Complete Generative AI With Azure Cloud Open AI Services Crash Course

hello guys so here is an amazing crash course on developing generative AI application using Azure open services so AZ open Services was one of the most requested video by all of you and I feel that this crash course will definitely help you learn a lot of things out there uh this entire crash course has been developed by S Savita uh who is an amazing Mentor uh I will be providing his YouTube channel in the description of this particular video along with the entire code materials about his GitHub everything will be available in the description so please make sure that you watch this video till the end and definitely make sure to implement each and everything because this crash course is going to be for 3 hours and along with that you will be able to find out the entire timestamps in the description of this particular video so yes go ahead and I've been just going to keep a target of likes for this particular video let's keep the target to 2,000 and I hope uh you will be able to achieve it as soon as possible yes go ahead and enjoy the scratch course thank you hey hello everyone welcome to this azir openi crash course in this azer open crash course we'll try to understand each and every aspect of the azer open and we'll try to understand it from basic to Advanced so guys first of all let me introduce myself my name is s Savita and I will be your host for this entire course I'm having around four year of experience with data science machine learning deep learning mlops and and from past one year I working with the generative AI so guys let's start so uh first of all let me show you the pp uh which I created for all of you uh just to just to introduce all the modules and all uh whatever we are going to discuss in this particular course so in the first module we'll discuss about the introduction of the Azure open AI that uh we'll try to discuss the differences between Azure open a versus open a then we'll see the how you can access the a account how you can create it and uh we'll see the subscription method and all so we uh this Basics first I will start this Basics only and then after this one I will come to this uh uh like little Advanced part where we'll see then how we can access the aure open a okay uh apart from this one resource creation model deployment understanding uh token per minute what is this limit and this uh request per minute and then we'll talk about the chat playground and and will discuss the different different model how you can access these different different model like GPD 3.54 Delhi whisper and all then we'll understand how you can use the python SDK for uh like for using this uh chat completion API we'll try to access the various model using this chat completion API we'll write the code inside our local system and finally we are going to create our custom web app and we'll deploy it also now after this one after this Aero I will come to the function calling using Azure open we'll use it using the python SDK only we'll try to write a code inside the local and we'll try to call any third party API then after we'll discuss about the finetuning concept of the azero we'll select the appropriate model for the fine tuning we'll prepare the data for it we'll talk about the complete cost and all and we'll see how we can find you the model inside the azer open AI now at the and we discuss about this a your AI studio so guys this is very important thing a your AI studio and uh first we'll see how you can create a your AI Hub and the resources we'll try to uh deploy the various model over here and then I will create our custom rag application so before this one uh I was talking about the fine tuning now after the fine tuning I'll be discussing about the rag so we'll create our own a complete pipeline using this prf flow it provide one uh specific it provide one specific uh service that is called prom flow with that actually we can create our own custom application we can deploy it we can generate the endpoint URL and we can connect with our own custom application also so yes at the end we we will create our own chatbot using flask and we'll try to consume the rag model from where from the Azure AI studio so I hope this syllabus is clear to all of you now uh let's do one thing let's start up from the Azure itself so guys once you will search over the Google let me search in front of you so if you will go and search aor so here I can write aor AI Services okay aor AI services so you will get all the services which is being provided by the Azo itself now uh simply you can sign in if you have account uh I will be coming to the account first of all let me show you all the services so guys uh these are the services or here I can search over the portal itself was AO AI Services okay so uh see guys this is all the services which is being provided by the Azure AI so first is openi itself then AI search then computer vision face API then spe service custom Vision language service translation document intelligence is one of them board service anom detective there are various services and here directly you can use the service and uh you can create your own custom application you can configure it in your own custom way and then you can utilize it guys okay now a your AI Services is fine so here first basically we are going to be start this a your open account itself and then I will come to the AI studio right this two thing this both thing is different this AZ your open AI account uh and this a your AI Studio okay so uh apart from this one we have this document intelligence also where we can upload the document and and for that we'll have to configure the database and all yes for these Services also like a your openi studio and a your AI Studio we'll have to configure the database and that service is also very very easy no need to do anything it's a very uh like simple simple thing and here itself uh in this particular tutorial I will show you now first of all guys what you need to do so see you need to sign up right if you don't have any account so here I already log in with my account account so if you don't have any account guys first of all you need to sign up to the Azure portal now let me write over here Azure portal so if you will write this as your portal guys you will get it this one so first of all sign up over here uh you can sign up with your Gmail ID your outlook or with your third party account as well if it is giving you the permission then after the sign up so once you will sign up uh then definitely you can login over here and then after the login you will find out this particular page okay this is the homepage of the azard portal now there are so many option guys so many services and all so first of all what you have to do you have to add the subscription now once you will click on this subscription guys see this is here are they are giving the option of the subscription just click over here so I already added the subscription see uh here this subscription plays a very important role uh definitely it will provide you the $ 300 free dollar credit but uh you cannot use it for most of the services it is uh like those credit you can only use for some services like for creating a basic B app and all okay free tire B app that's it nothing uh like else so if you want to use this AZ your open a or other storage account or aure Ai Se or different other services related to the AI definitely you will have to add the subscription yes guys this is very much important so uh first of all you have to create your subscription you can simply create click on this ad you can write your subscription name and uh yes you can add your uh you can add your card or you can add your like uh you can you can add your debit card credit card whatever it is asking and finally you can start using it so if you will see guys here it is providing you this a free credit but again uh 55 Services it is providing to you so you can go and check out but uh believe me guys this openi service you cannot exist inside this free tile you will have to you will have to add this pay as you go uh over here right now I think this subscription and account is fine to all of you and I believe you can do it you can create it now coming to the next point so uh here what we have guys see here we have this azer open AI now I will search it inside the search box definitely I will be getting it so let me search over here this azer open see I'm getting it so uh if you are uh if you are searching it first time the uh definition you will you won't get a option for creating it so first you will have to uh like make a request okay so first you will have to make a request and then only you can access this as your openi service uh if you are not making a request guys you won't be able to uh access this service and this request approval actually takes around 24hour time so within one business day they will approve your request and then you can utilize this particular service I hope this thing is clear to all of you now coming to the next point that what is the difference between this Azure open Ai and open AI means a your open API and open a API okay so first of all let me show you this open AI API so if I'm searching over here open AI API now uh you will get it uh so let me open it first of all explore the API and start building so this is what this is my h this is my openi API right so uh yes uh I think we all have used this openi API and even uh on this channel itself you'll find out many video related to this openai API in all and uh definitely you can generate the API key you can access the various model we have a various model basically here inside this uh aure inside this openi itself like uh GPT GPT GPT 3.5 gp4 whisper Delhi various model right now guys uh first of all we have to see the differences right because until unless we are not going to understand the differences there will be a confusion okay so what I did I created uh like one like here I written the differences for all of you I think it is visible now so a your openi versus openi so uh just just try to read over here what uh what I have mentioned so a your open actually it is providing you the managed infrastructure okay where it is providing you all the security right all the security and here you can see AZ openi offers private networking reasonal availability and responsible AI flood TR okay now uh on the other hand we have this open AI so if you will uh see inside the open so it is like I have written over here open itself is not a secure okay so here this Azure open like whatever Services it is providing to you uh like whatever security Feature alerting Feature metrix and all I will come to that I will explain you in a detail I have kep couple of more images okay so yes those feature is there inside the a open but if you're going to use open only open you won't get those feature definitely you can access the model whatever model they are providing to providing to you and you can uh use those model okay inside your infrastructure so whatever infrastructure you have created with whatever security compliances and all so you can use over there there but you won't be like get the so here actually it won't be a manage service you are doing it by yourself it's a custom one right which you are doing from your end and there you are consuming this open AI API but on the other hand if you if you're going to see this Azure open API it is providing you the complete manage service means it is providing the model along with that security different different matrixes alert alarm and all and here you can see the last one is what responsible AI content filtering so responsible AI content filtering means what so uh like no other harmful activity which uh like if you are performing any harmful activity uh with respect to that model so they will block you okay and yes uh they will take this action immediately so that like you cannot perform any unusual activity now here uh the second one is what functionality such as uh raising an alert whenever suspicious activity for instance changing of a primary key are supported so it is providing you it is providing you multiple key okay uh actually it provide you two key two API key so if let's say one key is going to be exposed so immediately it will rotate the key okay so it is having that a particular mechanism so automatically it will be done now other other other than this azeria you will find out this functionality like there is not supported so it is highly likely that data T happen means uh this open a actually it is not providing you that type of functionality it is just providing the simple uh key and uh not it is not providing you the other of functionality okay it is not providing the other functionality related to the security and all now coming to the security part so in the security part see here I have mention so first one is what multifactor authentication the second is what a data encryption at rest right the third one is what rbsc role based access management you can create a specific role for the specific Services as your keyboard for for keeping safe your keys and all okay Keys uh and whatever like your important information is there if you want to keep it somewhere uh you can keep it over here okay now the second next one is what service end point so here private endpoint public endpoint both endpoint actually you can create and you can expose it according to your requirement I hope this thing is getting clear what is the primary difference between this Azure open and open AI now model wise so it is provide the GPD engine GPD 3.5 GPD 3.5 turbo GPD 4 gp4 o also then Tex embedding right and then Deli engine then chat playground evaluation and matrixes so in short uh I can conclude it I can say here Azure open AI it's a completely managed service from the Microsoft this Azure open AI is nothing it's a collaboration of azure and open AI getting my point now this Azure open I what uh like it is providing to you it's is providing the many security features security alarms and all everything so I hope uh if someone is going to ask you this thing in an interview definitely you will be able to clarify it now coming to the next Point here so see so here if you want to uh access any model inside this a your open AI so first of all uh you will have to create it so once you will click on this create just do it just click on this create and here uh you will have to select the meth so that is what actually I was telling you in the free services uh like you cannot access this as your open AI you want to read more in more detail about the free services I think I already shown you about it you can go and check out now first select the subscription so this is the subscription which I have created for me then you will have to create the resource Group so as of now I don't have any Resource Group so let me create the resource Group over here so here it's going to be crash H now uh let me click on this uh okay and then here is a reason guys you can select the reason according to your requirement I will come to this reason and pricing also because it plays a very important role I will come to the documentation of the AIO AI Services there they have given this uh there they have given the documentation related to each and every service all right now uh let me write the name over here so name basically uh here I can write crash for okay so crash course models so this is what this is my name you can write any sort of a name you can write uh you can give any name over here then pricing tier so automatically it will give it to you what according to your subscription so I took the standard subscription uh they are having premium or like Enterprise level subscription also definitely can check out now off view full pricing detail I will come to the pricing detail guys now after filling all this information what you need to do you need to click on this next after clicking on this next here U keep it by default then again next and here no need to do anything then uh here if you want to check the resources you can check so first is aure eii service second is Resource Group so this two resources it is uh like showing you over here now coming to the next again and then uh simply you it is first it is going to be validate now once the validation will be done you can create it over here so let's wait for 1 minute okay now it is done see guys so just simply click on click on this create and here you can see your deployment is initializing so you can see your deployment over here right hand side now uh until it is getting deployed let me uh show you few more things related to this as your open AI so guys here uh see this is what uh this the portal let me close it it is not required as of now or simply search about this a your AI service document okay so once you will search it guys here you will get this very first link just click over here I'll just open it let me close this this one so see you will get the uh page okay you will get this page where you will get get the documentation related to each and every service so here is aerop speech language translator Vision custom Vision phase content safety video indexing uh video indexer Aur I search whatever you want whatever you want they are providing you the a documentation related to each and every service now what I have to do here guys see I have to open this documentation as your open AI now uh this is the documentation and here just click on this model so this is all the like details and all related to this AER openi but uh later on I will come to this Kota limit deployment types and all everything so first let me show you this model now once I will click on this model guys just scroll down here and you will get one table right so inside that table you will get you will see that which model is accessible in which reason okay now uh here I have selected uh East us this reason now inside this East us what all model is accessible just just see over here this GPT 4613 is not there 0613 means what it's a date it is representing a date okay 13 June so this is the date and here is a month got it now this gp4 is not accessible gp4 1106 this is also not accessible gp4 0125 this is accessible then gp4 Vision this is not accessible gp4 turbo this is also not accessible and here you can see the list of the model OKAY related to what related to the tell me related to the reason so you can easy easily check which model is accessible in which reason now apart from this one guys here if you will go and check with the pricing so we will get the complete detail of the pricing as well according to the according to what according to the reason so in which reason which model is giving you how much uh cost so if we're going to talk about this GPD 3.5 turbo 0.25 so this is the pricing of it in like based on the input tokens so this input token actually it is based on the R so 1,000 token uh sorry 1,000 token per minute actually you can uh process okay within this price 0.5 so likewise you can calculate other than this one I will show you the cost calculator as well later on I will come to that once I will explore the a AI studio so there I have involved the cost calculator also for the complete cost analysis got it because that is very much important now uh the next one is what Legacy model so these are the Legacy model different different Legacy model assistant API code interpreter and all now base model B Bas Lain and here you will find out that fine tuning is not available in this reason yes guys so if you want to do fine tuning model you cannot do in this particular reason definitely I will have to change the reason so let's try to check with some other reason so here what I'm doing so here I'm going to be click on this South Central us let's see whether it is available over here or not so here guys see it is not available in this this particular reason also north central reason yes so inside this reason the spine tuning is available now you can simply check the price and here guys you will find out the price is very very costly of the fine tuning yes so weage $34 training per computer hour and uh here you can see the hosting price input uses per th000 token so this is the price for the per th000 token okay then output use is per th000 token here is a price now just see this price $34 $68 for what for this dpd 3.5 it is a pretty much High guys because see whenever we are passing the data to the model okay so it takes around around okay minimum if you are passing minimum 5200 rows on now so it takes around like 30 35 to 40 minute I I already seen I already checked and in that in between that only you will get around $60 so this fine tuning actually it is very very costly right so other then this one I will show you the rag like how you can create the rag uh basically and rag is also like not a cheaper one it is also very costly because uh with the rag this