Azure AI Agent Service | Tutorial #1 | Getting Started Tutorial

Key Takeaways

Hi friends, Naji here. Welcome to my YouTube channel. We are starting a new playlist on Ashawa Foundry and this is a introductory video on Ashure Foundry and where where we will discuss on agent service and how can we create an agent service and we'll be using the ashure playground to do some kind of a trial line or some kind of a what we call the chat with respect to the the agent service. Okay. So let's understand how the architecture will be look like okay or in a high level architecture of um ashure I foundry or maybe like ashure a service okay so here we can see mainly we can divide it into consider it as a three-pillar architecture so one is a built-in enterprise readiness and the second one is the extensive ecosystem of tools and third is the model catalog so let's understand uh what is like enterprise readiness and features so they this can be divided into two parts. One is the security and storage and other is the monitoring and the authorization. So here we can use the BYO file storage and BYO search index. Okay. So here BYO means bring your own file storage solutions for complete data control. So we can use our own um data or storage uh for this uh agent service. Okay. And we also use the our own uh like the search index also can be bring here for the seamless integration. So similar way we can use the OBO authorization support and also we can use the observability tools. The middle part of this architecture is the comprehensy tool and ecosystem. So there are two type of tools that is available. One is the knowledge source. So we can use uh Microsoft fabric as a knowledge tool and sharepoint and bing search and ashure a search and the license. This can this all can be the knowledge sources and also we can knowledge tools. This can be tools and other is the action tools. Okay. So the action tools will be like ashure logic app maybe open a 3.0 tools and Ash ashure functions. Okay. So these all things that we can explain in detail when we do the practical examples. So next understand um the flexible model selection. So in the initial uh part of maybe like in the initial stage of the Ashure AA service or Ashure AI platform they were like only supporting the open AA service but now they are flexible and they are uh integrating a lot of other opensource and the closed clo closed size model so models also okay so in the Ashure open a service like they are proing different models like GPTO and different GPT 5.1 and 4.1 when all those model service are supporting and uh other things like llama, mistral and coher and lot other models are supporting uh other than the opera service. Okay. So that is um overall about like ashure AI foundry and it's in general it's a enterprise ready infrastructure with security uh security and the complaints building extensive ecosystem of knowledge source and action tools flexible model catalog from open meta and coher and seamless integration with existing ashure service and the workflow. Okay so this is overall introduction about the ashure AI foundry. Now we can see uh how can we uh create a agent in ashureai service. So I'm I'm using aa.ashasure.com and I already log in with my email id. Okay. So I have a outlook account and I used to my outlook account to log to the aa.ashure.com. Okay. So this is the uh I'm using new new theme. So let me uh Okay. So before going to that I have to create a project. Okay. So here this is my uh the initial screen. After login I can see this screen and here I'm going to create a agent. So let me go and create an agent. It's a straightforward and uh uh this is the project name. Okay. And if you expand this you guys can see my subscription is pay as you go as as you go. So that is the default one and this is my resource and region is US East one and public network as is enabled and resource group is also new resource group I have selected. Okay, now let me go and create it. So the creation process will take some time. Okay. Meanwhile um I can explain like um few more things. Okay. So these are the basic models that is supported by uh [clears throat] that is supported by the Ashure AI service. Okay. So but one thing that you have to understand this models all the models is not available in all the regions. Okay. So based on its um it's decision made by the ashure team. So [clears throat] not all the models are available in all the regions. For example if you look on this uh GPT5 2025807 which is available in only the few regions. You can see that right. So similar way the other models also available in different different regions. Okay. So I think GPT 4.1 is available in almost all region. So so sometimes like if you select a model you have to ensure that the model is supported in that particular region. So similar to that there is like another uh scenario that is ashure open a foundry the kas and limits. So that is also the kas and limits also will change with respect to the regions. Okay. [snorts] For example, here we can see uh the kas and limits are not enforced at the uh channel level instead of the highest. Uh let me show you here. Okay. So you see this tokens per minute and the request per minute limits are defined per region per subscription and the per model and the deployment type. Okay. So you also need to take care of this. So no need to worry about this. Look when we when we implement the project then you will understand like what what are the problems and how can we solve this. Okay. So let me go back to and see like if my um deployment is done. Okay. So here um I have created the project then they are asking about deploying