Train your Own Enterprise Data with Azure OpenAI Service | ChatGPT with Custom Data - PDF, Word, TXT

hello everyone in this video we are going to cover how we can train custom data using Azure open AI and also connect it through a custom website so there is a recently Microsoft interviews a new feature in Azure open AI Services where we can train our own custom Enterprise data and we can interact with that data and also we can also deploy that to a web website so we'll be covering all of that so before getting started let's create a new Resource Group once you have created the resource Group let's create the open AI resource that is azure open AI and here we are going to name our resource and also choose the pricing Tire make sure you choose the reason where it is applicable to you and next and review and create once our Azure open AI resource is created we can launch the open AI Studio that is oai Dot azure.com so once you have created the resource you can select that resource from here okay currently it is asking me to create a resource because my store my resource is not completely created along with this resource we also need a storage account and cognitive search so instead of wasting time by looking at the process let me create it before it hand itself okay and also uh there is a prerequisite that the cognitive search resource that you create should not be a free one so we'll be going with the basic Tire and create and also let me refresh this one because by this time our Azure open AI resource must be created yes okay let's create the cognitive search and yep choose your subscription and also choose the resource that you created okay so you have now go to chat and I'm going to create a new deployment for GPT 3.5 Turbo and I'm going to give the same name okay so once you have deployed the model you can see the deployments and you can see the model that you deployed and let's go back to the chat and this is a new feature that added by Microsoft that is add your data that is still in preview I can click on add data source and here I have three options as your cognitive search blob storage or upload your files so basically upload your files also uses blob storage so I am going to start with the Azure blob storage and also let me create a storage account okay and review and create all right um also now let me look for some files so that I can upload okay so I have a doc basically a Microsoft Word document with a Blog in it so this particular blog I'll be training basically uploading this to the storage account in Blob container so my storage account is ready let's go to storage browser blob containers add new container and let's give the name as data okay I can open that click on upload and let's upload this file okay so once this is done okay I need to wait for the cognitive search to finish but by the time let me show you how this is going to be look like okay so here we choose the blob storage I'm going to choose the Enterprise one that I just created now and I must choose the storage container and also cognitive search I guess it is completed okay still in progress it should throw me an error that resource not found something similar to that so we have our cognitive search results created now let's go to open AI Studio refresh the page add a data source select your blob storage container then select your cognitive search results and give a name for the index let's say [Music] um blog next save and close so it is it will take some time for to drain that data by the time this is happening I would like to deploy this to a new web application okay so what does it say your web app will be configured with Azure active directory okay so what it is going to do it will deploy by web application and enable the authentication on that for my tenant only if you want to share it with others apart from your own tenant then yeah you can update that authentication configuration um to include that common in place of your tenant ID but yeah let me first deploy this okay it's already there let's try this one I'm assuming that it is waiting for the deployment that is this particular deployment to get complete and then it is going to start yeah please wait for web app to finish deploying okay okay yeah so it is I think creating the web application and then it will start deploying my model let's see okay so my chat is completed basically it has imported my data that I have added to storage account blob container it has created an index blog in my search resource and it must have created some vectors in that but why this is not working okay it's fine uh we will come back to that later okay it is doing something in the back end but yeah let's not waste there okay let's start hi and also let me open that document to ask some questions from it creating strong and unique passwords so what is this [Music] um best practices for everyday cyber security that's us what are the best practices for cyber security it's not replying to my previous response a system message here you can modify it's fine let me reload the page okay there it is let's say hi to it and there we have our first response yeah okay let's ask this one there are many best practices including creating strong and unique passwords implementing multi-factor authentication that was the first yes it is going uh topic by topic keeping software and device updated yes being very efficient attempt perfect so it has given me all the headings main headings and let's ask something inside that [Music] um what is practicing data encryption okay so it is giving me something data encryption is the process of converting plain text