database cost is associated so we'll come to the rag and we'll try to explore the rag as well now I hope guys you got the overview definitely you can check out and according to your requirement you can select the model and you can calculate the price now one more thing you'll find out over here if I'm going back so here you'll find out the deployment types so first uh let me click over here on this deployment type so uh the basic one is the standard one okay now the next one is what provision and the deploy and the global one okay here is what here's a global global deployment so generally actually we going to use this standard only because my pricing the subscription which I took it's a standard one if I want to take this provision so specifically I will have to make a request for this provision now I will show you what all model is there inside the provision and what all model is there inside the standard and what a model actually you'll find out inside the global reason everything I'll be showing you where over the portal itself now I hope your all the basics is clear coming to the next part so once you will open your azer so here see it is ready simply click on this go to the resources once you will click on this go to the resources guys here it is giving you the different different option right no need to go through with each and every option as of now so here uh let me show you which is the like uh necessary for you so first is this key and uh Endo so this I was telling to you that it is providing you two keys right so yes uh here is the first key and this is the second key you can utilize any key so for accessing your model so this is called primary key and here is what here's a like secondary key now you can read over here these keys are used to access your as your openi service do not share your key right each and every detail that like sharing with you we also recommend regenerating the key regularly you you can regenerate the key also over here and one only one key is necessary to make API call only one key is necessary no need to use both key over here then here's a reason so East us we are working in this reason and this is what this is the endpoint URL okay so this thing is clear to all of you now coming to the next one so guys here if you will click on this model deployment see let me click on this model deployment and here see there's a option manage deployment now they are telling to you model deployment feature have moved to the Azure openi Studio okay so uh see here you won't be able to find out this uh model deployment once you will click on this manage deployment so you will get one studio and that studio is called a your open a studio it's not Azure Studio that is different one that's a different one different service aurei Studio it is azure openi studio so let it open and then I will show you how you can access the various model from here and how you can deploy it and how you can create your own custom chat engine if you are enjoying the session guys please hit the like button please mention inside the comment section uh if you have any concern any doubt or whatever in your mind guys so see my a your openi studio is ready and see there are different different option now let me uh zoom in and see left hand side you'll find out various option now uh first guys uh what I can do uh let me click on this model so what all model is there so if you will check in the model in the in this model section find out the various model let it uh load okay so these are the model guys which is available in this specific reason the reason is what the reason is the reason is esus2 so see the model version created at which day okay now here stat is succeed and deployment yes it is available for what for the deployment okay so uh now guys apart from this model one more thing you need to check so guys here is a deployment so I haven't deploy any model over here and if I didn't deploy it guys definitely I cannot use it so uh for the deployment simply I can uh create a new deployment from here and I can deploy okay that is completely fine now EX in subscription key or wrong in point make sure okay not an issue let close it don't know why it is giving to me now coming to the next part here Chad so once I will click on this Chad so this is what this is my playground now here you will find out two thing first is what first is chat and second is completion so definitely I will give you the differences between this chat and this completion what is the difference in between now uh here one more is there Delhi is there right Delhi I think we all know about Delhi okay so here see it is telling to me this deployment is successfully as soon as I opened it and is like uh telling to me your deployment is done if you will go and check inside the model guys you will find find out this uh sorry inside the deployment you'll find out what you'll find out this Delhi okay because whatever model I want to use so first uh I will have to deploy that model then only I can use those model so automatically it got deployed uh and it will be uh it will happen for you as well once you will click over here it will deploy now I'm coming to the chat completion API now inside the chat completion API see uh here it is telling to me you don't have any deployment see Delhi is just for this one for this a particular one but this Delhi is not for of this one we cannot create an image from here and similar to this one we have this completion API so if you remember guys uh I don't know you know or you don't know see uh this open AI if you will go and check with the open a so there they have two type of API the first is called uh chat completion API and the second is called completion API so completion API actually they were using they will they were using inside the previous version of the open AI I think in 0.26 maybe 0.26 or 29 the open a module you can simply check now this chat completion API actually it's a latest one okay and like still in the latest version they are using this chat completion API and I will be showing you the code as well by seeing the code you can identify it if you want to know more about this chat completion and completion API so it's see don't go with the completion API always go with the chat completion API but still you want to know simply you can search that completion API versus completion API so simply you will get how many uh like platform and all and even uh they are explaining you over the uh like openi developer Forum as well see here is a complete detail so now this chat completion API actually it's a latest one and we're going to use it even over the St overflow also you will find it out so simply you can click over here and you can read about it complete details and all okay I hope this thing is clear now coming to the deployment so guys uh before deployment I would like to clear one more thing over here so that is nothing that's a quota right so Kota or the type of the deployment anything you can say now just click over here and here once you will click so you will find out we have a different different type so different different type means what so different different type means standard provision okay global standard other so different different type right so now see whatever model is there inside this standard okay inside this particular segment I can access all these model but whatever model is there inside the provision for this one I will have to make a specific request okay or I will have to make my subscription okay so I will have to take this pay as you as you go plus one the premium one then only I can access or specifically I will have to make a request for increasing my quota right and yes definitely charges and all will be associated with this provision if you are going to purchase this subscription if you have a standard subscription then you can access these many model this particular model now let me show you how you can do it so guys once you will go in the deployment okay now just click on this create new deployment from here you can select your model so let's say I'm going to be select this uh GPD 3.5 turbo so yes this GPD 3.5 turbo is available for me inside this ex standard Cota now this is telling to me because this fine tuning is not available inside this particular reason and because of that it is giving me this particular warning no not an issue I can close it I can ignore it now uh see this deployment type is what standard now other than this one if I'm going to select any other model let's say if I'm going to select this dpd 332k now see it is uh coming to this provision manage and here if you will go if you will see this uh warning so it is telling to me requesting for more Kota just the existing Kota means I will have to increase the Kota specifically I will have to make a request for this particular model so once I will click over here I'll will be getting it so I'll be getting the complete form okay so yes you can fill out that form and then you can raise your quota see once you will go inside the provision and which model guys this one now just click over here this uh to this one so let me show you just wait I think I can hide to myself okay this one guys see here is the option request Kota so once you will click over here on this request Kota so it will contact your Microsoft account team to request provision through P Kota okay fine so I think I will have to take a advanced subscription for accessing this particular model I believe this thing is getting clear to all of you now uh for some model actually you can access from here itself so let's say if I want to access this particular model gp3 gp4 turbo uh let's uh see okay no for this one also I think I will have to make my uh I will have to like take a advanced subscription got it so uh whatever thing is there inside the standard I'm going to use only those particular model I believe this thing is clear to all of you now let's make a deployment of few models and let's start to use it for our chatting purpose or for the content generation and all so click on the deployment and here we can easily like do our work with our GPD 3.5 or Turbo as well and here in this particular regon see in the standard region this GPD 4 is also available so yes we can use this gbd4 for what for the images for the image related thing now uh let me select first GP 3.5 turbo and here I can write the deployment name so this is going to my chat engine and here guys see this is a token limit token per minute uh rate limit so what's the meaning of it token per minute rate limit so in 1 minute actually how many tokens you can proceed okay so you can increase it according to your requirement token means what the number of words uh so inside one sentence like how many tokens is there how many words is there it is asking about to this one right this second is one so corresponding to this limit see if you are going to increase the limit so this limit is also increasing so this limit actually RPM it is showing to you request per minute okay so in uh 1 minute how many request you are making right to your API so I hope this two thing is clear it is pretty much important now I'm keeping it low because I for this practice I not required this much of like RPM and this TPM I can simply keep it 30,000 only that's it now what I can do I can simply click on this create and see my deployment is going on so once I will deploy I can access the model inside this chat section see now apart from this one guys uh we can take one more model so the model is what gp4 so which model guys gp4 Omni I'm going to be select gp4 Omni because it is available inside this ex standard price so just click over here and then you can write the name of the deployment this is what your this is your Omni model then uh you can select the PPM and RPM and then finally you can create so see your model is getting created and now it's successfully deployed so what you can do you can go inside this chat and now you won't able to find out that a particular issue so see this is what this is my complete chat playground this is my complete chat completion API so uh if you want to select the model simply you can uh click on this drop down I think it is visible let me uh let me switch off my video this one guys chat engine in the deployment itself you can see this one so once you will click over here you'll find out this chat engine and this Omni model this Delhi it is only for this particular playground okay it is not for this chat Playground now uh what I can do I can use this particular model so let's say I want to use this chat uh engine simply I can write my message hi and what I once I will hit enter guys so here uh see it will be generating one error and this is not for me this is for the everyone so whoever uh basically it is uh using this particular model xess dened to invalid subscription key or wrong API end point make sure you provide a valid key for an active subscription and current API okay I think I was getting this uh 5 minute error because within 5 minute actually they are going to be active this deployment let me check with other message hello access deny due to invalid subscription key or wrong API endpoint is it wrong guys okay just give me a second let me check it so here guys you can see my uh chat is working now my model is working what was the issue so here I just restarted my payment okay my subscription basically whatever subscription I have added and it is working I think there was some issue with this uh subscription method I just restarted it now you can see it is active and if I'm going to make a chat here I'm able to do so let let me show you from scratch only so this is what this is my model let's see this is what uh this this three model I have deployed uh let me open this uh chat engine first of all I'm clicking on this chat engine now click on this open in playground so once you will click on this opening playground and here uh it is opening inside the playground itself see directly now I can ask any questions so I can say here uh hi tell me about the uh Indian Capital now see what I'll be getting so I'm getting about the Indian capital I can regulate my response as well so first of all let me give you this entire uh like overview of this entire interface so what it is all about now U see here you will find out this clear chat option just click on this clear chat and it will be gone now a playground setting if you want to set any uh setting over here anything right you can do let's say language subscription and all and there are so many setting you can manage according to your requirement I will show you the uh like speech also like how you can convert from piece to text like I will show you the use of the whisper model and that is in my pipeline only now let me close it and then view code so here you will find out the code uh the entire code guys so by using this particular code you can connect right you can call this API using the python python SDK I will show you this one uh like right after this configuration and all so first of all let me uh show you a few more thing and then I will come to this completion API chat completion API then uh apart from this one left hand side you can Define your prompt and all so see here left hand side what you can do if you have any sort of a template promp template uh you can select from here okay otherwise uh what you can do so like don't select any sort of a promp prompt template here simply you can Define your uh prompt okay this is nothing this is a behavior of your system system behavior and then uh here if you will go and check this code inside the Json so you will find out the complete detail who is replying who is giving the answers and all everything the role is system and here content is what so content is what content is nothing it's it is representing the prompt now let's say if I'm going to ask say something hi hello how are you so you will find out that who is giving this response so user is asking this thing user this is the question of the user and here uh this is the answer from the system so system is my assistant means my main model and here this system actually it is representing to the overall behavior of what over behavior of the model I hope this thing is clear to all of you now see here is what add your data I will come to this one also how you can create your custom application how you can add your data and then finally we'll deploy it from here itself now one more thing so just click over here uh and this one past message include how many message you want to include it is related to the history let me show you uh regarding which I'm talking about so see guys this one pass message include now see the option over here for what for the uh like messages how how many message you want to include then other than this one here is a parameter so the different different parameter Max response how many you how many one actually you want to get a response this is the maximum one temperature means what the creativity of the model here top P you can read about this top P actually it is also related to the creativity probability and all related to the generating word stop sequence if you have any specific sequence where you want to stop you can make you can write over here frequency penalty what should be the word what should be the fre quency of the word that also you can decide present penalty what you want don't want to present present right okay that specific thing also you can mention over here so many things you can regulate it is a complete uh API okay chat completion API now uh this is clear now coming to the completion API guys so this is a previous version this uh completion API is nothing uh actually this is what this is the previous version old version how we can check out let me show you so let's say if I'm writing High here and it will generate the answer no no doubt no issue the completion orientation the specified model Delhi okay so it is saying that first please change the model uh where is the model guys this is the option for the model uh chat engine is fine great now finally I can check so let's see whether it is generating or not so yes it is generating now I think see it is generating related to what what Rel to the code uh I think I will have to define the chat context from here like let's say I'm going to be Define this question answering so here what I can do I can ask any sort of a question and according to that is going to generate answer how many NE are is there let me generate it let's see what I'll be getting it's it it is giving me a predefined template okay now I can ask my uh my question basically so what is your name let's say so if I'm going to ask any custom question definitely it will be able to generate the answer so no the model consider model modifying your prompt and parameter to behavior okay so I think guys I will have to give in this specific format only means related to this one only uh whatever question whatever uh context they have given likewise you can check but yes again uh there is a chat board also you can check okay uh so let's see what I can ask over here uh hi so let's see what I'll be getting with this particular one no text by the consider modifying your prompt and parameter to improving the PRS okay fine so I think guys uh we'll have to check what I can do from here but yes uh you can explore it and you can run it now uh the main thing basically which I want to show you over here that is what there is a view code now once you will go and check with this view code so you will get the code and guys this code actually it is uh representing to the previous one okay the uh code related to the previous version definitely you can copy and you can check it now uh coming to the next Point here see Delhi so if I'm such if I'm writing anything over here uh inside this prompt let's say I'm writing uh give me a image where where person or where boy is cycling and girl City behind him so this is what this is my prompt now let me generate it and let's see what I'll be getting here will take some time and again uh whatever images you are generating like with it the price is associated okay fine so see guys how like uh like they have generated the image so here is a boy and the girl is sitting behind this boy anything you can gener you can download it from here you can delete it and one more thing show code so just click over here on this show code and here guys you will get the code as well now let's see how we can use this python SDK right for accessing this model so first of all uh these three things is required first is what endpoint URL then API key and then this code let me copy this endpoint URL and see already I have created this uh environment and all everything so here my folder name is what your openi crash course inside this one this is my environment EnV here is my file where where I didn't I haven't written anything and this is my require. txt file