Azure OpenAI and Semantic Kernel Development

1. Introduction: Hello and welcome to my course. Understanding Semantic Kernel SDK. I'm your instructor Tervor Williams. I'm a software engineer, lecturer, author, and a enthusiast, and it is my absolute pleasure to deliver this course to you. Before we jump into the course, let's review some of the main objectives that I hope to help you to accomplish from this course. Now, one thing we want to do is gain a comprehensive understanding of the semantic kernel SDK. This means we're going to explore the core concepts, architecture, and capabilities of this SDK relative to.net development. We're also going to look at how we can develop AI powered plugins that are model usable and enhance the application functionality with AI capabilities. We also want to look at how we can integrate AI models into applications by using LLMs like those that are available from Azure Open AI service and how we can integrate them into our applications. We also want to make sure that we leave this course understanding how we can use this SDK to implement real world applications like virtual assistants or career coaches or systems that can help with decision making powered by AI, of course. Also want to make sure that you get to apply all of the skills that you are learning through hands on activities and projects to reinforce the learning. So some of the key features of what this course has to offering through One, an overview of the semantic kernel SDK with in depth explanations of its purpose and its role and how it can be used to make AI integrations into your modern applications. Also have hands on guided exercises where we're going to put everything that we learn about the SDK into practice. And, of course, we're going to be building on top of LLMs provided to us through Azure Open AI. In this course, we're focusing on the GPT LLM. We also are going to have real world use cases because we're actually going to be building a coaching system that can process a user's query, retrieve historical data, and then offer a personalized response. We're also going to port this into an application for the web. And we're going to look at advanced features as we go along like memory planning, reasoning capabilities, configuring AI with long term support or retention and several things that you hear about with AI, we're going to be looking at how we can use the semantic kernel SDK to accomplish all of these. Now, before you take this course, I advise you that it is good if you are already familiar with Tutent and C Sharp programming, and you are comfortable. You have those foundational concepts like object oriented programming, how to incorporate packages and libraries and how to use ABIs. It's good if you have an appreciation of all of that before taking this course. Also, if you have an understanding of Microsoft Azure, that is good. Though you don't need an in depth understanding because whatever you need to understand, I will explain and will use only what is required for this course. It's also good if you have a basic understanding of AI and AI concepts and keywords. As I will be using some keywords, I'll try to explain as I go along. But if you already have that knowledge, then it will be easier for you to assimilate this information. Now, as a recommendation, I suggest that you take my course, generative AI for.net Developers with Azure AI services, where I go from scratch to completion going through the different concepts that make up AI and generative AI and the different services that Microsoft Azure offers to us. So by the time we complete that course, everything we're doing here would come naturally. Now, if you already have the foundations of the previous three points, you probably don't need to take the course. But I do make a recommendation so you can see the whole picture as you go along. With all of that said and done, I want to welcome you to discourse. And well, let's get started. 2. What is an LLM?: All right, so let us start off by discussing what an LLM is, and LLM is an expression or an abbreviation that you see a lot when you're dealing with artificial intelligence, especially nowadays. So a large language model or LLM is an artificial intelligence model that is designed to understand and generate human like text based on vast amounts of data. So you'll find a lot of generative AI applications are powered by LLMs so that they can perform natural processing or natural language processing tasks. Natural language means we are speaking to the AI the way I'm speaking to you. I'm using English. So it's not about using a programming language. That's traditionally how we talk to computers through programming languages. No, LLMs allow us to speak English or native language to the system, and then it can understand what is being asked of it and then produce a response as best as possible. And, of course, any AI engine is only as powerful as how it has been trained. So LLMs tend to be trained on a lot of data on a lot of things, so they can understand a lot of variations of our natural language prompts. No, models like Open AI, Open AIs GPT Series and Google's BRT are prominent examples of LLMs used in various applications, and pretty much an LLM uses deep learning techniques, and it is capable of performing things like language translation, summarization, and conversation and sentence completion. 