Learn Azure OpenAI - Chat and Code with Your Own Data!

hey welcome back to the code wolf and welcome to another video about Azure open AI in this tutorial we're going to explore how you can wire up your own data with Azure open AI so you can build chat Bots and other tools that are specialized around your own data and they're not polluted by more general purpose conversations like you might see with the standard chat GPT so if we look at the sample app that we're going to be setting up you can see there's a title to ask your own data questions but below that it says this chatbot knows about Hobbit and Super Mario but not much else so ask away so for example if I were to say what is azure and hit enter it's going to come back with an answer of the requested information is not found in the retrieve data please try another topic now of course the standard chat GPT would have all kinds of information about that but this bot is just configured to know about these two things because the data source that I set it up with has information about those so in contrast to fire to say who is Gandalf the famous Wizard and hit enter of course it'll say Gandalf is a wizard mentioned in the retrieve documents but we can do better than that we can say who is bbo baggin and it'll bring back a more thorough response so this only has some basic information about The Hobbit it doesn't have the entire book or all the extents of lore or anything like that obviously it has more data about Bilbo than Gandalf or if we say what powerups can Mario use since this also knows about Mario just one more question here you can see it'll give us pretty extensive information about this and I uploaded a decent amount of information about Mario and so we get a more uh complete response here and the data that I uploaded is also semi structured it's not like a complete essay or anything like that it's just data that I retrieved from various web pages on the internet and uploaded it into blob storage so we're going to see how to set up an Azure open AI service wire that up with a search service and a blab storage repository to retrieve data and see how we can consume those both from our own app and custom code and in the browser playground it's going to be great please hit subscribe to support the channel and let's dive in all right so let's just jump in and get started out in Azure there's two primary services that we want to create that will be dependencies for this project and those are an open AI service and a storage account to actually hold the data that we want our AI to work with or to consume so first let's go out and create our open AI service and you can get to that by just searching for open Ai and picking that from the drop down here so let's create our Azure open AI service and first just pick the resource Group that you want to use and I'll just call this uh the code wolf AI or something like that and choose the standard pricing tier whatever you're comfortable with make sure to check out the pricing details before you create anything and then I'll hit next and all this default settings here all these are fine and at the end I'll just click create and while that's going I also have another tab over here to create a storage account and this is even simpler I already have our code wolf AI Resource Group selected again and I'll just call this code wolf AI data or something like that and I'm just going to choose a standard tier uh pretty standard settings here and just go ahead and create that so we'll give that a moment to run and it looks like our Azure open AI service already finished so we can jump over there and the main thing that we're interested in here is these model deployments so remember with Azure open AI we can deploy different types of models that specialize in different tasks but those are all handled out over in this AI portal over here so let's move on out to here and give this a second to load and the first thing we'll want to do here is say create new deployment and this is where we select our model that I was just talking about and there's different options here depending on what your your subscription settings are and what you're paying for um but GPT 35 turbo is kind of a good multi-purpose model this is good for language based things like we're going to be doing and we can leave that setting as is and I'll just call this turbo wolf and click create and that's going to go ahead and deploy our model for us and let's also check on our storage account so it looks like this finished successfully so this is where we're actually going to upload Lo data that our model is going to consume so let's go out to our containers and let's create a new container here called AI data or something simple so I'll create that and that'll pop in right there so let's navigate into that container and now let's upload a few documents in here and these can be really whatever you want but I would recommend picking something that has at least a fairly significant amount of data in it you know maybe 5 to 10 pages of Text data or some type of structured or semi structure data so let's just browse for a couple files here and I have two files that we're going to work with one is uh Hobbit dxt which is the first chapter of The Hobbit so we'll see what it can gather about that story for us and the second one is actually some fairly unstructured data about the Super Mario franchise um it's just some data I scraped off of some Wikipedia articles and things it's valuable information but it is very Loosely structured so I'll just upload these quick and those will load in our container so that's all we have to do for storage we're all good to go on this front so let's move back over to our deployment and refresh this so under our deployments here's our turbo wolf and from here to set up our demo let's go over to chat this is where we have our playground where we can experiment with different models and different settings and talk to our Ai and just see kind of How It's behaving so by default this is kind of a general purpose AI in the chat here so if I were to just say hello how are you uh you can see we get some general AI assistant uh feedback here but we don't want a general purpose bot we want our specialized bot and for that we can use this add data Tab and this is a really powerful feature so if we click on this add data source it'll walk you through this workflow on how to get this going so we're going to pick blob storage but note that there's lots of other options here too so you could use a cosmos DB database you could actually just use a website directly or you could even upload files but uh those will end up in blab storage anyway so I just went ahead and created a blab storage account ahead of time so let's pick that uh AI data blab storage account now if we were to refresh this it's going to say that we don't have an Azure AI search resource so let's go ahead and create one of those as well this is easiest to do just from this AI workflow so that you know you're creating exactly what it wants so let's use the same Resource Group and I'm going to call this uh Cod wolf search or something like that and I'm going to put this in a region that's closer to me now on this pricing tier this part is actually kind of important for the search service I believe you have to be using at least standard for this to work uh I don't think the free and basic uh have the requirements to be able to use this for this AI chat that we're building now note that this does have a cost associated with it in my experimenting if you just create this and test with it for a little bit and then delete it the cost is very low it's nowhere near $250 but don't take my word for it make sure you investigate the pricing calculator down here and know what you're doing before you start spending in Azure so I'm going to click create here and let that run for a minute and that'll validate and while that's working uh we can go back to our playground here and as we refresh this eventually when this finishes um like I think it just did here now if we go back to our playground you can see that that code wolf