okay so what I'm doing I'm creating one more file the file name is going to be deli. py now inside this del. py uh the first thing is what endpoint URL so let me keep it this endpoint URL and guys this is a confidential thing so please don't expose to anyone and after the session I will also delete it so no one can access it so here is what here is my endp point URL I can do one thing I can create EnV file also over here and there also I can keep my confidential information and directly I can read it inside by file but as of now directly I'm mentioning over here in front of you only now the next thing is what API key so let me take the API key also from where from this code so I'm copying over here this API key now perfect now the third thing is what guys here I want to write a code so this is going to my code now let we copy it and here I'm going to be paste it so this is my code now see guys here uh we required what we required this see this this will be default by default for every uh like this will be by default for every model okay this 24201 actually it is showing a model date so you can check out also I think it was visible inside this model section this one model this one created date okay so automatically by by default you will get that date okay now coming to the next point so here is what as your end point so what I can do I can remove this one and I can give my variable name here so let me keep my variable name then the third one is going to be API key so let me remove this API key from here and I can directly pass my variable so here is what here is my API key now coming to the next one so what we required so here we required this open AI so let me mention this open AI inside this re. txe and and yes I can download or I can install this require. txt inside this particular environment don't worry I will give you this all the command related to do related to this environment creation or this uh like related to this re. txt download and all I think many video many project you will find out over here also on this platform uh like from there easily you can learn these things but I will give you all these commands so here if install then iphr requirement. THD so see uh this is running and it is installing until what I can do I can check whether everything is fine or not so p is done then uh here is API key grade so this is what this is my method client image generate here I'm passing my model name Delhi okay here is what here is my prompt give me a image where is a y great and N is equal to one means the number of image so it will provide me what it will provide me a image URL so let me print this image URL so image uncore URL let's see what I will be getting over here so my installation is done if you want to check which version I have installed so here simply you can check with the open AI API open AI Pi okay so just search about this openi pii you will get the version of this openi pii 130.3 now you can go and check inside the release history also so the completion API actually it was available in 0. X okay so it was available in this particular version in this 0.28 0.281 so uh like from 1.x actually it is not available you will find out this chat completion API you can check out you can go and check by U yourself as well I don't have any issue so coming to the point now so everything is fine I can cross check also now once I will write python over here so I can copy this a your open and then I can paste it so if everything is fine I will be able to load it yes it is working fine for me now exit and then I'm executing this particular file so for executing this file simply I need to write this Python and this file name Deli do3 I hope it is visible to all of you if not let me switch off my camera so here see what I need to write I need to write this Deli 3. py or deli. py only this is my file L now see uh I'm running it and if everything is fine guys it is going to generate this URL at the end it will provide me a image URL so it is calling that API and after calling this API I can access the image just wait just a second great I'm getting my URL now just click over here and see here is what here is your image this one guys so again like it has gener different image because second time I'm calling it and yeah I hope you are seeing the image and you will be able to generate it also now you can write a logic wherever you want to save the image wherever you want to generate a number of image but again I'm saying cost is associated with it if you want to check the cost definitely you can check out you can go inside the price section okay I think I already shown you that uh where is a price this one pricing now just go inside the price section and check the price of the Delhi the Delhi price is quite High uh where is the Delhi Delhi Delhi oh here is a Delhi image model is not available in this particular region I think I'm working in es us so see Del e uh $4 for 100 images okay and $8 for this particular configuration and for HD actually $12 so this is the complete configuration uh sorry this is the complete pricing and all for what for the delhi now I hope how to call this python or how to call this openi API Aur open API using this python SD now let me revise this code so what we required we required this endpoint URL and this API key and you have to mention inside this particular class okay we are creating a object after passing this API version means date then a your Endo here is a endo and this API key this API key and this is what this is my client now using this client I'm creating this I'm calling this image then generate and finally I'm passing my uh like config configure parameter and then finally I'm able to generate the image okay now similar to this one I can call the GPD 4 gpt3 model as well so let me write the code inside this chat completion API so what I'm doing guys I already written the code let me give you this particular code okay and then let's try to do it so here guys what I'm doing see this is the code now I have to remove this uh python. EnV because we are not going to use it as of now directly we are copy I am keeping this API ke all over here itself in inside the file so API version is this one oh let me keep it by default let me remove this uh API key and Ando because it is from my previous code I want it uh from my latest code so this is what this is from my latest code let me copy just a second here I'm going to be copy it and after copying I can simply keep inside my chat completion. py so here is what tell me guys here is my chat completion sorry here is my AP endpoint API and this uh endpoint API and this API key now uh let me keep this endpoint API over here so this is the one then I can keep this API key over here okay perfect so let me remove this unnecessary thing if it is not required uh then I can simply remove fine perfect so now client. chat. completion create and here is what what is my model name guys tell me chat in if you will go and check with the model name so you'll find out chat engine let me show you my model name so once you will go with the mod deployment you will find out your name chat e n g i n e that spelling should be correct otherwise it will give you the error e n g i n e great now here is what here is my user which is asking this question what is the capital of Delhi so let's see whether I will be able to get it or not so I'm converting this completion into what into the Json and I'm simply printing it okay so let's see what I will be able to get if if I'm running this particular chat completion API now uh let me switch on my camera so that you can take the access of the entire scam here I'm writing python python chat completion py now uh it is generating guys and if everything is perfect you will get a response see here is what here is my response in the content actually this is in know Json form so definitely you can extract the content from here so just uh like write your logic and you can ex the content so here it is giving you the answer Delhi is not a country so it is not having a capital however New Delhi is the capital of India so it is giving the correct answer because we are using what we are using tell me we are using the GPD model behind so let me ask any other question I can simply ask how many stars how many stars in the universe so this is my simple question guys let's see it's not a simple uh let's see what this model is giving to me so here uh I'm calling this particular API python chat completion. py and I'll be getting my answer so as AI language model I don't have an accurate answer to this question according to estimate by scientist there is potentially 100 billion star in our Milky Way galaxy only and there are billion of galaxy in the observ universe however the exact number is Stars and unknown great guys so yes my chat completion API is working fine using what using this python SDK now uh I hope this uh thing is clear to all of you if it is clear guys please hit the like button and you yes you can write uh in the comment section also sir everything is perfect and everything is working fine in my system also so now in the next thing what I have to do see we understood how we how we have to access this chat completion API now uh if I can keep my own document here and if I can chat with that so how great it would be tell me yes or no right so what I'm doing so here I'm going to be add my own data source okay so let's see guys if we want to add our own data how we can do that so first of all you need to click on this add a data add a data sources so just click over here and then it will give you the various option just click on this drop down and see here is a Azure blob Azure AI search elastic search URL web address upload file so guys I'm going to be upload the file from my local system so let me click on this upload file and then here you can see it is asking me a name okay so first uh okay not name actually it is asking me select a your blob storage resources so first I will have to select the resources as you can see it is giving me an error because I don't have any Aur blob resources I haven't created by created it by myself so what I will do I will click on this link which is uh presenting uh by over here which is it is representing over here now let me click on this link create a new Azure sources so I will be getting the page and apart from this one it is asking me about this AZ your AI search also okay so this both thing I will have to add over here so let me create this one as well now uh create a storage account so here you can see Resource Group this is the subscription now I'm going to write a name name for what name of this uh storage account so name is going to be crash codes okay so this is going to be a name now what I will do I will keep everything same over here I'm not going to change anything just click on this review and create okay so once you will click on this review and create first it will be validating and then it's going to be create this storage account now apart from this one what I can do I can create this AO AI search also so let me write let me select the resource Group crash course and here I'm going to write a service name so uh let me write the service name service name is same crash course so this is my service name and see the reason is what best us so keep it by default and then create it now uh here you will find out pricing tire so standard uh basically 160 GB partition per partition maximum 12 replicas maximum 12 partition okay and maximum 36 suchar unit so it is providing you this particular configuration for what for your uh system for your chat for your chatbot which you will be creating I will be showing you how you like in some time you you will get to know about it so just create over here on this review and create and then finally you can create it so let is uh let it validate and then finally you can create on this create on create one so see my storage is created you can uh go to the resources and you can check and in sometimes this will be also ready it is a submitting the deployment now both I can select over here so first let me check with The Blob storage whether it is coming or not just refresh it and here see crash course is coming coming now let me check as your need permission yes you can provide the permission also so it is asking about this term on course for what for the uh like cross reason so let's say if you have created your storage in the west us and if you want to access inside the East us how you can do it it is giving you the option let me switch of the camera see turn on course just click over here on this turn on course it will be done now here is what a your AI search resources just refresh it and you will get it over here the Scrat course see this one now you can write your index name index name is what like where actually you are going to be store your data okay that particular index so here the index name again what I can do crash course so the same every name actually I'm going to keep the same only so that I won't be having any sort of a confusion now just click on this next and it is asking about the file so which file you want to be store over here which file actually you want to keep it over here now just browse the file and here inside my Lo system itself I have one file let me show you so guys this is the file this shoe let me show you what I have inside this shoe file okay which I'm uploading so let it uh upload the status is spending until I can show you this shoe file so I created it by myself it is my custom file okay so here is what uh just a wait let me open it this is the file guys and inside this file you can see there is a detail of the shoe okay Nike ear XX plus sneaker description of the shoe and then style release date style code and every uh detail actually you will find out related to the shoe now the other one also then three shoes I kept uh just a small data so that it won't take too much time for creating the index and the final one because at the end it's going to perform the similarity search uh I will show you which uh search actually is available over here here you won't be able to find out the vector similarity search you will only find out the keyword similarity search okay so uh now what I can do uh I I kept the file over here so first of all let me close it and then this is the file guys is spending and upload file okay I think I need to click on this upload file and it is uploaded now just click on this next and then it is asking about the Chun size so chunk size what by default I'm keeping one24 means my data is going to be convert into this one24 Chun me inside one Chun actually I'll be having this much of tokens so uh now you can take the search type so I only have this keyword search over here I don't I'm not having this eding search so that eding search is available inside the Azure eii Studio this is a your uh like this is what guys tell me this is AZ your openi studio okay and I will be showing the updated version of it as your AI Studio let on so just click on this next and then here you can review your entire detail and finally save the dogs right so your dog is safe so see injection process is going on and it's going to be convert your data into what into the Chun I don't don't have that much of data so by default it will be taking care of taking care uh take care of each and everything whatever is there inside the file so let it create and after creating it guys what I can do I can ask any sort of a question over here related to this particular document it takes some time it will take like one or 2 minutes so let it upload guys now my data is ready now my index is ready see here uh it is turning to me data source is what so here is what upload file name chunk size every detail it is giving to you so what you have to do first you have to check and then finally we are going to be deployed so how you can check simply you can uh write your message and let me switch off this Jason so yes how I can assist you so I can simply ask can you tell me the price of so which price let me take a a shoe name from here Nike ear Max Plus so Nike ear Max Plus Nike ear Max Plus okay let's see whether I'll be able to get the answer or not it is checking from where from the document itself so the original price of this nikar Max is what 150 USD you can check out inside the PDF and yes the original price is 150 USD let's ask the different question Nike slide okay or Nike sandal so here what I can do I can ask that uh what is the store location so let me ask over here can you tell me the store the store location the store location of Nike or sending so let's see what I'll be getting over here so let me hit enter and see here here here here according to the the restore location of Nik s is bakur Nepal correct guys so it is working fine I can give the thumbs up and my chat bot is working over here now if I want to deploy if I want to make it available for everyone so how I can do it so here right hand side you can see the deployment option okay so just click on this deployment and here is it asking how you would like to deploy a new co-pilot in the copilot Studio or a new web app so I want to deploy it as a new web app so just click over here on this new web app you can select the name so here the name is what uh I can write to app okay so here you can select the subscription let me select the subscription then Resource Group here is my Resource Group location I can select East us only and then pricing plan is free so here uh which uh which app actually so which server you want to take or like it is asking about the pre pricing plan so I'm just keeping it free only I'm not keep like keeping any exard or something premium or something now enable chat history if I'm going to enable the chat history so regard to that actually I will have to pay so that also I'm not going to be enabl over here if I want to do it if I I'm ready to pay yes in that case I can tick mark on it but I I'm not going to be associate any sort of a cost other cost basically so let's deploy it directly after fck this menu of information just click on this deploy and see it is deploying two has already exist great so this is already there I can create a name other name so let's say cre Cod app this is what this my custom name now let's deploy it see whether it's going to deploy or not so here it is uh starting guys my deployment is started it will take some time around 2 to 3 minute so let it deploy so now see guys my app has been deployed so once it will be deployed right hand side you will find this option launch web app so just click over here on this launch web app and there you will find out one complete UI yes so let it load guys it takes some time and uh I hope you are able to understand this pretty simple process and don't worry later on I'll be using this Azure AI Studio which is updated version of this AZ your openi studio and we'll try to uh consume this Ando URL as well but as of now we are just seeing over here itself uh using this uh web app itself and automatically it will provide you an UI uh so let it load just a second and if you're enjoying this session please uh do let us know you can mention inside the comment section and whatever doubt you have directly you can uh mention in the comment or or you can ask me in the LinkedIn as well okay so let's wait guys just a second it is loading yes uh I think it is asking about the permission now once you will accept it you will get your UI so I think now it is done see this is a URL by using this particular URL you can access your UI see start chatting and this UI has been created by this contos so back in a back end actually this framework is working uh this quo framework is working you can check about it it is uh specifically for the UI and all now I can ask any question through my UI so once I will write hi uh definitely I'll be able to generate my answer so let's wait if you are running it first time it will take time and even for the like once will it will open also now it will take time in that also so yes I'm getting my answer hello I can assist you so you can ask the price of the Nike shoot can you share the price of Nike Air 2 Nike Air Max okay let's see uh whether I will be able to get it or not so it is $150 yes it is giving to me now uh I can ask now can you share the detail the detail of Nike sandal Nike SLE let's see what I'll be getting over here so let's uh look into the answer yes I'm getting my answer and it is giving me a complete detail related to this particular s and here it is giving the reference also so it is taking uh it is taking this output from this particular PDF got it I believe this is clear to all of you now see if you want to share you can share it from here you can copy the URL and you can share with everyone now uh through the Aur Studio we'll see how we can expose this endpoint URL in our uh application okay so how we can create the rest API and all