms/m365/community

A

A

Generator for Microsoft teams very very shortly about myself and Lee I'm Lisa Fraser I'm, an M365 and Power Platform consultant from Germany I have an MVP for M365 development and business applications. If you want to, you can find my blog at M365 princess.com, and please also do follow me on Twitter.

B

Okay and my name is Lee Ford I'm, a solution architect for a company called Smitty in the UK I'm, mainly work in teams app development, but it also sort of bleeds into other parts of Microsoft 365 I blog at Lee, hyphen 4.k, and you can follow me on Twitter at the underscore Ford.

A

Okay, what time is it I think it's demo time so I.

B

Went a bit early, sorry.

A

Yeah a little bit early, that's what he said: yeah.

B

Okay, so uh here we go all right, I'll leave it again: okay,.

A

So we generated um a good device generator, so purpose of that is to send positive affirmations on a schedule and whenever needed, to people in the workplace so where they actually work, which is, of course in teams. So we will trigger our bot right now.

B

Okay yep, so uh let me I'm going to trigger it through uh in Azure, but uh we can also trigger it in routine, so I'm just going to here. Let's just do a run, make sure it actually works.

B

We think about it. Oh God that doesn't look good.

A

B

A

Just like from teams so just like sending any message to that bot and it will return a very, very nice message as an.

B

Adaptive card to accept it I think so we should hopefully be getting a message at some point that.

A

Would be nice so last one right before the call it worked, hey that Saucy personality should be taught in school. You feel me, so that is the kind of vibe that we wanted to give our uh people in the workspace, which, of course, is a very, very cool thing, and we would love to show how we accomplished that. So how are we doing this? Like.

B

Now there we go there, we go okay, yeah yep, so that one's come through yeah. So.

A

It's your life's journey Rose like a line you Hidden Gem s, so whenever you it just like, send a message right now to the chat, because we want to know okay. So uh when we uh try to build that, we had just like several um design goals and I think we have a slide for that as well. So first of our design goes what we want it to be secret, less as much as possible, so I hate to deal with Secrets, because it's also like um rotation is an issues. Storing is an issue.

A

um It expires that then you want to demo it. That is an issue as well, so uh we wanted to go as secret less as possible. uh Second thing is: we wanted to be serverless because I don't want to host that myself, so um you can imagine that which is like super super big fans of azure functions, which means that if we don't use it a lot, it will not cost a fortune or almost nothing, which is very, very cool and third's design goal was to be open source because we.

B

A

In community and if you want to contribute to this bot, you can absolutely do this as well. Okay. So what is the architectural Lee.

B

Yeah, so the architecture is uh 100 hosted in Azure, um so the main component is azure functions. So that's all the application logic is stored. That's where, when we send out a message, it comes from Azure functions when the bot gets triggered, it goes into actual function, so that's the main sort of component and what you'll notice here is.

B

We do have a couple of managed identities and, like you said, one of the sort of key design goals was having no secrets so we're using managed identity in two ways of using the user assigned one to communicate with the bots, so we're not using any client Secrets or anything like that, and similarly with Cosmos DB, first database storage and for the storage account, which is required for the Azure function, to run again using managed identity from the Azure function, to those components.

B

And finally, a key vote. Is there as a backup for things that we can't really store as a managed identity, things like Uris and tenant IDs, and things like that.

A

Yeah, so for everyone here is that super familiar with what? What is a managed identity? Usually, you would need to have an app registration and that app registration comes with the secrets, and then you have that problem to handle that secret, and all of that is obstructed away from you. If you use a managed identity, because then Microsoft cares about that and the secret is not even exposed to you nor to your developers.

A

So there is no way where you can just like know that secret and need to handle this on your own, which is just like a perfect use case for that.

B

So um I think now, if we go back to the the Azure portal, we can quickly go through just the sort of the the basic construct of the the application. So.

A

um Show me some code I.

B

Want to see this, you want to see the Azure function.

A

B

So, let's, let's go into into vs code, so we have two main entry points for the bot to work to be able to get a good vibe. So the first one is where we, uh the bot uh bot, is listening on messages coming from teams, users and then it triggers this Azure function and this Azure function effectively creates a good vibe bot and the bot is there if I just show you the bot code. uh Essentially we have on message.