3. What is Generative AI?: Now let's us briefly look at what is meant by generative AI. So artificial intelligence is designed to imitate as best as possible human behavior by using machine learning to interact with the environment and execute tasks without any explicit directions on what it should produce. So generative AI describes a trench of AI that allows the system to generate original content. Now, people typically interact with generative AI through chat applications, and an example of this would be like Microsoft copilot or hat GPT. And generative AI applications accept natural language input like English and return, sorry, appropriate responses in natural language images or code. And what's most important is that it generates this on the fly. So it's not going to do a predetermined thing all the time, but it is trained on a lot of data, like we said, backed by the LLM, and when it gets a prompt, it can look at the vast amount of data, make the connections, and put them all together and return a relevant response. And that is why when you're using the co pilot or the Char GPT or any other AI engine based on generative AI technology and LLMs, you'll see that you can generate original content just by making a suggestion or asking a question. 4. What is Semantic Kernel?: All right so now that we've laid a little bit of the foundation, let us jump into the main topic of why we're here, the semantic kernel. Now, the semantic kernel is an open source SDK that empowers developers to build custom AI agents. The semantic kernel allows developers to combine LLMs with native code and create AI agents that understand and respond to natural language proms. It is an efficient middleware that enables rapid delivery of enterprise grade solutions, of course, AI powered. And best of all, there's support for it across C Shar, Python and Java. So here's a little graphical representation of how it works, and I've taken this diagram from Microsoft documentation. So there is what we call the AI Orchestration layer. This is the core of the semantic kernel stuck, and it allows seamless integration of AI models and plugins, and this layer is responsible for combining these components to craft innovative user interactions. Then we have the connectors. So the semantic kernel offers a set of connectors that enable us as developers to integrate the LLM directly into our existing applications, and these connectors serve as the bridge between the application code and the AI models. Then we have the plugins. Now, the semantic kernel operates on plugins, essentially serving as the body of the AI app. The plugins consist of prompts that you want the AI model to respond to and functions that can complete specialized tasks. So we have some built in ones, and, of course, we can build our own, and we'll be looking at both of them in this course. 5. Why use the Semantic Kernel?: Alright, so now we know what this semantic kernel is, you're probably wondering, Okay, so why should I use this thing? Well, for one, it integrates with LLMs like Open AI as your Open AI and hugging face to name a few. Basically, it allows developers to create plugins to interface with the LLMs and perform various tasks. So the SDK provides built in plugins that can quickly enhance any application, and it allows us developers to easily utilize LLMs in our applications without needing to learn the intricacies of each LLMs API. And that's the most important part. So it provides us an abstraction layer so that it allows for the simplified integration of these AI capabilities into existing applications. It also allows us to develop streamlined, reliable AI experiences while avoiding potential unpredictability associated with prompts and responses from LLMs when we use them directly. Finally, it supports the ability to fine tune prompts and plan tasks to create controlled and predictable user experiences. 6. Understanding AI Agents in Business Context: No, another key expression that you're going to see is AI agent. No, an AI agent is a software entity that can make decisions, take actions or interact with an environment autonomously to achieve specific goals. These agents use artificial intelligence, including machine learning and natural language processing, and they can understand and respond to their environment, often adapting based on feedback. They're designed to actually specific tasks or goals like recommending products, answering customer queries or controlling a robot in a physical space. So some tangible examples of AI agents include chatbots and virtual assistant, which are AI agents that interact with users. They can answer questions, provide information or help with general tasks. You also have recommendation systems like systems that suggest sorry products, movies, and content based on user behaviors and preferences. We also have robot process automation. These agents can handle repetitive tasks like data entry, form processing or scheduling, and you have robotic AI agents that operate in a physical environment like a warehouse robot or an autonomous vehicle or a home assistant. Now, in the context of a business, AI agents can help to streamline operations, improve customer service because they make applications smarter and more responsive to user needs. They can also help with decision making by automating tasks and providing insights. 