everything so uh let me show you one more thing see uh we are talking about this completion API chat completion API and we have seen around the different different model now if you will look into the model so as of now we have this three model om is also there now uh just uh look over here just go inside the chat and if you want to chat with your Omni definitely you can select that also just uh look at right hand side okay deployment side see this one just click on the drop down and select the model once you select the model guys uh now whatever you want to ask to your Omni model can write over here you can mention and see it is able to generate the answer so first of all Let Me Clear the chat and let me reload it also oh I'm doing hard refresh and see my portal is done now let's uh try to change the model so from here I'm selecting this Omni okay everything is perfect and uh now let's ask the question so hi and here let's see finally guys it is generating the answer and definitely you can select any sort of a model from your end now I would like to show you few other model like this for and text to speech and all because those model also is being provided by what by the openi itself now if you will look into the openi API just uh search about the model so the gp4 GPT 4. turbo GPT uh 3. five turbo and even this gp4 you can use related to the images and all okay if you have access of the gp4 over here you can upload the images okay you can upload the images to this portal and you can utilize it so uh one more thing I would like to show you over here that is what that's a uh like whisper model but this whisper model it is not available inside this model section see you cannot see this one so what I'm doing I'm uh opening one more time this uh particular open Ai and I'm creating one more API okay I'm creating one more service inside the different reason so uh let me go to the home here and here as your open Ai and let me create one more that is for what that is for specifically voice okay so I'm selecting a same resources CR course and my reason is what guys so this time actually I'm going to select this specific reason so here the reason is going to be no North Central us so in this particular region my whisper model is available so North Central us this is the South one in the South it is not available this one guys now I can write the name let's say the name is what uh transcript okay I can generate the transcript using the whisper model so here I can write this transcript and then pricing tier is a standard one now uh this is already there now what I can do my transcript let's see this is working or not so yes it is working no it is also not working here I can write audio to transcript so transcript ah fine I think this is perfect let's see my audio to transcript great now click on next and the provided subdomain name I think it is perfect now just click on this next then next then next let it validate if anything is wrong then definitely I will getting the error over here if everything is perfect I can do it great now just create it over here and let's see deployment is initializing and until what I can do guys here see I can keep my data okay so I'm going to create one folder my folder name is what data and inside this data folder I'm going to keep the data okay so what I did I created one file so let me show you my voice file and uh I'm going to be take a input from The Voice itself this time so let me uh show you that particular file I kept in my system only just allow me a minute where is a file file file your okay guys so here is a file now what I can do I can keep it inside this particular folder inside this data folder so reveal in the file explorer so here is what this is my project I just click on this data and inside this data I'm just pasting it so this is my voice file MP4 vile so I'm going to be open it and see listen what it is telling to you uh I think it is a you can hear it let me check my speaker is on or not so so it will be audible okay so is saying list out all the players right in the Indian team so this is the thing which is the recorded one now what I can do I can keep it this one let me uh contr C and here I'm going to be paste it down just a second uh okay it is already done now I'm going to be close it and my data file is over here so let me create a f let me create a file over here whisper. py so what I have I have my whisper. py now what I will do so let me go to the resource here and let me deploy the model just go inside this model deployment manage deployment and then deployment section so I think it is created great now just click on this create deployment new deployment and from here actually you can select the model now see text to speech is there and this whisper is also there whisper is for what for generating the transcript from where from the audio or video I just select this model and you can write the name so here I can you can write chck uh audio to text now here you can select this request permanent limit it is by default three keep it this only and then create it deployment is creating guys here you will get your whisper model now until what I can do see here I'm going to be access it using this python code so I'm going to be write my python code just a second let me give it to you so this is got guys this is my complete python code and here I required what I required my data file so let me keep it and yes everything is perfect see this is my complete python code for what for the whisper everything I will let you know everything I will be explaining you one by one now this load d.v is not required here Json is required request is also required I'm going to be remove this particular part because I'm keeping everything over here itself so first of all guys see this whisper model name Chad model reason is required okay reason in which reason we are working North Central us now along with this whisper model we'll have to create one chat model also so let me do one thing let me create one chat model and that is going to be GPT 3.5 turbo only so just click on this create deployment and here click on this GPD 3.5 turbo and you can write the name chat model okay and then create it so let it create guys I think it is done now uh we required couple of things so first of all what you can do you can select this audio to text and open inside the playground once you will do it you will get endpoint URL and uh you will get what tell me you will get the API key as well so scroll down here and let's see whether it is there or not oh I think you can directly pass the data over here you can put it down and you can check it is giving to you A View sample code yes this is the code guys maybe I'll be getting okay this is the complete GitHub which they have providing to you I required API key and what I required the endpoint URL so just a second I think I can get it from homepage itself just a second let's go back and let me check with this particular model open AI playground and if you will look into the view code yes I'm getting over here just copy it keep it over here this is going to be in point URL and I required the key also just copy the key from there and here is a key then keep it over here I require two thing guys this key and this endpoint URL now both are done now what is the name of this whisper model so in my case the name is what audio to text audio to text okay or video to text whatever you can do now here is what chat model I think this both are fine let me cross check whether it is correct or not just go back and you will get your model name inside the deployment S Audio to text and chat model both are fine now uh reason name is also correct So This is My URL okay so here I have to create my fin URL so endpoint URL and then whisper model name and this is my complete URL guys I have to request to this particular URL okay now here API key where actually I'm passing this particular key so this is my header and URL I need to prepare this two thing now I'm going to read my data so my data is available over here so let me change the name this is going to be data only not a voice data my data is available inside the data folder right I hope it is visible to all of you this complete thing then the next thing here is what here is my data then I'm going to open my file this particular file data file okay I'm reading my data okay so here is what here is my file and this is what this is my uh like URL so here is my data file which I'm passing over here and here is my head header inside the header actually I having a API key and this is my final URL this particular URL I'm requesting to this URL and I'm getting my final response so this is My URL guys okay which is being uh responded now uh after this one see here uh I'm getting my final response so from this final response I'm generating a text and here is what here is my user prompt now what I'm doing so this is going to my prompt me whatever we are saying inside where inside the video okay so that thing I'm able to get from here from here itself then I'm calling this a your open okay the chat model for the chat model only see client chat completion create and here is what here is my chat model this which I have defined because I have to generate the answer now whatever we are saying in inside the audio so if we are asking about the Indian player name I'm passing it over here this is my user promp okay so see I'm passing it inside this chat engine so uh this is my client guys along with endpoint URL and key we are going to configure we are calling this chat completion API to the chat model we are passing this uh system prompt and here is what here is my user prompt now finally I'll be printing the answer let's see whether it is able to generate the final answer or not here I'm going to clear my screen and let me hide to myself so that you can get the access of the entire if I'm searching over here if I'm writing here Python and python whisper. py with. py oh let's see whether I'll be able to get or not no module name request okay fine so here I can mention the request also inside the re. pxd request is here it is request now I can install it on my terminal so pip install request q u let's see guys uh whether I will be able to okay so it is installing so finally I can run my file so python whisper. py it is saying file not found error I think there is a issue with the file let me check guys what is a issue with the file whether it is correct or not so here I'm going to be copy the path and let me paste the complete path because I think there's a issue with the folder name uh here I'm going to keep it and inside this data the file name is what voice mp. MP4 perfect guys I have my file now I can put the double slash because it is required in the windows system and here is what here is my file guys okay let's see whether it is working fine or not now so Python with. py and here I'm going to execute it list out all the player name in the great I'm able to get my text and see guys here is what here is my answer I hope it is perfectly visible to you now you can pass any sort of audio file you can generate the text from there using the whisper model and here you can generate the answer using what using your chat model I hope this thing is clear to all of you now let me summarize what all thing we have learned so far so I'm going through with the pp itself now uh we understood guys about this Azure open a okay the complete detail introduction and after this one we have seen so how you can access the Azure open AI so we have seen the playground chat completion API different different model gbd4 g3.5 I shown you standard quota provision quota everything pricing detail and all I hope this is perfect now my next point is going to be a function calling using a your open a so if you want to make a function calling how you can do that now apart from this one fine tuning and this a your AI studio so this will take some time this a your AI Studio but fine tuning and function calling it won't take any time okay like that much time let's start with the function calling so guys let's try to understand the basic and the fundamental concept of the function calling and then we'll go for the python implementation so here uh let me open my Blackboard on my Blackboard I kept one image and by using this image I will try to explain you the fundamental concept of the function calling so uh here let's say my user is asking a query so here is my user and he is asking a query uh through the chat interface now what is the chat interface guys so chat interface uh it could be any UI UI of the chatboard and QA system so here it is calling okay so whenever I'm writing a qu query it is calling to my openi API through the chat completion API itself so I'm going to be Define my chat completion inside my Cod and then we are calling this openi API now inside this chat completion API I'm going to Define one function okay so this function it is extracting the information from where from the user prompt so whatever prompt I'm going to be defined So based on that particular prompt it's going to be extract some information and then it is calling to some third party API okay and then from there itself it is fetching the information got it now uh I hope this thing is clear but let's try to understand it by using the example and then we'll try to implement that same example okay in the python code only so uh let's say uh here is a user now this user actually uh this uh wants to understand or this wants to know about the latest weather of the Mumbai now uh here is what uh here I'm calling my GPD model GPD uh let's say 3.5 now this GPD 3.5 it does not know the latest weather of the Mumbai or any other city like Delhi bangaluru bopal indor whatever right now this uh openi actually openi model it I don't know the like current weather of the model so what it will do guys so inside this one what I will do so here I will be defining one function now this function okay it will take this location it will take this location and after collecting this location what will happen you know so it is going to be called this third party API for what for extracting the external information so here I got the location let's say I got the Mumbai now uh based on my code whatever code I'm going to be write whatever description I'm going to be write based on that let's say it's going to be called this weather API we e a t h e r weather API okay now now this weather API based on the location it is providing me a information and then this information is coming uh the response is coming to my chat completion API itself and finally it is going to the user itself I hope the understanding is clear to you now if someone is going to ask you in the interview why we use this function calling what is the use of this open a function calling so simply you can say that so this model this uh this this particular model this llm model it does not aware about with the it does not aware with the latest information okay or the recent information so those information we can fetch externally from outside and there this function calling comes we can extract the like necessary inputs from where from The Prompt itself and based on that I can call my further code for calling the external API for fetching the information I hope this thing is clear now let's try to implement it in a python so here guys uh let me do one thing let me so here all I created one file function calling uh f c okay fine this is fine now here itself I'm going to write my code so uh let me write the import statement import OS then uh here I'm going to be import one more module that is going to be a dequest then from uh open here itself I'm going to be open uh I'm going to be import this as your openi so let me import this as your open guys perfect so I have imported this now now what I'm doing I'm going to be create one method method name is going to be main now uh here inside this method uh what I can do I can test it whether it is working fine or not so I can simply write a testing over here so let me write this testing now what I can do I can call this particular method so let me call this method Main and here I can call it inside this if condition so if a name is equal to is equal to M okay now perfect I hope everything is fine uh so what I'm doing guys so first of all let me switch off my camera so that you can get access of my screen now what I will do so here uh actually I'm going to be run this particular file so simply uh I have to write this python function calling uh where I'm writing it guys just a second what is happening let me open the command prom I think that is fine now let me write here python function call great so if everything is fine guys I'll will be getting my testing okay perfect so I'm getting my testing now I have to write my further code so let me write uh the further code so first of all what I have to do guys so I have to Define my client uh I already shown you this uh chat completion API right so from here itself I'm going to take this client now uh let me copy this CLI and here I'm going to paste it inside my main function got it guys now this client actually it is required two things first is endpoint and the second is what the second is API key so let me Define this end point and this API key over here so guys this API key and endpoint from where I will collect tell me whatever model I'm calling okay uh so whatever model I'm calling from my aure open a so that particular endpoint and API key I'm going to be paste over here so uh I'm going to take it from here itself so let me take this endpoint and this API key I already shown you that how to uh created all right this uh model uh how to deploy it and how to generate this endpoint and API key now uh let me paste it over here so here is what here's my Endo now this is what this is my API key okay uh so I got both now fine now what I will do so after calling this one the next thing is what I will be defining the function okay now before defining the function let me do one more thing Let Me call this chat completion API over here so for the chat completion API simply I can write here client c l n great now I can write this chat okay now I can write the completion so here is what here is my completion then dot create right so here I got my create now inside this one I can mention about the model so where is my model guys here is my model now my model name is going to be chat model I already created it I think you know about it then uh what I'm going to do so here I'm going to be write the message so let me Define my message over here my message could be anything let's say my message is this one I'm going to be copy and paste uh this prompt which I already uh created now with that I think uh the idea will be more clear to you so here is my message guys see I'm saying you are assistant which is going to be replying me a weather information and then here I'm asking a weather of the Bal okay I can mention any City and it will provide me a information so instead of Bal let's say I'm writing Mumbai over here got it great so this is what this is my uh chat completion API now let me call it and let me check whether it is giving me a response or not or uh what it is giving to me actually now I simply I can collect this thing inside where inside this completion so let me write over here this completion now uh what I can do so let me uh get the output further so here I'm going to be print this Json that's it guys so everything I took it I took from the previous code itself and I haven't defined any sort of a function over here now I will be defining the function first let me check what will be my final output so uh let me clear my screen and here I'm going to be call this function calling. py uh let's see the output so here it is telling to me missing argument model model and Str Str okay so it is showing me some error guys just wait let me check line number 25 uh line number 25 then here it is telling to me create okay then missing required message and model St Str uh message and model maybe I missed something guys okay messages messages here the argument name is wrong messages great uh now let me check whether it is working or not great so here I'm going to be run it again python function calling and let's see so see what I'm getting here deployment message the API deployment for this does not exist if you can deploy within 5 minutes fine guys so here I think this API is giving me issue let me check whether it is correct or not so what I'm doing I'm just going through what through my aure API as your open API uh not this one okay session expired let me check guys fine I'm going to be login again so here simply I'm going to be write this as your open AI just a second let me open it here let me sign in then I will be getting my Azure open a fine then uh here is my crash course model and this is for what this for the transcript now now let me open this one and here is