a model for model in that particular project. Okay. Particular agent service. So here I can see I'm going to select a model that is the GPT 4.1 mini. So let me select this. So I can see the description here. So if you want you can go and read this. But let me confirm this. >> [clears throat] >> So it's in progress. Um so it's uh giving me the deployment type. I'm going to select it as a developer and I can customize if I need it but I'm not doing anything. I'm just going to deploy this. So let's deploy. So it says um deployment is failed. Uh invalid uh resource properties for the specific SKU developer tire. The account deployed is not supported. Okay, so this is not supported. So let me cancel this one. Let me select a deploy model. Maybe I can select like another model because I selected the developer that's why it's giving me that error. I'm going to select GPT4 of mini. Execute it. Deploy. Okay. It's also at least um the specific capacity zero for account deployment should be at least one hour but more than I tried few before that may be the reason but let me see uh I have also have done few clean process. I'm trying it again. We see I made it to standard. Okay. Then I am trying. Now it is deploying. Okay. So now we can see. So let me go back to the agent agent menu. Okay. So here we can see this is the agent that I have deployed and we can see GPT4 mini is my model I have deployed. Now if you come here in the right side we can see different options here. Okay. So I can see the agent ID, agent name can be seen here and deployment can be seen here and this is the instruction. Okay. Okay, this is the instruction that um by default it is using if needed I can update this. Okay, and uh the agent description is mentioned here. Let me go back. Uh let me scroll it down. See this is the knowledge. So there is no knowledge is added here. So if I click on to add we can see different knowledge can be added. I can add a file. I can add a service search. I can add a Microsoft fabric. I can add anything listed here can be added. Okay. But I'm not going to do that this point of time. And actions as we said uh I can do the actions like code interrupter can be used. Open a 3.0 and aure logical app can be used. Okay. And uh if you want to connect like other agents I will able to do that. Then definitely I can use the uh temperature and also topy. This also can be customized based on our need. Okay. So that part is done. Now what I can do? I can go back and do some go to the playground and I can do some kind of a testing here. So let me go back to the playground. Okay. So this is my playground here. I can type hi. Okay. It says how can I help you assist you today? Okay. I can ask who is the uh what is um the capital uh maybe I can say who is the first uh president in US let's see okay so it's giving me reply like the first president of US is the uh was George Washington and it's giving me the details as needed. Okay. And if you look on the thread, we you will able to see the logs also. So, let me go back and see the thread. Okay. Uh playground. Let me go back to the agent. Let's see my thread. Here we can see the thread. So, let me click on this. Here we can see thread is a kind of a kind of a log that we are keeping. Okay. Here we can see the uh the chart that we have made. Okay. So in this way what we can do we can create our own agents and we will able to do some kind of a uh trials using this playground. Now this is the old version old uh UI style and if I want to see the latest version I just click into the new foundry. So this will give like more uh more beautiful UI. Okay. So let me show you that. See this is the new version. I can see now let me go to the home screen. Where is my home screen? Uh I can go and see the build. Okay. And I can see let me click on the agent view the classic agent. Okay. So this is the agent that I have created in the other view. And if I want I can go and create from here also. Let me go and create a agent here. I can give agent as a test one. So in our coming videos we will be using this UI because I love this UI better than the other one. Okay. So I'm creating a agent agent name as test one. Let's see. Um here also we can do the same things. I can attach the tools. I can attach the uh knowledge and the memory and I can also add the guardrails. Maybe let me see a hi uh I want to see that it's working or not. Hello. How can I assist you today? Okay, I can ask what is your name? I'm called. How can I help you today? Okay, my name is Naji. Let me see if it has a in memory built-in memory is there. Okay. Uh do you do you know my name? Let me see. Yes, you just told me. So it's it's a memory is also the context is it is keeping the context also. So that is also good. And we can see different things like traces will be there, monitor will be there, evaluation will be there. Okay. And if you go to the models, okay, so we can see the different models deployed in this um region. And if you want to do the finetuning part, we can do a finetuning of a model that we can discuss in our coming videos and tools um knowledge and data. Okay. So any data set that we can see here and evaluation and the guard. So these are the few menus that is available and in our coming videos we will be discussing more about this. So as a introductory video I just want to introduce this and I want to show you guys like how to create an agent and how can we do some kind of a uh trial chart with this in the uh playground. So that's the agenda of this meeting uh I mean uh this tutorial and I think I have covered that. So that's all about. Please do like and subscribe. Have a nice day. Bye-bye.