search is now available for us to pick and now we have to give the index a name so this just creates a new index in our search service that helps with data ingestion so I'm going to call this code wolf index and for the scheduler you can really put this at whatever you want if you just do it once it'll never run again obviously but you also have the option to do hourly and daily depending on what your needs are but I'm just going to leave this at once since we're not going to be periodically updating this just for the demo and then you have to to acknowledge that some costs may occur and then let's hit next and we're going to set this to keyword search type and it'll give us kind of a nice summary of what we're doing here and then let's say save and close now over on the left here you can see there's this ingestion process so it's starting to index our data and pre-process it and kind of get this ready for use by our AI bot so this can take a little bit of time it's not too bad but I'll just pause here for a second while this finishes all right so once that finishes indexing let's first test out our setup here in the browser in this chat playground to make sure that things are working as expected so over on the right if I were to start a conversation here such as what is Google the chatbot is not going to understand what that is because it only understands Concepts that live in our data that we uploaded so if I were to switch this up and say uh what is Super Mario because remember we added a file about that in our blab storage you can see now it's going to give a more meaningful or substantial response now this is great we're able to see that everything is working but this chat playground isn't all that useful in a real world scenario when you're building your own app you're obviously not going to want to use this in this browser tool here so now let's start to explore how we would add this to a custom app like we looked at in the beginning of the video so over in Visual Studio I have that app open and this is available on GitHub in the description and you can actually use this app as if is if you just replace some key connection configurations which we'll look at in a moment now as a side note I have a separate video that goes into greater depth of how to build an app that connects to open AI programmatically so in this video we're going to just review a few key Concepts and then focus on the part that configures our API to use our own data source so if you're looking for more information just check out that other video on my channel that'll give you the answers you need so here we have a fairly traditional razor Pages project and we have our form here and this is what the user fills out to submit their question there's just a regular input component here and that binds to a property on our model called question and then when the response comes back it just displays that below the form pretty simple here I'm not going to go into how razor Pages Works in this video and this is a pretty simple flow so that code behind lives in our index. CSH html. CS file and so if we open this up you can see where we're binding that submitted question from the user from that input field and then this response content will hold what comes back from open AI as we'll see in a moment so the most important part of this setup is the on poost method and this is what fires when the user submits the form to handle their request and run some logic and there's just a few key steps to this whole setup the first step is to configure our open AI client so this is the class that will actually go out and talk to Azure open AI but to do that we need a few configuration points so we need the open AI endpoint the key and the deployment name and you can find all of those easily out in Azure so if I were to go out to Azure I have this Cod wolf AI Resource Group open so this holds all of the different Azure resources we created for this to work and so if we were to navigate down into this uh code wolf AI Service First and finding these Keys is pretty easy so if we just go into keys and endpoint you can copy the key value out of here as well as the endpoint so those will map to the key and the Endo so that takes care of two of those and then the third one one is just our deployment name which we had called turbo wolf so if we go back to our playground and we go over to our models or if we go over to our deployments then we can see our deployment name is Turbo wolf so those are the three values you need to set up your open aai client and then you pass those in when you're creating a new instance there so the next part is to configure the search service so we have to set up a couple configurations here so that the open AI client will actually use our search Service as its data source rather than just the default model data so it's a similar deal here we just have to set up a few key configuration values which are endpoint key and index so if we go back out to our Resource Group and open up our search service on the overview page we can find our URL right here so that one's pretty easy to grab and then in our keys we can grab one of these admin keys and that'll be our search key and then the final value is this search index so if we go back to our indexes remember we created this code wolf index when we filled out that workflow in the browser so all of these key values are here now we can then use those configuration values to set up our search Chat extension configuration so this is an object that holds all of these uh search configuration points and we'll pass that into our open AI client when we send out a message and that's the final step here right here so we set up these chat completion options so this kind of builds the object that will be sent to open AI to get our data back and so first we create a new chat request user message so this simulates the request from the user and we pass in the question that they submitted remember this gets bound at the top here when they submit the form so there's our message getting sent over and then we also attach those search configuration options so this code wolf config that maps to our search configuration up here and then finally we also pass in the deployment name so it knows which of those model deployments to talk to remember we also Define that up at the top here with our turbo wolf and then the final step is just to send that out to open Ai and we do that by calling this get chat completions and finally we pull the response content off of what comes back from that and that gets displayed in our page here in the browser so I already have this running and so just as a refresher if we were to ask another question such as what type of game is Mario and you see we get a nice little response here with some but if we were to ask something more General such as what is C it's going to tell us the requested information is not available now remember this is the most important part for our discussion here this is what sets the data source as our search service if I were to just remove this actually and just take out this extensions here and then if we were to restart the app let's see what this gives us now if we lose that search configuration so now if I were to say what is C ask it the same question you can see now it actually gives us a full response back just like a standard chat GPT prompt would if you want your app to only use a specific set of data sources you just have to include that configuration so I hope you enjoyed this video I have more Azure open AI content on the way so please hit subscribe to support the video check out the other two videos on open AI on my channel for more information and more detail about coding and getting set up with this service and I'll see you next time right here at the Cod wolf thanks

Channel: The Code Wolf

Views: 10,513

Keywords: Azure OpenAI, Learn Azure OpenAI, Azure OpeanAI use your own data, Azure OpenAI coding, OpenAI own data, bring your data, Azure OpenAI upload data, Azure OpeanAI Chatbot my data

Id: OdpqciXDKjY

Length: 15min 31sec (931 seconds)

Published: Sat Dec 09 2023