my model deployment and then manage deployment now uh inside this one I'll be having my models so let me check whether it is there or not great Telly chat engine Omni model and everyone is every model is here now okay perfect this is also fine now let me open inside the playground yes this is also I open now let me check with the view code and here is what here is my API sorry endpoint let me check with the end point whether it is correct or not I'm going to be paste it over here I think it is same now let me check with the op API key so I'm going to be copy it and here I'm going to be paste it guys this one if you want to keep it inside the EnV file you can keep it fine so this both thing are fine Now API and all everything is perfect and the model name is what chat model so let me check guys model name is correct or not so here we are providing with a token name now let me take the model from here okay the model name is chat engine that's why it was giving me a issue so let me mention the chat engine here it is going to be chat engine perfect guys n and my spelling is also correct great now let me run it and let me check whether it is working or not so simply uh I'm uh going to be clear the screen now let me switch off the camera and see python function calling. py yes so fine I'm getting my output and here you will find out so it is telling to me I am sorry I'm a AI language model and do not have a real access to the current weather right and however you can use device built in be app or search online for the readable beather sources so guys this uh this particular code actually it is not giving me a current weather as I told you that it is not capable this model is not capable so how we can do that so we are going to be fetch this information from the weather API okay so open a weather API now let me write over here open by the a map API so from this particular API I'm going to be fetch the information okay now here see once you will uh scroll down so it you will find out this current weather okay current weather data now just do one thing just uh click on this subscribe now once you will click on the Subscribe here you will get this free option okay free option Now using this particular code what you have to do you have to fetch the current weather data now uh here you require this API key okay for fetching the data now just click over here on this get API key now once you will do it guys you will be getting uh this particular page you can fill out the information let me do it over here so here I can write my username and uh I can write my passport already I did it guys so that's why uh it is uh not asking to me I saved it and it is uh giving me a directly now uh okay so I think create account maybe I can directly log in over here because again it will give me a issue so just wait let me log in directly I think I already logged in and here is what here is my API key just a second guys okay so see guys if you haven't created account you can create the account and you can get it now let me click on the sign in and here I can enter my email and password great I sign it now here is what here is my API key so this API key will be required for fetching the information for fetching the data okay so uh if you haven't created account guys first you will have to create the account then only you can get this API key now let me copy this API key and here I'm going to keep it somewhere inside my file only so let me open the file and here I'm going to be keep this API key as of now now one more thing will be required guys I have to create a URL okay a specific URL so uh let me show you from where actually you can get the URL so once you will go back so here just uh click on this one current weather just uh once you will click over here on this current weather so see this is the URL which will be required okay so how to make an API call for getting the current weather this one and and for that you will be required this latitude and the longitude and this API key API key we have already collected and latitude and longitude I will show longitude I will show you how you can get it so just copy this URL from here and then keep it somewhere inside your uh keep it somewhere inside your file only so this is the second thing now one more thing will be required see if you want to get this latitude and longitudes for that actually you will be required the location by using that particular location you can get this latitude and langit longitude now just scroll down here and there itself actually you will be getting the option you will be getting one more URL for fetching the latitude and longitude okay let me show you that as well so just scroll down scroll down and here see guys city name so once you will write the city name means once you will write a location so what you will get you will be getting the uh latitude and longitude of that a specific City just copy it and keep it over there I'll be showing you how to call each and everything after defining the function and all so uh I'm going to be keep it inside where inside my file itself got it so this three API so this three thing now what I can do I can replace this uh API key okay or what I will do directly I will be pasting inside the code itself and you will be getting now one thing I haven't talked about the function okay so first of all let me Define the function so that function basically it will fetch the location from where it will fetch the location from The Prompt itself right right and then we are going to be pass that location over here to the city name and based on that I'm going to fetch this latitude and longitude and finally I'll be generating my current weather based on my query getting my point now let's try to perform it so uh let me create a function guys so I already created a definition of the function I'm giving to you and here uh this is the definition so inside the definition the first name is what the first thing is the get weather this is the name of the function here is the description of the function okay description is what you can see you are you have to retrieve the time information data about a particular location and place simple uh description now parameter this is very much important guys this particular parameter now see uh here in the parameter actually we have given this location so this is going to my parameter and this parameter I will be fetching from where tell me this parameter I'll be fetching I mean this location I'll be fetching from The Prompt itself automatically it will take it or automatically uh like whenever we are going to be Define a function it will collect it from the location itself or if we are going to be Define the a few short prom okay let's say we are going to give a description and based on that I have to fet some information so that uh thing in that specific manner I will have to mention over here okay with those specific parameter so later on I'll be showing you that particular example as well but over here I hope this thing is clear this is a simple this is a simple function definition so which thing we required we required this location okay so function calling is done now coming to the next part so here I have to write a code for what tell me I have to write a code for the uh latitude and the longitude so first I will I will be giving the city name means the location uh which I'm fetching from where from The Prompt itself using this function and then I'm going to be pass it to this latitude and this longitude okay so uh one more thing guys I told you that over here inside this completion API I have to Define the function also so let me Define the function so here is what here is a function parameter and which I'm going to be defined inside my completion API fine so everything is perfect everything is clear now coming to the next part now if we want to collect okay if you want to collect this uh latitude and longitude so we'll have to write the further code so first of all let me remove this print statement as of now it is not required now just focus guys so here is my code okay I already return it let me give it to you this particular code and one by one step by step I can exper you so uh see uh this is the code guys okay just just focus now and I just just try to understand each and everything whatever I have written over here okay now the very first thing so let me align this entire code and uh I have to write it down just for saving the time and uh let me explain you this one this is a specific function for fetching the data great guys fine so let's uh discuss this particular function okay so let me switch off my camera so that you can get the access of the entire screen and here let me remove this also fine now just focus here so first of all guys what we are doing see we are going to be uh load this input okay sorry this output whatever output I'm getting from where from this completion API okay so here uh what I'm getting tell me I'm getting the output and this particular output I'm going to be defined inside this initial output okay inside this particular parameter so I have this Json let me check whether I have Json or not so here I have to import the Json got it so this is my Json now see here is what here is my initial output which I'm getting now from this particular output I'm going to be def I'm going to be F the location okay so I'm getting a location over here I'm seeing over here okay if location is exist so please do let me know the location I'm simply printing it and then I'm going to be call it okay I'm going to be call what I'm going to be call tell me I'm going to be call this get weather method so here to this get weather method I'm passing this location okay now based in this location based on this particular location what I am getting I'm getting uh my response okay once I will hit this particular URL and from this particular response I can collect the latitude and longitude based on my location only so I'm printing my latitude and longitude also and then this is my final URL uh with the latitude and longitude and based on this latitude and longitude we are going to fetch the tell me guys what we are going to fetch we are going to fetch the final response and here we are printing uh this particular response getting my point yes or no guys tell me I hope this is clear now uh here in front in place of this particular URL whatever is there so previously I was testing with the API it was working fine so I have I had generated this URL now let me remove this URL and let me put mine whatever I have created so first of all let me keep this API key so here is my API key I'm going to be copy it and I'm going to be keep it over here that's it now the next thing is what so over here after keeping the API key guys what is required tell me so here the location is required uh so this side also I can mention the API key perfect guys let me keep it inside the double code great now uh here city name so city name actually I can replace with the location so let me keep the location over here this one and here perfect now what I will do I'm going to be concatenate this particular string this one and this one guys so let me keep it like this great and here this I can keep inside the double code okay perfect and here I can keep this particular value inside the double code so that is going to create my entire URL okay so I can remove this uh this one because it is not at all required fine so this is my complete URL which I got now simply I can copy and paste it over there because I want it with this particular API key so here is my URL let me remove it first of all and perfect guys this is my latest URL now same thing I have to do with this one also so first of all let me remove this uh last one because I already pasted and here I require this latitude and longitude so let me keep this latitude and longitude because I want in a Str strr form so here is my latitude okay great now what I will do I'll be keeping it over here and then inside the double code that's it so this thing I'm keeping inside a double code and now plus and then s Str longitude okay I have to create a URL in this particular form only so that's why I'm doing this each and everything now uh this is my SDR longitude now I'm going to be click on this plus and perfect guys so I got my this URL as well fine so I'm going to be keep it inside the double code yes perfect so here is what here is the uh longitude and now let me keep the latitude also guys just a second uh so here is what here is my let SD uh weather is here weather and then s Str perfect guys so this is going to be my latitude okay so just a second I think I did it wrong now let me keep it over here that's it guys perfect so uh now let me cross check this particular URL whether it is a correct or not and here for this one I can simply check with my WEA open AI so weather open Ai and this is the one just give me a second and I'm finally going to call it so here is the lead guys so this is the thing basically which I was missing after the weather and please check out with the original one if H something is missing or something is there perfect guys so this is my complete URL now I'm going to be copy and paste where tell me over there itself so here is my URL so let me remove it and this is going to my request perfect and yes so see guys I have pasted my URL with the API key and all everything so let me remove it and then everything is fine everything is perfect so uh let me revise the code once and then I'm going to be execute it and we'll get the weather from the different different cities and all so uh first what I did I defined this main function this is my endpoint key key endpoint and this API key this is my client okay and and by using this client I'm going to be create this chat completion API there this is my model and here is what here is my message then I have defined the function for what for fetching the information from The Prompt itself so this is my function and here I'm fetching the location and based on this location okay I'm going to be fetch the uh like necessary information okay so here I'm going to be F the location from where from this particular response and then uh based on this this particular location I'm going to be called the open weather map API and here is my final weather so let's see whether it is working or not so first of all I'm going to be clear my screen just a second CLS and then let me call this python function calling. py great guys so let's wait and check fine it is telling to me uh Json decode Json coder expert value line number one okay at which line I'm getting error guys let me check over here line number 65 line number 65 okay and then line number 42 maybe I missed something over here uh line number 42 Json loads and here is what function argument great guys this is perfect Now function argument is the location itself okay wait guys let me check what is the issue so here guys you can see we are getting our output so what was the issue so ISS issue basically it was was with the model so here there what I did uh I just changed my model means I redeployed that particular model and now I'm getting the output because uh see I'm passing the input uh prompt to the model and it was not able to generate the final output this uh sorry not final one this initial uh like output this initial response and because of that it was not able to decode this uh Json basically that's why it was generating the issue so what I did I did nothing I just like uh redeploy the model maybe it was not able to connect and now I'm able to get my answer so let me show you how it is going to work I will give you the entire code and in a similar way you can test if you are getting the same error which I am getting in that case you can restart your model or you can delete the previous uh deployment and then you can redeploy it and then finally uh it will give you the answer so uh let me check with the other location over here so here uh I tested with the Bal uh here let me write the Delhi so after writing a Delhi I can uh write over here this python function calling now see uh once I learn it guys here you can see the city name is Del latitude is this one longitude is this one and here is the weather condition so the weather condition is has now uh see if you want to uh if you want to know more information about the weather you can get it so here it will give you the complete and a detailed information about the weather I'm just going to be check the like type of the the weather the weather condition which I'm going to be uh collected from here and finally I'm printing it that's it but if you want to get a like a temperature cloudy this that whatever right so here you'll be able to find out inside this particular variable now uh what I can do I can show you so I can I'm going to be print this final response so that you will be getting what all information actually I can get with respect to the weather so uh here uh again let me run it python function calling. py and uh now see it is Delhi ltitude longitude great and see guys here weather condition is has but apart from this one it is giving me each and every information so here's is the description now humidity pressure temperature temperature minimum feace like okay every information so uh see guys over here uh you will be finding out finding it out for the console itself I hope this thing is clear now uh what we can do so we can understand the next module now in the next module guys let me go back to the uh let me uh go back to the PPT so we are done with the function calling in the next module I'll be showing you this fine tuning so the fine tuning first we'll select the model then we'll prepare the data and then uh we'll do the cost analysis so uh here guys I will be doing the cost analysis and on top of it I'll be performing the fine tuning because the cost of the fine tuning is very very high and it takes lots of time as well so here definitely I'll be guiding you related to this one and finally we'll be starting with the Azure open a sorry AZ your AI studio so let's start with the fine tuning so before starting with the fine tuning let's try to understand the fundamental concept of the fine tuning so here guys uh let me show you my Blackboard and over my Blackboard what I kept I kept the different different model and related to those model you find out the parameter so uh let's look into the parameter so this GPD 4 uh it is having around 1.70 1.76 trillion parameter now this GPD 3.5 is having 175 billion parameter GPD 3 it's having again 175 billion parameter now here you can see some other model as well so this Bard is having a 345 million parameter this a t511 billion parameter okay uh let's look into the Jimmy model So jimy Pro it's having around 175 billion parameter now this llama right so we have a different different variants of the Llama like uh 2B sorry 7B 13B 17b so it is representing to the parameter only now first of all we need to understand the meaning of this parameter so let me write over here this a parameter is nothing guys this parameter is uh called bits and biases okay so bits and biases itself is called called what it's called the parameter now whenever we are talking about the model guys so here let me write the model or let me write any sort of a model let's say here I'm writing this GPT or maybe jimy so this model actually they are using Transformer as a base model I think you know about the Transformer right this attention all you need research paper uh you can check out with that research paper you will get architecture of the Transformer there you will find out the uh inside the Transformer architecture you'll find out the neural network there are various stages of the Transformer so first is uh embedding okay encoding Ming then uh there is a multi detention and then from the multi attention it is going through the neural network itself and then uh the normalization layer and all everything so uh inside the Transformer actually you will find out the neural network and this neural network actually uh it's a collection of what weights and biases so this weights and biases is nothing it's a trainable parameter this is what this is trainable parameter now just think over here if we are saying three uh so if we are saying like 175 billion parameter so what would be the size of the architecture okay now if we going to if we going to train these type of architecture so how much time it will take it's a like a very uh typical question okay so like no one can answer uh other than that person who has trained the actual model so here we are talking about the llm model which is a very very huge which I just shown you so it is having a lots of parameter like bits and bies it is called the trainable parameter and this bits and bies is nothing it's a part of the neural network only whenever we design the neural network there actually we connect this input layer to hi layer with a bait