7. Visual Studio Code: Now we have Visual Studio code as the alternative. Now, this is a great cross platform alternative, which means it will work for Windows, Linux and Mac versus Visual Studio, which is Windows only. And, of course, it is compatible with the Net. And then there's this extension called C Sharp DevKit, which is very useful for helping to give us that visual studio feel and kind of bootstraps on for.net related operations like creating projects and interacting with our project files. But for some things, you will be required to write some CLI commands, which is fine. We'll go through those, and actually for this course, I'll be using Visual Studio code more than I'll be using Visual Studio. Now, to get Visual Studio code, you can go to code.visualstudio.com, and from here, you can download. You can download for Windows or you can download for other platforms. I think this will actually default based on what platform you are on when you browse to this page. After a very simple and straightforward installation process, you'll be given your visual studio code window. And it's really just a text editor, but it can be enhanced through extension. So once you're here, you can go to the extensions tab, and from here, you can look for the C Sharp Def kit. So once you click on that C Sharp Def kit, you'll see the button there to install. Here, mine is updating, so you'll see that it's currently install ink. But for an extension you don't have, let us say the Scubnt is one. You'll see that button, we'll say install. So when you go to the C Sharp Def kit, you'll see Install and click that. Of course, if you don't see readily, you can always search and once you find it, then go ahead and install it. It may take some time. Now, you'll also need to install the.net SDK manually. So this comes with Visual Studio, but if you're using Visual Studio code and or you're not able to use Visual Studio, then you have to go and get the dotnet.miicrosoft.com and get the SDK that you want to use. Now, you can install both. This one is the recommended one. And as I said, standard term support, and it was released very close to when I'm doing this recording. But what I'm saying is that either one works, you can install both, and you can use either one to create your projects, and it's quite easy to upgrade from.net eight to.net nine if you decide to do so later on. Now, you'll see here this is same for Windows, but you can get all versions. All you have to do is go to A and I'll go to all.net eight downloads, and from here, you'll see that you do have versions for different operating systems that are not Windows based. So once you have all of those installed, Visual Studio code, the.net SDK. And I recommend actually, I got the order mixed up. I apologize. Do Visual Studio code first, then install the SDK, and then attempt to install the C Sharp DevKit. As it does say that the dotnet SDK is one of the requirements. So to avoid any problems, do it in that order, and then you can continue. 8. Microsoft Azure Account: Now, what is Microsoft Azure? Microsoft Azure is Microsoft's flagship Cloud computing platform that provides services like computing, networking, storage, and databases. It offers seamless hybrid cloud solutions, and it allows businesses to integrate premises data centers with cloud resources. Also provides easy scalability, so business can quickly scale resources up or down based on demand and other factors. Now, most importantly, Azure includes built in AI and machine learning services like Azure machine learning, cognitive services, bot services, and others, and all of these can be used when developing intelligent applications. Now, I'm just going to quickly walk you through what it takes to set up an Azure account if you don't already have one. First of all, you need to create a Microsoft account or sign in with an existing one. So your journey starts right here at azure microsoft.com, and I'm sure they're going to redirect you to the site that is most appropriate with where you are in the world. So from here, you can go ahead and sign in if you already have a an Azure account? Well, sign in. If not, you can click New to Azure and start free. And then you have to choose your subscription. So you do have the option for a free trial, which allows you to explore the platform and its services without any upfront costs. You do get some amount of money for, and they do give you the lowdown down here. So let me not try to quote and mislead you, but you get there we go, $200 for 30 days to try out services. But for the course of a year, you actually get a lot of free services and access to free services throughout that 12 month period. Alright? So you can start off for free. They do require that you add some payment information, but there's no upfront commitment. You can cancel anytime, and your credit card will not