B

So when we receive a message from the team's user, we store the conversation uh reference and the reason we store computation reference is that when we want to send out messages sort of uh scheduled, we have a conversation, reference I. We know where to send it to it's, not just an email address, it's a it's an ID and a name and a and a few other bits. So we store that in Cosmos DB um and then we get a Vibe.

B

So if I just quickly show what we mean by getting a Vibe, we have a list of phrases so we have um uh and effectively we we um combine them um to to generate a good vibe and once we've created a good vibe, we we send it on an Adaptive card out to the uh out to the user, and then the final bit in here is on a member's added.

B

So when a member is added to the to the the or the bot is added to a person's chat, we send a welcome message, say: hey we're the good five spots, send us a message and we'll send you a good vibe of that sort of thing. That's one type of interaction and then the other interaction is, which is what I showed you earlier, where we trigger the the bot to run um on a scheduled uh interval.

B

um Obviously, I manually run it that time, but that's where we're using durable function, which means we can use it at scale or we can kind of send out a thousand messages rather than sending one message at a time we can kind of scale out and what we're doing.

A

Here we wanted to just like sound thousands of.

B

Them yes, exactly right, yeah! So so so, obviously in our scenario, it's a couple of us, um but if this was rolled out to an organization in theory, anyone could um interact with the bot. You know and all of a sudden we have a thousand people wanting Good Vibes every day. Then then it's kind of built ready for that so similar to the way it worked before we're getting the conversations which we stored in the previous method, where we communicate with the bot.

B

For the first time we uh generate a good vibe for each conversation, and then we send that good vibe for each conversation uh using the the bot framework.

A

So um the interesting thing is we randomize all these phrases, so we have four kind of phrases. First, we have just like a greeting phrase, which is something we have a slide for that exactly.

B

A

Have a slide for everything, yeah yeah, so uh first is uh a greeting phrase like hey or listen up so just like to to have a nice hook uh for our new conversation. Second part of that is we want to address a personal trait or something about your personality. Just okay, your hair, your personality, so something about the person that we want to send the affirmation out to uh third part is the complement itself. So the nice thing that we want to say about us like is awesome or absolutely rules the world. Something like that.

A

C

That yeah yeah.

A

It should be a teachable skill, because I believe everything is a teachable skill um and then we end with something we end the vibe with, uh and that's uh just a fact, and we um well, we close this Vibe again, and what we did is that we we store all these phrases in Cosmos DB and then we randomize that so that we can just like have a lot of different Vibes.

A

Even if we have just like a limited set of phrases but they're, just like yeah a lot of just like combinations possible, so that it does not feel always the same for our users, even if they get more and more and more of these five speakers, they can trigger the bot by just like texting to it in teams or it runs on a schedule, and we have a chrome stop in place for for us. It's now at 12 pm every single day, but um you can do whatever you like with that.

B

Absolutely okay um and yeah.

A

So we still need to show the Adaptive card because.

B

A

How this then arrives in teams right? We.

B

Wanted to send this.

C

Yes, as an Adaptive.

A

Card and of course, the Adaptive card has some minor code as well, so this is like super easy because uh we only sent the git Vibe and then the future which says made with love by working on it. So um you can get that as well and customize. This message, of course, and then it arrives in teams and our user is happy to make you now happy. We of course support you with deploying uh this git Vibes generator.

A

So if you feel like hey, this is something that I want to try out as well, or maybe you do not want to send git Vibes, but anything else that can't be just like stand on a schedule or just like by triggering something, and you would think hey I can reuse that architecture. We um we have something in place, so uh we have Powershell script, that that deploys everything for you. uh We use infrastructure as code with bisat, because, uh let's quickly brackets I really appreciate that so this is.

A

um This is the way how to do this. Also, it's the the energies uh that joke in the world, respect from arm templates, the bicep and I really like that.

A

um There is a ZIP file still that you will need to unpack. This is because sometimes it can be really really slow when you um have this cold start in the very beginning, so we provide you with some files to make that easier, and last but not least, we of course have a team's app package, because otherwise it would be a teams application. So we provide you with that as well, and all of that, of course, is available on Kitsap.

A

uh We will submit a PR uh right now, it's uh on github.com working on it, slash git Vibes generator, and we hope that um you can just contribute to that as well or find more use cases and find this valuable. Thank.

C

You excellent thank you Lee. Thank you, Luisa, really, really cool, stuff and and great to see. um Well, don't forget to say good wipes. That's always really good, so the positivity is the the engine of the world to keep us moving. So thank you for that really really cool stuff.