and vises and Hider layer to Hidden layer with a bait and biases okay so I hope this thing is clear to all of you now see guys let say we have train our model on top of one specific data uh what I did uh let's say this is what this is my GPD model and this particular GPD model I train on various data so here this is the entertainment data now this is the sport data okay now this is the politic data now this is the historical data of various data from where from the internet and now it is capable to answer related to this particular data related to each and everything now tomorrow guys let's say there are one more data which came so here uh one more data is related to what uh related to Indian election 24 okay so here guys see let's say tomorrow uh we came with a we came up with a new data and I want to ask something okay I want to ask something as of now see my model is just is just train on this e-commerce data or uh basically this uh uh basically this particular data right Sports and this uh politics and history and all but it don't know anything about this Indian election 24 so uh let's say if I'm going to ask any sort of a question that who will Who will win the election in 24 right so who will win the election in India in in 24 so this is my simple question to my model now will it be able to answer this question tell me will it be able to answer this particular question no it cannot answer to this question why because it don't know anything about this particular data we haven't trained it we haven't trained my model on top of this data okay so my weights and biases actually it don't know anything about this particular information so what I can do so in this case I can perform the fine tuning fine tuning actually what it does see so here we will be passing this particular data so again this weights and biases means in short I can say parameter so my parameter actually it's going to be adjust okay my parameter this parameter is going to be adjust and now if I will ask this particular question to my model definitely I will be getting the answer why because now I have retrained my model on top of this data I have adjusted my parameter this bits and biases and now I can ask the question related to this particular information getting my point and this retraining okay additional training additional training it is called fine tuning I hope this thing is getting clear to all of you now let's do the F tuning of the GPD model using this aure open AI first of all what you have to do guys you have to open your Azure open AI uh a studio now inside this one just go and check with the model so here you will find out the various model and this is all the pre-trained model which someone has St and we are using just click on this create a custom model okay now here you find out it is telling to you this fine tuning is not available in this particular reason and I think I shown you initially that uh you can check the reason in which reason fine tuning and all which is available so what you can do so here uh simply you can uh go and check with the documentation let me show you again uh so this fine tuning is not available inside this East us uh but I have created one more a your openi service with the c uh with a South Central us okay I will show you that just a second so let it load or what I can do uh here actually I can just wait so let me close it first of all and I'm going to be open one more as your open Ai and with that basically I can show you oh just a second I think my Internet is slow great guys so here I have opened my Azure up here I I opened my Azure portal now let me open this Azure open Ai and here I already created two service so this is my first service which is inside this East us and the second was inside the nor North Central us sorry I was telling the South Central but it is inside the north central us once I will open it so here you will find out uh there is a model deployment now just click on this manage deployment and then uh okay until I think it is loaded now okay fine so I got it so see inside this particular section SE this fine tuning is not available we uh inside this fine tuning model and what is my reason my reason is the East us so what you can do just click on this drop down and then select this South Central USC still we are not able to get the fine tune model whatever model basically which I can find T so just again click on this drop down and select on this uh select this nor Central us now see guys we are able to get this model okay which I can find you so web with the $34 okay D in with the $40 uh GPD 3.5 turbo $45 GPD 3.5 turbo $68 got it now uh what I can do so first of all let me show you how you can f tune how you can start your F tuning process and then uh we'll perform we like calculate the cost and all what might be the cost okay the minimum cost so first of all see we are inside this particular location north central us now just click on this model and then create on this custom model now here there is a various uh stages okay various stages which you need to fill out so first you need to select the base model so here you have a b DaVinci and the different different variant of the GPT and this GPT 4 now what I'm doing I'm selecting GPT 3.5 turbo and here you can decide your model suffix if you want to add something it is optional only so let's say here I'm writing uh my fine tune latest right so this is what this is my suffix means at the end of the model I will be getting it now just click on this next so here it is asking about the data so I already kept the data guys let me do one thing let me remove this data I'll be showing you from scratch okay so I'm going to be close it here is what here is my data file and then let me delete this data file fine great so again I'm going to the deployment and then custom model here actually I deleted it why I will show you uh basically why I uh stopped it I will letting you know oh let me delete it from here as well so perfect this was my previous job which I was creating okay now custom model and here is what base model this is my suffix my find un latest great now what I can do so here I can click on this next now it is asking me about the data so how you going to provide the data there's a three option choose from the existing file which was there but I deleted it now upload the file from local and here is what as your blob or other shared web location so from anywhere you can keep the file now what I'm doing so just click on this browse of file okay I'm uploading it from the local itself and this is my file now see here the here the extension of this file should be Json okay do Json L getting my point so uh let me uh do one thing let me check with the Google itself in front of you so here I'm searching about this Json L okay Json l file so this is the specific file okay Json lines actually it's the extension of the Json only which this uh GPD model is using okay it want the data in this specific format only in this Json lines now let me show you where I kept the data so I kept the data inside my local system inside my local system only I created one folder find model uh and inside this one I have my data so let me reveal it inside the file explorer itself so this is my data guys inside this fine tune folder test do jonl okay Json L Json line now I'm going to be open it inside the notepad and see guys so if you will look into the data so guys this is very mandatory and necessary thing please be careful here you have to keep your data in this specific format only okay this is here I have kept 10 lines because one more thing if you're providing uh less than 10 line okay here is a 10 10 line basically inside my data if you're providing less than 10 line in that case it won't accept it means the GPD model it won't accept it it will reject it so inside your data there should be minimum 10 lines okay let let me show you one line uh so what I kept over here so what I'm going to do I'm going to select this line completely and here after selecting it I'm going to be open my Json viewer let me open my Json viewer and there I am searching or sorry there I'm pasting this particular line so here is my line right now format so see guys so this is the data okay in this specific format I have to pass the data first is what first is the role so here I'm defining the role of the model so I'm saying you are a customer support agent for the smartphone company who who whose primary goal is to help user with the issues they are experiencing with their smartphone you are friendly and concise you only provide factual answer to the query and do not provide answer that are not related to the smartphone right because I want to train my GPD model on a specific data okay on a smartphone data which maybe don't know so let's say if I'm going to ask anything to my GPD model into the smartphone and if it is not able to answer I can train it specifically I can train on train it on that particular data and that is what I'm providing over here so the first thing you have to provide the data in this specific format and the second thing is what at least 10 line inside your data the third one is what the extension should be Json L Json line got it so perfect and here is what here is my user user question and this is the answer so uh yeah now what I can do I can uh upload the file which I already selected now after uploading the file see it is getting uploaded I can simply select it it is in progress guys great I got my file now just select it okay done then just click on this next and it is asking about the validation just select this file great validation is done Okay now click on this next and here these are the like some option which is which it is giving to you how many number of epo you want to run in number of epo EPO means what forward propagation plus back propagation that is called one EPO so uh if you will go and check with the custom one you will find out you can increase the number of AO by default actually is two only now B size if you have lots of data if you like would like to uh provide the data in the batches as of now there is only one batch means we are providing the entire data but if you want to increase the B size you can do it that is also possible now our learning rate like multiplier so yes that is also like you can you can like put and here you can uh Define the seed value also okay so these are the thing basically which I'm keeping by default only and here I I'm going to be click on this next and see this is the like final information and then you can start your training job now once you will click on this training job so here see guys it it has started with a fine tuning so this is what this is my fine tuning progress now let it complete and uh I won't complete it guys because it's going to be charge you a lot and I don't want it and guys uh believe me not even charge it will take the time as well so if I want to get this particular model so the charge this uh trading time it will take around 30 to 35 minute and after 35 minute I'll will be getting the model after 15 minutes so this is a 1 hour effort okay 1 hour process with a 10 line only just just think guys how big the model is and how much time it will take with a like just a small it's just like with a 10 line only if you are passing like 100 line or 150 line so it might take uh maybe 2 Hour 3 Hour 4 Hour it depends on the data only so here first of all let me calc the cost and then uh I think this is done uh so yeah then I will be stopping the stopping this particular process and see after uh like training this particular model so you'll be getting it over here itself okay inside your model then definitely you can uh deploy it and then you can use inside your chat engine or else you can consume it inside the local in a similar way we are going to be we are consuming the other model getting my point so on let me repeat it once you will uh like train this particular model okay so you will be getting over here along with all the model you can deploy it over here inside this section and then you can utilize it directly inside your playground or else you can access it inside your local system as well using the get completion API fine now uh let's discuss about the price and all so let's say I have decided this particular model so the price is what training per compute R is what 45 tell me guys $45 so here I am going to be open My Epic pen uh let's try to calculate some sort of a price so here guys let's say um great perfect so uh let's say guys here the price for the per computer R is what $45 okay so if we going to talk about here or instead of this one I can open my calculator just a second because I think it is giv some problem $45 for the per computer let's say if I'm going to be uh if my process going to be for the half an hour so you can uh divide it by two itself so there will be around 23 okay now apart from this one here hosting per hour let's say I'm going to be hosted for the uh this particular model for uh let's say 2 hour okay so here uh what is this tell me $3 so here I'm going to be add more $6 now apart from this one see input uses per th000 token so this is fine this is are very less charges I can add $1 more so here around $30 so if I want to use this model if I just want to test it in that case I will have to pay at least $30 guys okay where I'm and where I'm calculating the training time just for half an hour let's say if I'm having a huge amount of data in that case uh it will take around like 2 to 3 hour or maybe more than that so the charge is going to be very very huge and guys you won't believe me I got $120 will within half an hour only so that's why I'll be sto sorry $120 will within one and a half hour okay so that's why I will be stopping the process and yes uh rest of the thing is fine now if you want to calculate into the Indian rupees so simply uh what you can do 30 into let's say I can take 883 only so $1 is equal to 83 RUP now here is the prices around 2.5k right rupia so that will be a price for what for the fine tuning just for 1 hour that's why I was saying it is costly and I think you got to know the process of the fine tuning so let me stop it because it is going to be run uh now here see here you can see inside the notification and all it is creating the fine tune model now uh here I can delete it so so perfect it is not at all required so yes I hope you got to know about the fine uni now let's try to understand the next thing uh let me open the PPD itself so now we'll be talking about this a your AI Studio guys now in this Azure AI Studio guys first we'll start with the Azure AI Hub I will create the Azure AI Hub and the other resources then uh I will deploy the model or I will see that how I will show you that how we can use the model from the AZ openi directly from the AI also you can connect okay inside this azui studio we'll create our own rag uh mean means our custom rag and uh then uh here we going to steam like the entire pipeline using this promp flow it is a very amazing Service uh which is being provided by the aori studio then we'll create an endpoint URL okay means I will deploy the application finally and we we will be consuming it inside our local application so here we're going to create one flask application and there I will show you how you can consume this endpoint URL okay now before it starting with this Azor EI Studio you should be aware about this azori studio guys so uh what you can do you can check out with the uh Microsoft itself with the official page of the Microsoft now just see uh what they are saying about this Azure AI studio so Azure AI Studio it's a unified platform for developing and deploying generative AI application so if you you want to create generative related application if you want to deploy it okay so this a plateform will help you from starting to end it's a unified it's a single platform for everything and it's pretty amazing guys definitely you will get to know once I'll be showing the entire functionality of this a AI studio now you can check the overview so inside the overview they have mentioned explore the model craft it then design it then roliz is okay so you can uh like like like get a evaluation and all evalution metrics and all at the single place after deployment and all then uh here you can see the benefit so what is the benefit it is going to be improve the customer experience it is reducing the organization risk okay improve work quality and then enhance productivity and all capabilities so yes uh these are the capability you can build your own co-pilot Enterprise check okay that we are going to be create here incorporate multimodality means you can uh like keep all the models the different different type of model and one very amazing thing I will show you and you will literally Sol so uh okay just wait I will be showing you that now analyzis spe yes you can analyzis the speech also so yes you can go through with it and pricing and all everything they have mention and yes definitely you will be able to get each and everything now coming to this AO AI studio so for that what I will what I have to do first I have to open the Azure portal so I'm going to be open the Azure portal inside the new new uh like uh inside a new browser new window guys so this is what this is my a portal now let me click it and here guys just a second let me do the sign in and this is what this is my a portal okay great now here what I will do I will simply search this a your AI Studio okay so let me search over here a your AI Studio got it now after searching this azui studio I'll be getting this particular page so I already created this my chatbot my project this uh to basically but here in front of you I'm going to create it from scratch okay so if you want to create this azer UI studio so for that first you have to create this Hub new a your Hub and inside that you're going to launch this studio okay so just click on top of this new a your AI Hub and after this one it will ask you some sort of information so first is what first is a subscription the second is a Resource Group we're going to use the same Resource Group which we have created then reason is what uh East us only okay name you can write any name let's say here I'm going to be write crash course and uh here uh see friendly name is the same you can change it also then default project resource grou okay so this is what default project resource grou now uh after this one it is asking about the aure AI service for the base model so uh See by default actually it is creating a new service new service for what tell me for the base model for the openi model but already we have created that service right I told you now that we can connect with the previous service also so we can connect it over here now just click on this drop down and select this crash course module see there is two one is my audio transcript and this crash course module I'm going to select this crash course module now let me select it and then review and create and see guys it is getting reviewed and then I'll be creating it now I'll just wait and the surprise is for all of you and here the surprise is going to be uh let me show you that so once I will create it so after creating it see the deployment is getting initialized and uh after this one after the deployment and all I will be able to access my a your AI Studio okay so let's wait uh let's wait for a minute I think it will take around 30 to 40 second or maybe one or 2 minute so yeah it is uh getting prepared now so see guys still is creating and once it will be done I'll be getting the resource option see by default actually it is creating these particular resources okay by default I'm not doing it automatically so one is the key w we are actually is going to be save all the keys and all the second is what second is a storage account okay so automatically it's going to create a storage account that blow blob storage account which I shown you right previously so yes I can utilize this particular account and I can keep my data over here so with every Azure AI Hub okay with every Azure AI studio uh this resources will be connected one is what one is this uh like Vault second is a storage third is this Azure open Ai and the fourth one is the Azure AI search that I will show you whenever I'm going to be create an index and all uh at that time it will come now see it is ready so simply you can create on this go to the course sorry go to the resources and there is so many thing right in the left hand side a project netw workking property lock alert metrix and all but here this uh project is uh required okay now uh see guys what you can do so first of all you can see this particular icon launch AI uh launch AI studio so uh once you will click on top of this launch AI Studio you will be able to launch this AI studio right so let it launch and after this one uh see guys here you can see all the options similar to the Azure eii