Episode 20- Azure OpenAI Assistants via Logic Apps with Konstantinos Passadis - podcast episode cover

Episode 20- Azure OpenAI Assistants via Logic Apps with Konstantinos Passadis

Nov 02, 202413 minEp. 20
--:--
--:--
Listen in podcast apps:

Episode description

Send us a text

Join us as we explore the fascinating intersection of AI and automation with our special guest, Konstantinos Pasad, a solutions architect hailing from Athens, Greece. Kostas shares with us the journey of integrating cutting-edge Microsoft technologies like Azure, OpenAI Assistant, and Logic Apps into real-world projects. 

Transcript

Speaker 1

Good afternoon and good morning , where you are . Welcome to Microsoft Community Insights Podcast , where we share insights from community experts to save the day in Microsoft . I am Nicholas and I'll be your host today . In this podcast , we will dive into Azure , openai Assistant , azure Automation with Logic Apps .

Especially , before we get started , I remind you to subscribe to our social media so you never miss an episode and help us reach more amazing people like yourself . Today we have a special guest called Konstantinos . Can you please introduce yourself please ?

Speaker 2

Hello , nikolas . Yes , I'm Konstantinos . Pasad is my last name . Kostas is a shorter name , an easier one . I'm from Greece , athens . I work here as a solutions architect for a multinational company working remote . It's a nice thing actually to work remote and I'm so glad to be here . It's an honor for me . Nikolas , thank you for the invitation .

Speaker 1

Okay , thanks . So before we get started , we just want to dive into your background . So what do you normally do ? Has your everyday job as a solution architect .

Speaker 2

Yeah , basically , Basically , I'm a multi-role architect . So , as a description in the job , role is to design and create solutions for specific demands by the customers , right ? This is the description for a solutions architect and this spans across many vendors .

It is a vendor agnostic role but since I'm mostly on Microsoft , most projects are coming from the Microsoft side of things for customers as well . But I'm also an engineer and that means that I'm quite engaged into actual hands-on , meaning I create Terraform configurations . I engage into different projects with data , AI , many things , Also M365 .

Recently I had a project quite a complex one with Power BI and M365 and Graph API , complex one with the Power BI and M365 and Graph API . I never wanted actually to leave behind the engineering part of me . I want to have hands-on , I want to create things . So I came with an agreement with my current employer and I love this flexibility .

I'm able to design and provide the architecture , but also engage when needed and if I have the time to Okay yeah , because you've got lots to learn , if it varies from different areas like 365 , as your architect as well .

Speaker 1

So what are some of the projects interesting projects that you've worked on where you are ?

Speaker 2

Interesting projects , apart from migrations , because you know everyone is doing migrations . Interesting projects are a data warehouse quite interesting A new project with Databricks and PromptFlow Azure PromptFlow , where the client had to decide about how he would do it and I was there for that . Quite interesting that , and the recent one with Azure .

Speaker 1

Virtual .

Speaker 2

Desktop and specific requirements about that deployment . A variety of diverse projects . To tell you the truth , Nicolas .

Speaker 1

That's brilliant so how's our work today , How's our team ? Is Azure , OpenAI Assistant and Logic Apps Automation in your opinion and experience ? How do you think Agile OpenAI integrates with Logic Apps ?

Speaker 2

Yeah , let's talk in reality , right , because I was quite excited to see the AI assistance and vision in AI studio , right . But then I started trying it and understood and realized that we are in the genesis of actual AI automation .

I think this term is going to be introduced more and more into our lives , especially for us engineers and people that are into the Microsoft Realm and into the AI , of course , practice . So we're talking about AI automation , right . I was expecting to see something quite easier for users to be able to utilize . I think it is still under development .

Of course , it is in preview , and that means that we still have to expect new things to come better user interface , better response times , better everything . I'm very satisfied , but I think it still needs work to be done . From the perspective of Logic Apps , we have a mature service , something that is working . Many people already are using it .

I'm not afraid of that service . It is quite mature and it is very fast as well . From the OpenAI perspective , we still need some things to be done there and I'm seeing changes . That's the positive thing . I'm seeing changes .

Speaker 1

Yeah because I know that you can do other functions with Flowchic as well with OpenAI . Because you can do , you can automate the deployment as well for OpenAI .