be charged when that free subscription is done. So you don't have to worry about the credit card being charged. Once a free subscription is done, then you'll just lose access. But then, if you don't want the free and you just want to go straight into the paid or when the free is exhausted and you have to start paying to continue using the services, you do have to pay as you go option. Now, this gives you more resources, more features, and of course, it's paid. So there are far fewer restrictions on what you can and cannot do, and you only pay for what you use. That is a great thing. So a lot of these services you can spin up, especially while you're learning. You can spin them up, and for the few minutes or hours that you may use the service, you will be charged. But once you destroy the service, which I will be showing you how to do, how to decommission the service or deallocate the service, then you will no longer be charged. So this is a great way to go as you're learning and you don't want to spend too much money, but you do not want the restrictions that the free account might give. However, if you're just starting with Azura, I do suggest that you start with the free account, so you can see what it is capable of, and the free account is good enough for this course. So after you've selected the suitable subscription plan based on your needs and your budget, then we're going to go ahead and set up what we call a resource group that is used to organize our resources into one logical container, and then we're going to explore and utilize different Azure services like machine learning, the app service, the SQL database storage. 9. Semantic Kernel Overview: Or now we're going to get started, and we're going to build our very first kernel in this section of the course. Now, brief overview of what we're going to do. We're going to look at the kernel, what it is. We already discussed the concept behind semantic kernel. Now we're going to get a little more in the specifics of how exactly works. We're going to look at using Azure Open AI, and we're going to complete a simple chat up with the semantic kernel. So stick around. We're going to explore some concepts and get into some practical examples in this section. 10. What is the kernel?: Now, what is the kernel? The kernel is the central component of the semantic kernel. So remember the semantic kernel is really the SDK. What exactly is the kernel? So it can be considered as a dependency injection container that manages all of the services and plugins that are necessary to run the AI application. So remember that the semantic kernel on a whole is really a wrapper around the underlying SDKs that each AI engine would provide. So open AI as your Open AI, hugging face, all of those have their own SDKs. The semantic kernel on a whole is a wrapper around them, and the kernel itself manages the different services in your application. So all the services that you expose to the kernel will be seamlessly used by the AI as needed. Any prompt or code that is run in the semantic kernel will use the kernel to retrieve the necessary services and plug ins. Now, the kernel is extremely powerful because it means that as a developer, we have a single place where we can configure the most important parts of our AI agents. We can also monitor them. And Okay, let's say, for example, we have an application. So we're going to look at a high level diagram that is taken from the Microsoft documentation. So we have our application, and then we need to run at AI prompt. So the kernel is going to select the best AI service to run the prompt. It will build a prompt using the provided prompt template, send the prompt to the AI service, receive and parse the response, and finally return the response from the LLM to the application. And throughout this entire process, you can create events and middleware that can be triggered at the different steps. So this means you can perform actions like logging and provide status updates to users, and, of course, implement everything that is required to make sure that you're practicing responsible AI usage. And once again, all of this can be done from a single place courtesy of the semantic kernel. 11. How to build a kernel: Now, let us review how we build this kernel. So using the semantic kernel SDK, it takes minimal setup. So we need the semantic kernel SDK package. We need some endpoint for an LLM, one or more, but in this case, we're just going to use one, which is Azure Open AI. We need the SDK to connect to the LLM via endpoint and run the prompts, and the semantic kernel SDK supports different LLMs like Hugging Face, open AI, and Azure Open AI. And if you're wondering what's the difference between open AI and Azure Open AI and what LLMs mean I want a bit more information, you can always check out my course on generative AI for dotnet developers. I go in depth in that course and take it from scratch about how to develop AI solutions. So for this example, we're going to use or for this course, rather, we're going to be using Azure Open AI and even more so for this example that's coming up. Now, while we're on the topic of what is needed to build our kernel in this course, and in this