into coded message to predict sensitive information okay let's see you secure messaging app consider using file encryption so basically it has summarized it and this is from where it is getting me the response okay that's pretty cool yeah and also you can limit responses to your data content so that it does not create the content apart from your own data so let's ask something which we always ask it was a fastest man alive so that we are sure that it is not replying from its own data yes there it is I'm sorry but I don't have any information related to the fastest man alive in my retrieved documents that's pretty cool yeah okay so that is all here and you can now deploy this to a web application [Music] okay now it is deploying my web application also if let's wait for that to cut completed also if you change some system messages so it will try to restart the conversation from the beginning so that your system message will get into effect into this chat session basically it will restart your chart session you can also upload a multiple documents in the combination of any type word PDF text all of them will be accepted here in our previous videos we had gone through this one that is azure open AI search demo here so this is our video that we created last time where we explored this particular sample how to set that up there are many limitations here where we cannot update the existing document basically if you try to update it will insert as a new one and also you cannot upload document that is of different data type that is different extension that is Word document or text file you can only upload a PDF document so those limitation has been handled over here and this will do this does a very pretty good job and also yeah also it created a very fine user interface to try it out and even this particular functionality also create a very good user interface not much of the functionality compared to this one but yeah it's a basic user interface that it will create okay so our web application has been deployed and let's try it out also let me copy some of the questions that I have asked so that I confirmed that this is the one that is coming from my own data yeah it might be restarting the Azure app service so it will take some time for this to load up let me show you what all resources that are created as part of this demo one is the app service plan that runs our app service that we just deployed now next is the Azure open Ai and the cognitive search and last one is the storage account where you keep your data and I have also seen that it also does incremental upload in one of the blogs that while introducing this functionality um there is what's new yeah this one this is a documentation and the new feature for the demo that we did today and somewhere it is written that it is incremental yeah these are the different file types that are supported text markdown HTML Word files PowerPoint and PDF and your maximum response semantic search index field interacting with the model using the API here so when this web application is deployed right since it is so once we test this web application there is a API call that is being made and here we can do the same API calls from our custom application to interact with this custom data so you did not rely on this particular web application you can call the API directly from your own custom application so these are the details how you can use the API and somewhere it is incremental let me search that yeah here it is you can send a streaming request using stream parameter allowing data to be sent received incrementally without waiting for the entire API response this can improve performance and user experience especially for large or dynamic data okay oh no sorry this incrementally is for the response not for the data upload maybe I misunderstood that but yeah it should be handled okay there we have it so since it is connected to so it has enabled the authentication on the web app you can disable that by going through your web application let me show you that this is your web app you can go to Authentication and you can remove that or if you want to support multi-tenant then first you need to go to the Azure active directory app registration change that to single tenant to multi-tenant and come back here and modify this one that is instead of this tenant ID just modify it as common and it will turn into a multi-tenant application okay all right so this is a chat interface that we have and let's say hi okay let's ask our first question and here you see that is what is uh it was mentioning about incremental it is live streaming yeah so that's a good feature if you have large data so that user did not wait for it so you can have this stream on and also let me show you that API and let's ask something else so best practices of MFA and this is what it is being done that is conversation and that's the payload so it also remembers what you have typed so it maintains the user State as well okay so that is all from my side thanks everyone

Channel: Dewiride Technologies

Views: 5,598

Keywords: Azure, Open AI, Search Demo, ChatGPT, enterprise data, training, chatbot, AI, machine learning, NLP, natural language processing, artificial intelligence, tutorial, setup, custom knowledge, cloud computing, Azure services, Microsoft, azure search, openai, bot, train own data chatgpt, chatgpt train custom data, use own data in chatgpt, enterprise data in chatgpt, cognitive search, vector db, train custom data, chatgpt train own data, azure open ai train data, open ai

Id: -sr44ZldZoI

Length: 21min 28sec (1288 seconds)

Published: Thu Jun 22 2023