Azure Open AI Search Demo: Custom Enterprise Data for ChatGPT Setup | Train your Own Data

in our previous sessions we were exploring this open AI search yeah I think Azure open AI search so I found what was the issue last time we when we were trying this out I found what was the issue and I was able to set this up so yeah let's start from beginning this is a basic architecture diagram where we get the files from the various sources such as files databases that is fed to the Azure cognitive search and then yeah some embeddings happen connected to open AI and then stored in the server and from the application we can just query that knowledge base so this is the sample application that will be shown all right so let's get started with this one first of all we need Azure developer CLI so here you will find all the steps how to get installed for Windows Linux Mac okay next one is python Sleepless that is again requirement node.js get and if you are on Windows then Powershell 7 plus so if you have an Azure account and if you are not an administrator or owner then you can have this permission assigned to your email account also if you are not working on your local you can make use of GitHub code spaces what it does it deploys the whole solution to a remote container you can say yeah so now let's start with the running as that is the earth login basically I'm just logging into my Azure account using this command so let me open the terminal let me go to my folder and let's create a new folder and let's CD into that folder okay now we have our folder ready now let me just run this command well you can run this command without the folder as well but on your next commands you will need the folder so this login can run without that so it will open a web page on your browser asking you to log into your Azure account logged into your azure next command this will just download the source code what that is present in this particular repository on your local and also set up the configuration for the Azure resources okay so it has downloaded the template code to my this new folder and it is asking me to create a new environment name so let's it is also suggesting me the environment let's go ahead with that and if I open my folder so here you see all the files now added to this one so whatever files are present in this repository everything is here so next command will deploy the whole solution to Azure and also configure everything but before that I just need to change these document these are the custom documents that Microsoft has provided so let's replace that with our custom documents okay so let's delete them and let's download something from Google like up-to-date information so that so we are sure that it's not coming from the llm model let's dial this one let me copy first in text editor okay this seems to be fine let me copy that and paste it here okay critics have branded the UK governments delete one billion package of support for the semiconductor industry as in significant okay let's pick only this small and let me just print this and convert that to PDF and let's put them into our knowledge base data folder okay there we have our PDF let's start with the one PDF and then we can move to some different ones okay this seems to be fine now let's continue with our process that is our next command is a z d up and it is asking me to choose my subscription so make sure the subscription that you choose have access to your open Ai and yeah this was the mistake that we did last time okay before that again we are making a mistake it will not work unless I just type sudo as an admin to run this command so now let me choose the location so here it says the for the Target location the regions are currently support the model used in the sample are yeast us or South Central us and once this is done so we can choose the South Central U.S as the location now if this process will take a lot of time approximately half an hour because it is going to deploy so many resources around I think five to six of them so it has created a resource Group with this name added a app service plan form recognizer this we need to explore this formed organizer it's a very powerful service which reads the PDF documents and extracts the text out of that this is a very good service that we can utilize in one of our applications so initially what it will do it will upload my documents to storage account from Storage account it will be picked up by the form recognizer extracting text and after extracting text it will go to the Azure search and now currently Azure service search service and Azure open AI are getting creative and we can also check the status on the Azure portal but this is a resource rule and here are the resources are that are created and here you can see the progress one deploying 11 succeeded and the search service is getting deployed now this approximately takes I think eight to ten minutes cell service okay open here is also deployed after this search service it will deploy or app service that will contain all our application code where we can just chat with this chatbot okay so its services created now it will be just creating my backend app service so if I just refresh this page now it will succeeded one deploying and it would have been deploying the web service yes okay even the backend service is declined that's good now what it is deployed three more that is open here roll back in search rule okay just configuring them but yeah these are the main surveys that will be deployed and there we have some error that's good okay yeah uh I miss this one I missed to change the script since I am just targeting Python 3 not python so I need to change the script so I go to Scripts I click on pre prep docs.sh and I just change python to python 3. this is not a requirement it varies from system to system my system does not require recognize python so I just replace that with the python 3. save this one and rerun the script but this time it will not deploy the services again it will try to update them since we already have Services deployed it will just try to update them now the process of uploading the documents will start to blob storage and then from blob storage to search service here now it is just installing all the python libraries here you see processing my document uploading blog for page 0. then extracting text using form recognizer and yeah that was all I think we are deploying Services now it is deploying so deploying service backend so it has completed with the Azure open AI search form recognizer now it is deploying the front-end application along with its own backend to the Azure app service so here is the code for that application if I go to app yeah this is the one so it is deploying this to Azure app service and let's see the models again and GPT 3.5 turbo is the first model and Vince is the second one okay we will save it so once this deployment will be completed we'll be receiving tab service URL where we can test our application along with the models created I think it's about to get computed fetching endpoint perfect now we can just go to this URL and test our application okay it's still processing something in the back end it will take some time for the actual default deployment to get completed and then we will not see this app service python developer page let's refresh after a minute and then it should be working and apart from that so we can also run this application on our local machines so here are the different steps one of them is if you are using existing resources then you can set your environment with these configurations and then you can deploy it and for redeploying your local code this is the command okay so this is the command to run locally let's try to run that and see if it works but yeah even this will also work okay now it's loading so chat with your own data this is the UI Microsoft has created let's ask some questions let's go to our question Bank okay let's generate our own question that will be much better critics have branded the UK government's delayed one billion package what was the word of package that was delayed okay the UK governments delayed one billion package of support for the semiconductor industry that's good and if I have a lot of document then it will give the situations as well like from which document this is coming from this is very powerful and let's ask something else as well has just unveiled a new 10-year strategy what is the sub strategy and wait by UK government Ah that's a strong spelling but um since there is no much of the context in my data and just repeating of where that text is coming from but even that is also powerful so but it is facing allocation it is not enough the US and you have announced support closer to 50 billion how much support is advanced to give the amount and support closer to but there is no mention of exam is it just copy paste in the exact text not right but there is no mention of exact amount of support announced by UK government okay and let's see who is the prime minister [Music] foreign our regular testing question who is the fastest man alive okay that's a good answer so basically it should not answer anything which is not in my data source this is one of the chat option and there is one more I think this ask a question I guess this um this is I think this is that DaVinci model okay this was powerful and love let's see how we can just run this locally so that we can customize the application so AZT login next one is change directory to app cd2 and and run dot slash start dot PS1 I think PS1 will not run on this machine because that's Powershell command let's run the start.sh okay so what's error loading is a d dot environment file from current environment um okay um yeah I need to change that as a star dots sh so I need to replace the python with python 3. okay it's a node.js application that's along with some python maybe node.js will be on the front end and python will be on the back end see now we can run it again but installing the node modules is another error command not found x d g Dash open this is a development server do not use it in production environment okay so it's running on this port number let's copy okay that's a good one let's ask something ideally it should give me answer okay that's powerful now we can just change this code on our local machine and redeploy it using this command where is that yeah simply run a ZD up and yeah okay before that let's see how we can upload the document if so we are able to do it directly on the storage account let's see so on the storage account we go to storage browser containers content yeah so this is my file that was uploaded here so we need to explore this code so from here it will be just called from like this form recognizer will be called but yeah I think that is all in this session okay thanks everyone

Channel: Dewiride Technologies

Views: 3,090

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

Id: PpFrKXKBvRE

Length: 23min 22sec (1402 seconds)

Published: Mon May 22 2023