example that's coming up, we're going to just explore what the Azure AI services are. So we kind of went through what Microsoft Azure is and how to set up the account, and assuming that you've done all of that before, I'm just going to go ahead and let you know what the AI services are all about. So you do have Azure AI Cognitive Services, and you have Azure Open AI service. And I'm going to focus on the Open AI service since that's what we're using in this course. Now, the Open AI service offers access to powerful language models like GPT, and I'm sure we've all heard of or used hat GPT by now. And these can be used to develop intelligent applications like chatbots, content generation and language translation applications. So natural language processing is at the helm of LLMs work. And Azure provides robust NLP capabilities that allow us as developers to analyze, text, extract insights, and perform analysis like thorough sentiment analysis and other kinds of analysis through services like text analytics and large understanding or Louis for short. A little more about Azure Open AI service. Well, let's take a step back. Open AI is an AI research company that has produced several groundbreaking, AI powered applications like hATGPT to name one, very powerful one. Now, Azure Open AI is a service that gives us restful APIs and access to open AIS, powerful language models, and you may not know that these are different models that are available in the Chat GPT. But if you use CHAT GPT, I'm sure you'd be familiar with at least 3.54 and 40. But just to show you that there are other models that you can access from a developmental standpoint, and as you're open, AI exposes these to us and giving us the backing and the guarantee of Microsoft as you're hosting an infrastructure, of course. Now, with these models, we can complete things like content generation, summarization, image understanding, semantic search, and natural language to code translation. So there are several things you can do with these GBT models. Once again, if you want to go in depth in them, you can check out my course generative AI for.net developers. All right, so now that we have an understanding of what the requirements are for building the kernel and starting to build an application, we're going to jump right in and start off by provisioning our Azure Open AI resource. 12. Create Azure OpenAI Resource: Alright, so I'm assuming that you've already set up your Microsoft as your account and you are able to login. So once you log in, we're going to go straight into creating our Open AI resource. So I'm going to search up here for Open AI and actually have it right here, but you can always just type in the search, and you're going to click on Azure Open AI. And from here, we're going to go ahead and create that Azure Open AI resource. Do note that this resource will incur costs. So if you're on the free tier, then of course, you have the credits, and you might get very limited capabilities, but otherwise, you will incur costs. So after you're finished with a demo, do remember to tear down your resources to make sure you don't incur anything unnecessary. So for this course, I'm going to create a new resource group and I'm going to call this resource group semantic Kernel RG. So just something that I can identify with. So when I'm removing the resources, I just need to remove this resource group, which I've named very specifically based on the examples that we're going to be using. So I'm just going to go ahead and do that and then give this a name. And this name generally want it to be unique. So you could use something else to make it unique. And by the time you're doing this course, my resource will no longer exist, but you can always append your name on it to just to make it a bit more unique just in case they say it's not available. And for the pricing tier, I'm going to go with the standard. Now, if you want to see the full pricing details, you can do so via this link. And then I'm going to proceed. I'm going to confirm that I want all networks to be able to access this resource, and I don't need to tag it and then hit next. And from here it's going to ask me, or it's going to validate everything that has been entered, and then I can create. Once that completion step is done, we can go to resource. Now, this resource gives us access to Azure Open AI Studio. So let's go to studio. And while it loads, you might be required to reauthenticate, you can authenticate using your Azure credentials, and in no time you should be allowed in. So once we're here, what we want to do is create a new deployment. So this is us setting up the LLM that's going to power our kernel, right? The kernel is really just a code based connector to some LLM that we have. So once again, we're using Azure Open AI. And here, I'm going to go to deployments, and we're going to create a new deployment. So we're going to click Deploy Model and deploy base model. So just for context, you can deploy a base model because Azure Open A has several base models that are pre trained with certain information and for certain things. But we can also fine tune that base model or another model to make it more unique to our