studio uh AZ your open Studio but it is not this a your openi studio it is a your eii studio let me show you if you are getting confused so this is what a your Open studio just for the openi model and all and this is a your EI studio for the end to end thei application so what I'm doing so first of all uh what I will have to do over here I will have to create a new project inside this one so once you will scroll down you will be getting a different different option permission description project connected sources and all everything so what you have to do first guys click on this new project and I'm going to be create my own project let me write the name also the name is going to be a trash course okay same name I'm writing uh for everything so that I won't be confused and I can easily identify so this is what is my project now project is getting created guys it will take some time again uh like uh one or two minute so let it create and after the project guys I'll be showing you that from uh where actually you can find out all the resources and all so inside the project all the resources will be there if it is not there then simply we can connect it also okay so guys here you can see my project is done I created a project now after creating the project you will find out these many option left hand side right so now uh just see uh let's uh look into the options what all option is there first uh let's look into this component so you will find out the data so whatever data basically you want to keep you can keep the data similar to the Azure uh openi Studio then you have a indexes indexes with respect to the vector database okay then here is a deployment so whatever model right whatever model basically you have deployed you will be able to find out each and every model over here now content filtering so if you want to filter any sort of a Content you can create over here and you can filter that right uh means uh you can connect it with your model and application and you will filtering that specific content now here you can see the fine tuning yes over here also you can do the fine tuning and all now evaluation metrics evaluation is also there you can perform the evaluation you can create the evaluation so ground Ruth relevance coherence fluency everything basically whatever metrix is required for evaluating a model now tracing right so here you can trace yourm application now one more thing prom flow so prom flow guys it is very much important and with this also with this only we are going to be create the Anto application now code is there if you look into the code so yes uh like you can create this code okay you can select your CPU or GPU you can select the virtual machine and over there you can run your own code right so these many functionality they are providing to you now chat wise so chat completion API completion API assistant is also there image related like Delhi and all everything is there and other than this one model catalog model Benchmark prompt catalog and all so this thing is not required just for the information and all you can check out so first of all what we have to do see guys you have to go inside the setting and you have to connect with all the services and all so guys once you will go inside the setting here actually you will be getting the option okay option for the new connection so see these are the services which is there okay connected resources crash course model this means my Azure open here is what here is my artifact storage okay means my blob storage and the other one like the Vault and all uh and I will show you one more a your AI search Once I will be creating my rank with my custom data so just click on this new connection guys and here uh see there are like a different different option you will find out uh now you just need to click on this aure openi service so here already we have a service two Services first is what crash course and this my audio transcript so you are able to see the service but it has not connected it okay it has not connected so if it if if it didn't connect it you can connect it manually okay so after click on the connect option you can like connect it manually you can add the connection now I want to add this particular one so just click on this add connection and after adding the connection guys you will be able to find out see it is connected now just close it and go inside the deployment so once you will go and check inside the deployment so now you'll be able to find out the same model which I'm having inside where inside the a your open AI okay inside this specific service so no need to do it from scratch u means uh all the model deployment and all directly you can f the model over here those particular model over here now one more thing guys here see if you will go and check with this create deployment so here actually you will be able to find out two option real time deployment and pay as you go okay so you can select any one from here so let's say if I'm going to select this real time deployment now here inside the real time deployment see uh if I will scroll down so along with this GPT And all I'll be getting the other model as well like fee llama is here okay and apart from this one m is here and this uh Nemo is there from the Nvidia okay Desi is here and many different different hugging face model and that is what I was saying to you so inside this azur studio it's not like that you are just going to be use the opening model no you can get the other model as well and that's the beauty of the AZ AI Studio I hope now you are getting my point so let's do one thing guys let's try to use this model so here I'm not going to be uh like use any other model other than this uh open a and all because uh yes this open a model is a sufficient for me as of now for creating the application but if you want to use uh you can use it I shown you the way how to do it so uh let's do one thing let's try to uh open this chat engine okay this is my model and after opening this chat engine what I can do so see with respect to this one I'll be getting my key and here is what here is my target URL and all everything so just click on this open playground I will be open it I will be open I will be able to open this particular model inside the playground okay inside my chat playground so let's wait uh yes you can see the model over here now I can test whether it is working or not so once you will simply write here hi so if everything is fine I'll be getting my answer so yes uh this is working fine now if I can uh ask some other question so how are you so it is giving me answer that yes I'm fine or whatever so okay now let's try to ask something else I hope it is visible to you let me keep it over here and uh yeah great now let me ask uh can you tell me the detail uh can you tell me about the T okay which is capital of India let's see whether this model will be able to answer or not so here I can see uh this model is able to give me answer uh it's a GPD model so definitely uh it can give me answer okay great now uh I think it is perfect now one more thing guys so once you will go inside this deployment so here see already we have this particular model now let's say some if anyone model is not there so what you can do you can just simply click on this create uh deployment and let's say this GPT is not there so just click on the GPT GP 3.5 turbo and then confirm it that's it okay you will be able to get it right now let's say if you are not getting this confirm option so first connect with your a your openi resources okay a your openi resources and then you will be getting this confirm option now I hope uh this is perfect to all of you so let's do one thing let's try to create our own custom chatboard so what is the procedure of it let me show you step by step once you will go inside the chat guys here see so uh just look over here that we are only able to get so I think it has removed okay fine now I'm going to be ask one specific question hi can you give me a price pricee of Nike Air Max shoes Okay so this is my H question which I'm asking so let's see what I'll be getting over here so it is telling to me that as a AI model I don't know about it okay I don't having a specific information uh I hope it is visible to you guys let me hide to myself see so it is saying it is not having a specific information with respect to this particular question so I want to create my own chatboard okay so which can like answer for these type of question so what I did here let me show you I kept some data okay inside my system and on top of that data basically I'm going to be train what I'm going to be train my model okay so uh just wait let me open this so here is what guys here is my data and I hope it is visible to all of you see so this data actually it is having an information related to what Rel to the shoes so I just kept only three uh like detail okay three shoes detail because unnecessary it will take a time if I'm going to be create a rack system or if I'm going to be create a like index and all so Nike ear Max Plus Nike Dunk and Nike slide or sandal so this is the information like style code or Price Right date and here you can see the manufacturing unit and all location every sort of a detail you can see over here with respect to this document now what I'm doing I want to create a custom chatboard which is uh like which will be like giving me a answer with respect to this true and all so what I can do here so first of all I have to add this particular data now for adding the data guys here you can see this option add your data now just click over here on this add your data and click on on top of this add a new data sources okay now once you will click on this add to the new data sources now just select the data sources from here I want to upload this file from my local so here I selected this local now I I'm going to be upload the single file so here let me take this particular file so the file name is what Sho so here I got my file uh the file name is Sho now once I will get this file I will click on this next and here see it is asking me about this a your AI search service so let's try to see what is this a your AI search service okay so here if you will simply check with the Google let me uh search over here as your AI search search okay now uh what is the Azure AI search so just read about it Azure AI search okay it is also called a your cognitive search so guys just listen over here whenever we are talking about this uh aure cognitive search it's a very important part of the rag okay because it is uh this a your AI search or a your cognitive search actually it is only giving the uh permission right for the similarity search and all okay so based on the query whatever query will come it is keeping inside the database and uh it is keeping the data inside the database now whatever query is coming it's going to perform the similarity search keyword search okay and the name is what so before the name was the AER cognitive service and now the name is what Z your AI search provide secure information retal at a scale over own contain inational and generative AI Search application so uh here in short you can see it is nothing it's a information retrieval Source okay it's a information retrieval system if you want to read more about it definitely you can check over here so how it is going to work and like the full flag uh uses and all okay with respect to this uh a your AI search right so what I have to do here I have to configure my AIO eii search now already I have so here what I can do I can connect with that so already I had created guys so this was the one now this crash course now I'm just simply click on clicking on this add connection and and here I am getting it so if you remember I've created it when I was running the azeri studio when I was keeping the data okay the custom data and I created one chatbot before this one as well so uh here my source name is what crash course now index name I can keep uh anything over here I can remove this one so it's going to be my Vector index my Vector index now here you can select the machine also so so let's say uh if you want to select the machine for the entire process so this is the like options now yes you can go through with this particular option and you can check now I'm keeping it auto only now just s on click or just select this Auto and click on this next now here is what here is your embedding model which is taking from this crash course models itself this is AI service AZ your openi service now just click on this next and after this one guys it is checking for the model it is there or not yes it is there then finally click on this create okay so here actually you will be getting this create see so this is what this is my create button so I'm creating this index and based on this particular index I'm going to be perform the similarity search over here so uh let's wait uh it will take some time around one or 2 minute uh for creating this particular index if your data is very very use in that case it might take 10 to 15 minute also so that's why I kept the small amount of data but still it will take some time so meanwhile what I can do I can uh set up my environment and all everything for creating my final application now first of all guys here I'll will be required some folder and file so I'll will be required one file that is what app.py I will be required one more uh file over here that is HTML which I'm going to keep inside the templates okay templates and that is going to my chatbot UI so now here inside this uh template I'm going to be create one file so let me give it to you the code so here my file name is going to be chatbot HTML chatbot do HTML okay now uh here inside this chatbot HTML I can keep the code and this is but nothing this is my HTML page okay so a very simple HTML page I just created it for what for the uh like for the queries and all it's not a full flag chatbot and all means a full flag chatboard page it's a simple uh like UI okay with respect to this chatboard and all I will showing you you will get it now uh inside the app actually I have to write a code with respect to what flask and all because here I'm going to be create my flask application so uh let me give you that code also or as I can write it down one by one so the very first thing which is required the flask only so simply I can write the flask and guys here in this particular uh environment flask is not available so here I can simply say FL sorry render uncore template so here is what template okay now what I will do so here I'll be writing the further code so here is what from flask request this is also perfect now here uh I'm going to be take this particular module as well so I have this URL Li and this Json and this OS okay so this particular module is required now what I can do I can create a my flas application so for that uh actually I simply need to write Here app is equal to flask so here flask and inside the flask this name got it now here I will be creating my URL so my URL first URL is going to be uh app. route and here I can simply write or okay I can simply pass the SL which is representing to the Home Route and then I can pass the method so here is what here is my method so method list so I can get I can pass the get and uh post method so first is post and the second method which is going to be get only got it guys I hope this is clear now coming to the next part so uh what you can do so here you can create the method which is associated with this particular route so here my method name is what handore form so it is going to be handle this particular form any anything you can write any method name I just this method name came in my mind I'm writing it that's it so inside this particular method you will be coding and or what you will be doing you are going to be return the particular template okay so which template guys this chatbot do HTML so simply you can write over here return uh this uh return render template okay render template and here you can mention this chatbot do HTML getting my point guyss or no see I hope this is uh like fine to all of you now guys what I can do I can test my application okay so here you can see it is uh still loading it is creating the index okay let it complete now now what I can do I can test it whether it is working fine or not so uh what I will do here so simply uh I'm going to be write this uh python app.py but before that I will have to install I will have to install this flask okay inside where inside this particular environment so what is the command for it f install plus simple now uh let me install it and see it is installing don't worry I will give you this each and every command inside my GitHub I will be keeping this entire code to my GitHub and the link will be available inside the descript deson guys so from there itself you can download it so I got my flask now what is the next thing I can uh run here python app.py so once I've learn it guys here you can see everything is fine but there is one thing which I need to add if name is equal to main so here let me write this name is equal to main okay so this is perfect now simply I will write this app. run and yes perfect so if I'm going to be run it definitely I'll be able to like get my URL okay and with that I can access this sttm this chatbot page so uh let me clear my screen CLS and I think it is visible to all of you let me hide to myself and here I'm going to be show you this Comm python app.py now see my application is running now if I will open it so guys see this is my chatbot page a very uh simple and very easy uh very simple page over here just to test the queries and all nothing else so if you want to create any full full flag chatbot and all full flag chatbot page I can do that also but uh later on I'll be showing you in my some other video here just for the testing around so guys uh let's do one thing let's close it as of now and then uh let's stop this particular process and I'm going back to where I'm going back to my Azo open AI now see here uh it is showing to me that it has created okay this index is created now where actually sorry the chunking is done and here now it is creating the index and all so the first step is done chunking and the Second Step what like it will be creating the eming and all okay and then on top of it like this entire thing basically it's going to be keep inside the index and then I can perform the retrieval operation and all so uh I hope uh this is done now if you want to check how the process is going on so just do one thing so just go inside this uh let me show you where index is so once you will click on this index just open this particular index okay now inside this index here is a one option that option is called job detail once you will click on this job detail you will find out the entire detail of this with respect to this particular job so whatever job you have seen right you will be finding out the detail step by step detail with respect to that particular job now here uh it is telling to you my uh recent workspace my live project just a second guys yeah this one now just uh scroll down and see these are the particular jobs guys inside this Vector so what is my Vector name my Vector index I think my name I written in this way okay fine not an issue and yes so this index is also completed and now registering this particular index so uh first of all let me uh like tell you one thing let me talk about the rag architecture until it is Crea I can give you like little information related to this rag architecture and all so in the rag architecture what we do so let's say we have a data now what I will do I will be collecting this data and this particular data I will be uh like storing somewhere okay let's say this data I kept inside the database then uh from this particular database I'll will be collecting this data again in the form of chunks okay chunking I'll perform the chunking because this data might be very very huge so let's say this is what this is my chunks now with respect to this chunk I'll be creating the embedding Okay so so here I can write chunk this is also my chunk here also my chunk okay now I'll be creating the embedding so embedding is nothing it's a numerical representation of the data so here is what here is my embeding this is also repres to the representing same thing and here also embedding now this particular embedding guys whatever embedding we have created I'll be storing inside where inside the vector database so here here is what here is my Vector database or like this Vector database or you can say Vector index because this Vector database actually it will provide you the vector index so the same process is happening over there itself so what they are doing you know so whatever data I have provided the shoe data they collected it means they kept inside the database aure blob they created a chunk from there okay if it is required they