Speaker 2

Automated deployment of whatever you mean .

Speaker 1

Yeah , so like modules and things as well .

Speaker 2

Yeah , what are ?

Speaker 1

some . What are some of the use cases that you used OpenAI systems for ?

Speaker 2

I tried to do the easy thing summarize a blog , blog post , for example , or get the latest from something and summarize it for me , and it worked . And the other one was to create a new virtual machine just from the prompts , and I was amazed that I saw it really worked and there was the potential I discovered . We have the potential to have , I don't know .

You may think of anything actually , from learning platforms to whatever , where the user just prompts the interface , the assistant and the assistant creates resources for the user . I think that's a very powerful connection there .

Speaker 1

Is there any benefits from using like Logic Apps with OpenAI systems ? So is it faster deployment as well ?

Speaker 2

Yeah , I think the benefits are on the automation part from the Logic Apps and from the interaction part , coming from the open AI , meaning that you can keep that interaction between the AI model and the user and you can have , on the other hand , the full-blown automation provided by logic apps , where really the limits are the sky or really the limits are your

ideas . I think we will benefit from that . Again , I'm returning to the term AI automation would be something that we will see more in the future .

Speaker 1

Hopefully the product will get a little bit more mature and easier to understand . Yeah , understand . In your experience , is there any challenges that you encounter when you integrate Logic Apps with AI for AI systems ?

Speaker 2

The basic challenge was that how can I say that ? You have to get your hands dirty and get the code from the Logic Apps ? Create correctly the schema details that really make the difference . It's not that simple . Starting with it I thought it would be more simple , but while I was start using it I realized that it's not as simple .

You have to get your hands dirty .

Speaker 1

You have to test the flows as well , if each of this if matches schema outputs . Yeah , so I've used Logic Apps for a while . So , instead of speaking of AI systems , what are some of the best practice would you recommend people to be aware of when using AI systems with Logic Apps ? Yeah , be aware of when using AISs with logic apps .

Speaker 2

Yeah , the first thing is be very careful with your token consumption . You may need to expand a lot of your tokens and also so the assistant knows and provides the interface . But this is again my own perspective here . Right , have a longer system message and a smaller interaction in the chat .

For now , we can only work with consumption logic apps , but I have seen a tech community blog post that we will be able to move on to standard plans , so that will be a blessing , because not everyone wants to work with consumption . For example , we don't have a high availability here . You could have consumption plan . Also , be careful with repetitive tests .

The consumption plan can easily skyrocket your costs with repetitive tests . Each time you run a workflow , it is taken out of your budget , so be very careful on that . You may need to apply some budget or some mechanisms to control your costs . I think it's a service that will have a future and we will see big things with this one .

Speaker 1

That's brilliant , so just before we wrap up this episode , is there any last-minute tips you want to give to viewers when learning about Logite Apps and AI assistance ?

Speaker 2

No , no , no For now . No , I'm just happy that we're able to see the thing . I'm happy that we saw the errors . For me , it's a blessing that we can see errors , that we can see things happening . That's what I mean , and I'm also happy that we finally met Nicolas . I hope we will meet also in the future .

Speaker 1

So how can someone get started with using AI Assistant aside from checking your blog posts ?

Speaker 2

You need to be able to create an AI Hub in your portal , or go to aiazurecom and start creating a new hub and then a new project , and you will find the assistance here in the playground menu .

Speaker 1

Just be careful , okay yeah , so we always like to get to know yourself . So are you going to any events , community events , where you are ?

Speaker 2

for now it's a difficult . Uh . We are doing some online things . We are planning for the global azure 2025 . We are planning for the Global Azure 2025 . And we are planning some internal stuff here in Greece . I hope I will be in March with the other MVPs in USA . Okay .

Speaker 1

Yeah .

Speaker 2

That's the plan for now . Maybe maybe in Adverb .

Speaker 1

Next one . Okay , thank you . So , yeah , so , thanks for joining this episode continues . So I hope you that was very good . It's a good experience to know that you could use logic apps with open ai assistance , namely , you can do that . That's why you're going to use it with open the air itself .

So so , in few episodes , in in few weeks , is going to be on YouTube and social media . So thanks for joining this episode . Bye .

Speaker 2

Bye-bye .

Transcript source: Provided by creator in RSS feed: download file