needs. So the needs of an airline company would be entirely different from the needs of a school. Right? So the base model can be a general, you know, go between for both of them because the base model would have been trained on a lot of every little thing or a little off every little thing. But then you really want to fine tune your model to fit your specific business needs sometimes. In this example, though, we're going to go with a base model. So let's deploy a base model, and we're looking for the GPD. So you see here, have several GPT models that are available to you and several versions. And each version has its own capabilities and would have been trained on certain information and up to a certain point. So when choosing the models, be very, very deliberate. However, I'm going to go with GPT 35 Turbo, 16 K. And if you don't see 16 K, you can go with the regular GPT 35 Turbo. So from here, I'm going to confirm. And then I can give the deployment a specific name. So in the case where maybe I have multiple deployments of the same type of model, I would want to give them unique names. In this case, I'll just leave it with the default name, which is indeed the name of the model anyway. And I don't need to change anything. If you want to change stuff, you can go to customize and you can modify the number of tokens being sent across per minute. You can change the version, right now, default is the only option, and we have default and default version two for content filters, and I'm not going to enable the dynamic quota. And yeah, I can now deploy. 13. Build a chat completion app: Alright, so now we're going to be writing some code. And in this lesson, I'm going to go ahead and use Visual Studio code and the.net CLI, just because it's a bit easier and it's easier for everybody regardless of your OS to follow along. However, if you're more comfortable with Visual Studio, feel free to follow along because we're just simply creating a new console application. So using the.net CLI, I'm going to run the command.net NU Console O Build your Kernel. So this means create a new console app and output this up to a folder called Build your kernel in whichever destination you have selected. So let's go ahead with that. No I can see the into build your kernel, and then do code dot Launch Visual Studio code in that location. So once Visual Studio code is launched, the next thing I want to do is bring up the terminal inside of Visual Studio code. And I could also do this inside of the original terminal window, but why not use the one inside Visual Studio code. So bringing up this terminal allows me to now type the command.net add package, miicrosoft dot Semantic Kernel. So this will now go ahead and fetch that package that allows us to support semantic kernel in our application, and then it goes the nuget fetches that packet. And once that is done, I'm just going to confirm the version that we're working with. So the version, if you go to the CSP file, we are now on version 1.29 point oh. Now, you can use this Edit view or you can use a Solution Explorer, which is a bit more natural. If you use Visual Studio, it feels a bit more natural with the way it's laid out. And this is available if you have installed that C sharp DevKit. Alright, so let's get into our example. So I'm going to start off with a using statement to use that semantic kernel. I'm also going to go ahead and initialize a builder, and this builder is going to be kernel dot create Builder. So remember that when we started off discussing the kernel, we said that it's like a dependency injection container. So if you've used that Net core before, this shouldn't seem too unfamiliar. Every time you open a program that sees in an A speed on a core application or at least let's say an A speed on NCR, Razor pages or MVC up that is preconfigured, you would always see some builder object that allows you to add dependencies. So this kernel Builder is allowing us to add those LLM dependencies that we need for our application. So here, I'm going to say add open AI, and then you'll see add Azure open AI and the different kinds of Azure open AI, services that I can add off the bat, right? I can also add plugins and services right here. So for this example, we are doing an AI chat completion app, and then this is going to have some it's going to have some overloads. So if we look at the overloads, it's expecting a deployment name and endpoint and different things like token, credentials, service ID, et cetera. So you have different overloads for this method. But before I fill those in, I'm just going to finish this up and say Var kernel. Is equal to, and then we say Builder, go ahead and build. So this is saying, Alright, take all of the different LLM endpoints and everything that we've added and then just create one final kernel project. Sorry, object. My bad. Now, just to speed this along so don't have to sit and watch me type. I'm going to go ahead and say console that right line into your inquiry. So this is a prompt to the user, and then I'm going to receive their inquiry, which I'm going to store in a variable called prompt. Now, once I have that input from the user, I'm going to go ahead and pass it along