Azure OpenAI BYOD: ChatGPT with Your Own Data!

companies typically have a lot of data out there that could benefit from some of the chat capabilities available in AI today that's what we're going to look at in this video so if you had a bunch of PDF manuals or word docs or even HTML markdown files and others what if you could actually chat against those so that your users could ask questions against those documents and not actually have to scroll through every page or do a manual search well we can do that with Azure open Ai and bring your own data and that's what we're going to look at here now to get started I'm going to walk you through how to set up the Azure open AI resource if you want a more detailed video go check out a previous one I did and this one I'm going to go through a super quick so let's get started first thing I'm going to do is just type open AI we'll select that and then from here we'll create the resource [Music] now that that's created let's go ahead and go to it and the first thing we're going to do is go to model deployments this will take us into Azure open AI studio and I'm going to click on create new deployment and I'm going to select gpt35 turbo now I'm actually going to create two models and I'll show you why but I'm going to go with the default name it the same so let's go ahead and create that first one and now I'm going to create another one because this can be used for vector-based searching and this is a different model called Ada so we're going to select text embedding Ada we'll go with the default and then I'm going to name it the same again and we'll create it okay so now that we have our models let's go into chat now as a heads up right after you create the model you typically can't chat even though it says the models are deployed we need to wait just a little bit of time but luckily we have some other things we're going to do to demonstrate the bring your own data feature now I'm going to scroll up to the top though and you'll notice that I have add your own data preview we're going to add our own data source and we're going to select what do we want to do here do we already have a cognitive search out there do we already have blob storage out there well I don't I'm just going to upload files and what I'm going to need to do is either pick from some existing storage or create a new one I'm going to create a new one in this case [Music] [Music] so it looks like that's being created so I'm going to come on back we're going to hit refresh and we'll select that all right now you do need to turn on cores access to make this work so you just hit the button here turn on course and there we go and the next thing I need is a Azure cognitive search resource so I'm going to click create new and do the same thing [Music] so that's all being completed we'll say go to Resource and you can see we're good to go so now we'll go back again we'll hit refresh and we'll go ahead and select that now the next thing I need to do is give it an index name it is a search so therefore we have indexing of documents going on I'm just going to call this open AI index now from here I can add Vector search capabilities now earlier I added the text embedding 802 model and that's why is because this will help us with that once we've selected that I can acknowledge that yeah there's going to be some charges for this of course and hit next now the next thing we're going to do is upload our documents so I have two documents one is how to install a clock I literally used AI to generate it so it's AI using AI some Inception going on and then I have a company FAQ document and this will answer questions that customer service reps may have so let me go ahead and upload those all right those have been uploaded and we can move next now the next thing is what type of search do we want to do keyword or semantic now if we select semantic I can acknowledge here hit next and then we can save and close now you might wonder though what is semantic versus keyword well keyword is what you would expect if I type clock it's going to search specifically for parts of the document that have clock whereas if I type time well time is semantically related to clock therefore it may find some references kind of like dog and cat yeah they're different but they're both animals they're semantically equivalent if you look at it from an animal perspective so I'm going to go ahead and go with semantic in this case and acknowledge and then we'll hit next and then we'll save and close here now what's going to happen is this is now going to index our two documents and that'll take just a bit to go through but once that's done we should be able to come down to our chat and ask questions against those documents okay so it looks like our indexing process is done in fact you can select this to go get some information about your index but normally based on timing we can now come down to the chat and start asking questions like how do I install a clock let's go ahead and run this now you'll notice we get an error notice the message up here says semantic search is not enabled for this service so what we can do is go back to the portal let's just do that in a separate tab I'm going to go into my resource groups mine is sandbox and let's just find this in here and there it is now you notice we have semantic search I'm going to go ahead and select this free plan for here and there we go we now have selected it so if you get that be aware it might take a little bit of time for it to catch up when we get back to the chat but that's what you can do if you get that type of air now I'm going to come back in well let's clear the chat out and we'll do the same thing again and depending on timing we might need to wait a little bit so be aware of that okay and this is what you should get so this is actually pulling information from my document all this was in the actual word doc or PDF that I uploaded and you'll see down here here's a reference to that document actually and we could open it from here but this gives a nice summary and I don't actually have to go scanning through that document myself now let's go ahead and clear this and let's do how do I process a refund this is in the other document which is the FAQ for customer service and there we go if a customer requests a refund or wishes to return a product blah blah blah notice here we have the company FAQ stock that we uploaded now this is all great but at this point you're probably like okay cool I can't let my users just run to the Azure open AI Studio obviously well you'll notice a couple things first off right up top we can actually create a new web app if I come on down you'll also notice view code here we could programmatically call this API and so when they type something in a custom chat box we could run some code whether it's they have Python c-sharp and then just curl or Json here let's say you're doing a net app well this will give you the starting code to actually call this API send up that chat and then get back the response but if you don't even want to do that and you just want to create your own app we can say deploy to a new web app and then you can come in and create that web app so let's go ahead and do that [Music] looks like it's done so we can launch the web app you can log in using the same account for the portal and there we go we now have an actual chat app that we can embed or use in some other app if we wanted let's try it out how do I install a clock and we get the same exact result so with very little work you can actually get this going get your documents out there and allow your users to chat against those documents really really powerful and this is just going to get better and better over time of course so thanks for tuning in and check out the other videos in my channel

Channel: Dan Wahlin

Views: 15,401

Keywords:

Id: u0AUwOKxUsg

Length: 9min 7sec (547 seconds)

Published: Tue Sep 26 2023