to the kernel. So I'm going to say result is equal to, and then we're going to await the kernel. And just so you know, kernel has several methods, asynchronous and synchronous alike. So you can create function, it can create plugin. And we're going to get into all of those things as well. We can invoke prompt, invoke acing, et cetera, invoke streaming. So without getting into details of every single method, what we're going to do here is invoke a prompt. We're using the asynchronous method. Hence the weight, and we're going to pass in that prompt variable, that value. And then we're going to go ahead and console that right line, the result. So once we've done all of this, let's backtrack now and fill in the missing information here. So let us look at the different parameters that are needed. So firstly, we need deployment name. And I'm going to kind of give I'm going to state the parameter name here so that we know what value we're putting in. So we need the deployment name. What's next? We can do the Open AI client, but I want to use the API key. So I'm looking at the next overload. And we can put in. So we can put in credentials. That's not the one I want. I can also put in that API key, right? So I can put in the string endpoint and then the API key. So let's put in endpoint there. And right now, I'm just making them blank because, of course, we're going to go and fetch the values, and then API key would be the next parameter. All right. So these are the three minimum bits of information that we need in order to allow the kernel to connect to our chat completion. So let's jump back over to the studio. So in the studio, we can first fetch the deployment name. So whatever name you gave it when you created it, that's the name that we're talking about. So back in the code, deployment name is GPT 35. I didn't change it, so it's the same name, right? Then the endpoint and the API key. Now, it would be easy to think, Oh, here's a target URI, that's endpoint, but that's not quite the endpoint. We have to jump back over to our portal and in our portal in the resource as your open AI resource, let's go down to the keys and endpoint and we'll see here the endpoint. This endpoint is actually the first part of the target URI here. It's actually up to this point, and then the rest of it is the target URI for this resource, which you can use in certain situations. In this situation, we don't need all of that. So it's easier or better to target the open AI endpoint rather than the specific deployments. Point, right? So we're going to come back, and that is our endpoint, and then we can use the API key. And in testing, I realized this key is actually the same as the key one here. At least that was my experience. But either way, it's better you use the key that is associated with the endpoint that we want. So let's go ahead and copy that one and paste it. And do note that whatever values you see here, you don't generally want to store them directly in code. You generally want to put them in an environment variable or use secrets. This is a simple example, simple demo, so I'm not going to go into all of that complication right now, but do know that whatever values you see here by the time you're doing the course, I would have expunged these from my resources. So make sure you're using your own and don't try to copy the exact values that you have here. So now we have our simple console app that will take some input and return, some result, courtesy of our GPT three, five, Turbo deployment, which we have in our open AI resource on Microsoft Azure. So let's go ahead and test this out. All right, so to your inquiry, let's say, list the best places in Jamaica to visit. All right. So I think by now, you know I'm Jamaican. So let's see where in my country would be the best place for you to come and visit. And see they list out some really nice places Montego Bay Grill Oraios. Kingston is the capital. You can always hike or blue mountains. Go to Port Antonio to relax. Trelon is rich in country. Culture, apologies and flora and fauna. Treasure Beach is great. All of these are great places. So yes, I approve this list, and thank you to our chat completion app that we have just created using our semantic kernel. 14. Section Review: Alright, so that section was short but hands on. So what did we do? We reviewed what the kernel is, which is the pivotal or the central piece in the whole concept of the semantic kernel. We looked at how we can provision and open AI resource, and then we looked at the Azure AI studio where we created a GPT model based on the 3.5 version. And then we connected to it using our semantic kernel, SDK. Encode and created our very first chat completion application using the semantic kernel. So that is really how simple it is once you have the LLM and you have all that information for the endpoint and so on, you don't have to worry about any SDKs and any complications with our uniqueness relative to the LLM that is being connected to, because the semantic kernel abstracts all of that away and gives us a simple way to connect and interact. So now that we have seen a practical example, let us go on to the next section where we'll look into plugins.