will create otherwise no because I kept every setting by default so I'm assuming that they have created a Chun because the process I can see over there then they have performed the iding with respect to that the model right so emding model which I loaded so by using that particular model they have performed the iding and then they stored it right where inside the vector database means by default they are using the vector database or uh basically with respect to that they are creating this Vector index okay Vector database actually it provide you the vector index getting my point now what will happen you know so whenever user is asking any sort of a query so here let's say is my query which is being asked by the user so it will perform the similarity search based on this particular query so let me mention the similarity search and here this uh context I will be getting the context means I'll be getting the detail this operation is called retrieval operation and this a particular context along with a prompt I will be adding my prompt and this two information I'll be passing to what I'll be passing to my llm I think it is visible now I'll be passing to my llm now llm what it will do so it will read this prom it will read this context and finally it will be generated it will be generating the answer with respect to this particular query so this is the whole sort of a process now uh just see over here everything is done with respect to my uh data now what I will do if I'm going uh if I'm going now with a CH chat so there I will be able to find out my index okay somewhere uh it has uh this particular index in some database I don't know about it uh if I'm going to check the detail definitely I'll be getting but here you can think that it has created the vector index now I can uh select it okay so here is what here's my Vector index which is already there now I can select how it is going to be find out the similar search so here I'm going to be select both option one is Vector and the second is what keyword okay so hybrid option now uh let's say uh uh let's ask the same question to my chat bot over here to this uh okay so it is tell me content cannot be empty validation error message zero why it is so guys I think I will have to select the model so model is also there now let me ask one more time just a second okay great now I'm getting so here uh what I can do I can ask what is a price what is a price of nice sorry Nike Air Max shoes so previously actually it was not giving me the answer but let's see whether it is giving me answer or not so now here the price of this Nike Air Max blush true is what guys 150 USD if you will go and check with the PDF you will be finding out the same price okay now you can ask any sort of a thing uh to your uh model with respect to this data and here see uh it is showing me the reference also from where actually it is taking a reference now I want to use this rag system the rag this one retrial argument generation in my own in my own chatbot see here is what here is my chatbot which I have created so let me run it python app.py it is running on top of this particular URL so see guys here if I'm going to be run it so I will be getting my chat board this one how I can integrate this Rec system into this chatboard now we'll see that particular process so here guys if I want to do it so uh I can do it right but how so this prom flow is The Rescuer okay by using this prom flow I can do it now uh let me show you how it is possible so once you will click on this prom flow uh this prom flow basically what is the meaning of the prom flow guys so by using this prom flow you can schedule the end to end pipeline okay end to end pipeline from data injection to this uh like uh retrieval and then this uh chatbot generation and all means this data generation and then deployment also so what here basically what I did I have created this pipeline now I'm going to be deployed by using this prom flow so I'm going to write my own name the name is going to be crash course over here now uh let me open it and here you can see it will provide me the location as well just simply click on this open and see it is going to open this particular folder so just a second uh it will take some time so now see guys uh it has opened uh this particular prom flow so uh inside this prom flow see there are lots many option okay so which option you need to select so first of all see let me show you the files and all so here uh you will be able to get all the files because see at the end this is nothing it's a code only right so whatever I have I have done over there uh the data uploaders that data upload and all with respect to that it has generated all the files and folder see it is a search it is with respect to the search this one see now uh here you'll find out uh this one with respect to the chunking and all see this is the require. txt if you want to see the content also uh you can like you just need to double click on top of it and here is what here is a content of the file all it now what I'm doing see uh for deploying this particular application first of all you need to select this drop down okay start comput session and then just click on the first option is start compute session now once you will click over here guys so here you will be able to find out it has started the computation of this particular session okay now once the computation will be done I this chat option will be eligible for me I can chat to my model OKAY over here now until what I can do I can give you the complete walkth through of this prom flow it is creating so uh see guys here actually what this prom flow is doing it is going to be create the entire flow so from this llm so this is my llm name okay now here is what here's the output what output I will be getting okay from my model then here you can see this is the script and all the complete uh like script it has created in a back end then other than this one here you'll find out the other thing like prompt and all how it's going to Define The Prompt okay in a back end for handling the for handling each and everything now apart from this one see extract search means like keyword search and all whatever we are going to be per form see with respect to that there is a code so this each and every code actually it took from where tell me from my configuration so whatever configuration I made over here inside this chat I directly imported it inside the prom flow and from there actually it took this particular configuration now let's say if we are not doing like this if you want to create our own custom flow okay so yes I can do it once you will click over here and once you will click the prom flow and then you can fill the information over here manually and according to that it will be generating the code that's it okay now see my computation is done now I can chat with my model okay so whatever model and whatever data I have provided the rag which I have created I can chat over here so once I will click on this chat I hope it is visible now let me keep to my uh keep myself over here and see if I'm going to ask any sort of a thing the same thing to my chat about over here so definitely I will be getting it so it is running just a second it takes some time if you are running it first time definitely it will be taking the time so simply click on this uh this particular option or if you will hit enter then also you will be getting the answer taking time guys taking time yes running is still running and here over the prom flow each and every service will take some time if you want to deploy the application it take around 10 to 15 minute okay fine guys so here you can see we are able to get an answer so here my answer is hello how I can assist you now I can ask something what is a price what is a price of Nik here Max 2 let's see uh I'll will be getting my answer on not so it is taking some time over here guys but don't worry uh once I will deploy the endpoint URL it won't take this much of time and we can configure this latency on all everything I'll be showing you how because at the end it will give you the option to select the instance okay so let it are done okay fine I'm getting my answer and the price and all everything it is providing to me so I hope guys uh you are able to get that how to run this computer session and how to test it now what is the next thing I have to deploy I have to public it okay I have to generate the endpoint URL so that I can integrate in my particular chat Bo so for that uh here let me do one thing let me not this one this one guys so what I can do guys here I can click on this deploy okay so let let me adjust to myself first of all great uh fine now uh see just click on this deploy here and after clicking on this deploy it will ask you the uh like some sort of a detail so first uh like it is asking about the endpoint URL you can uh keep you can write manually this particular URL then deployment name you can write the deployment name okay manually but I'm keeping it I'm keeping everything like by default only now here is a virtual machine so 2 core 8 GB 16 uh 2 core 8 GB RAM and 16 GB is a dis and here is the charges for the virtual machine okay 0.1 per hour so I'm going to be select this standard only but if you want to select any custom one then definitely you can select that also uh let me show you where this one so once you will click on this drop down you will be getting all the option with respect to the machine so that is what I was telling to you if you want to make your process faster you can like select the instance according to your requirement okay now what I'm doing so here I'm keeping it this by default uh this a particular machine only so what is the which is the one so here is this EXT standard one now after this one guys the next thing is instance count so I will be keeping three instance okay now inferencing data collection okay so here I'm keeping enable only all the information by default now review and create now uh see you can review all the information one more time and finally you can click on this create now see here my application is going to be deploy guys so it takes time it will take some time now how you can check you can check inside this notification so this is a notification guys and my model is getting deployed so once uh it will be deployed you will get the notification over here inside this particular section so let it deploy let's wait uh for some time until what I can do I can write my code okay for the further thing so uh here I was returning this render template but uh I don't want as of now this one I'm writing a full flag code so first is what first uh is a if else condition so here I'm going to be write the condition my if condition so inside this condition guys or I can do one thing I can give you the entire code and we can understand like a step by step I think that would be great so see this is my complete code guys it is very simple very easy nothing is here and uh easily you can understand this particular code so inside this code see what I have I have my if and else condition so if request. method is post in that case what I'm doing I'm getting the user query okay from my user side I'm converting uh this query in this particular form okay this uh Json form then here is a body okay means I'm going to be dump it Json do dump data and then I'm going to be encoded right so in this particular format actually I want my query this is a specific format which is required okay to sending the query to my end Point URL now here is my URL guys so here actually I'll be keeping my URL which I'm generating so let me remove this one from here and let me keep my URL I will be showing you how you will be getting it now remove this API key you can keep your API key I'll be showing you how you will be getting it then if not API key I can generate this particular exception now this is my header guys okay so in the header you can see there is my uh like content this is what application SL here is what here is my API key and then this is my application name okay the deployment name now apart from this one see here what I have I have my request so I'm going to be request to this particular end you endpoint URL here is my URL and here I'm getting the response okay now from this particular response I'm collecting my text okay whatever uh thing I required and then finally I'm going to be return it to my HTML page getting my point if it is going to generate any sort of error in that case there is my exception module now you must be thinking from where I got this particular code so guys I didn't get it from my end I didn't write from my end I got it from The Tell Me from where from the uh like uh prom flow itself I'll be showing you after deployment you will get the entire template over there now here uh in the else uh it's nothing so yes if it is not a post request in that case uh yes we are just showing this particular template that's it now if you will look into the HTML code now just go and check with the HTML code so here uh what I have so it is very simple code now this is my header okay head and here is my body in my body I have one div this one right which is showing me a reply this a particular div and this is my form okay you can create form before the DI this is also fine now here is my form which is a like which is working on this particular URL uh this slash only and here I'm taking a user input okay now after taking this user input what I'm doing I'm uh like I'm click like we are clicking on this submit and finally uh I'll be getting my reply over here based on the like query okay so this is a simple HTML which uh I have written for all of you and here is what here is my app.py file now I have to keep this two information over here so uh what I can do I can check whether it is deploying or not uh Let me refresh uh this particular process and uh here I think uh it is deploying so just a second let me check guys whether it is deploying or not reloaded uh yeah see deployment is uh yes yes yes it is deploying so just go and check with the deployment detail you will be getting each and everything over here itself so detail is here guys this one and uh here is what here is my URL this one and this is my key okay right hand side you can check and is still it is creating see if you are not getting the deployment and all inside the notification you can simply refresh it getting my point so uh here see it is deploying and now this is my endpoint URL which is required to me I'm just going to be copy from here and simply going to paste inside my code okay now the next thing is what so here the next thing is this uh API key so just copy this API key from here and then keep it inside your code okay so here I'm going to be keep my API key now uh both thing is done now what is required so here I required my application name so I can remove this my project this one and what I can do I can keep my name so here is my name guys crash course r y DD so this is the name basically which was there and if you remember then at the end basically there was one so uh crashed course r u y dd1 so this is the particular uh requ request header and I hope this is clear to all of you now once it will be done guys I'll be making the request to my endpoint URL it takes some time so if you have initialized uh maybe uh at 150 and it will be done within like 10 to 15 minute this is a very minimum time for it it might take more than that uh as well so if you want to increase the time in that case you will have to play with the instance configuration and all everything so let it complete and then uh we'll hit the URL guys so here guys you can see my deployment is done so uh uh guys uh whatever name uh basically I given to the deployment it was giving me the issue uh I just I check with the log and all because uh previously during my practice I already created with the same name so what I did I uh changed my name over here rest of the process will be same nothing uh will be there so uh you can check out uh inside the log okay here is a log or else uh you will be getting the notification also right so uh it was giving me the issue that's why I changed the name and again I redeploy it with the same process which I shown you so after the deployment you will be able to find out this type of page okay now uh here you can see provision succeed and then uh here is what here is my endpoint URL the same one and is what here is my API key now what I can do I can keep this endpoint URL where inside my code and this key as well so let me copy the key and uh already I pasted this URL here you can see this one now let me keep the key as well got it so I have my key and I have my URL over here and along with that this is what this is my name okay as your model deployment name now one more thing I can show you over here see you can test it you can test it inside the uh Azure itself okay over the portal as well so let me write High over here so definitely it is going to be the answer now let me ask can you give me a price of Nike Air Max sh true so let's see what I'll be getting so if everything is perfect then it will generate the price yes I'm getting it and here is the price is what 150 USD now I was telling to you that if I want to connect it with the my application right I connect it with the flask and all so from where I will get the template so guys see just click on this consume after it here it will give you the each en detail see here is what here is the endpoint URL primary key secondary key now JavaScript python C s p is giving you all a sort of a languages okay four main languages and other uh thing also it supports so later on basically in one advance or R and all I will be showing you but here see for the consumption it is giving you four option now just click on this Python and this is the entire code so the entire reference I'm taking it I'm taking from here itself so try Okay the same uh from the the same code basically I pasted over there then accept Okay this exception the same code now here you can see the head so same header basically I WR written over there okay inside my code now everything's same and I took the reference from this template only now without wasting a time let's test our application so if everything is fine I'll be getting my answer so here I am going to stop it first of all and let me I'll do one thing let me stop my camera so that you can get access of the entire screen so here I'm simply writing python app do py now once I will write it down it guys so see uh I'll be getting my URL okay great now just click over here on this URL here is my chat Bo now simply write hi over here that's it so let's see perfect guys perfect perfect so here I'm getting my answer and it is in front of you only and now I can test right so here actually basically I can test with the different different question and all let's say here I'm going to be asked Can you tell me uh who is uh Gandhi so this is my question let's say I'm uh submitting it and see back in back end actually GPD model is working now it will provide me answer the request information not everyone the retrieve data okay so it is a giving me answer based on the retrieve data now let's see if I'm going to be asked the question can you tell me the price of the Nike ear Max shoe so here once I will submit it so I think I will be able to get it now here yes I'm getting it so see guys previously I was uh talking about uh previously basically I was asking this particular question but as a GPT model it was not giving the answer okay why because every information is trying to fetch from the uh document and is providing the answer so here what I will have to do tell me so here I will have to rectify my prompt okay whatever prompt in a back end uh like they are passing right so here in the prompt flow basically you will find out they're passing the prompt okay uh where you will get it you'll be getting inside the project itself uh let me show you where uh in the deployment this is my project okay prom flow this one prom flow and here is this two uh okay this one now just scroll down and you will be getting the prompt and all everything see so you will have to rectify Guys these particular prompt if you want to get your desired answer means uh with respect to the other question and all that's it nothing else so I hope uh you like this series guys I think uh I fulfilled my promise I created my own uh rag uh system okay I created my own chatbot and yes it will be able to provide you the answer with respect to each and every question in all you can ask about the nice sendal also so let's see whether it is working or not so once you will click on this submit and here you will be able to find out the detail means uh yes it is a patching from the document itself and based on that it is going to be argument the response so I hope uh you like this series and I started from the introduction and all but uh I kept uh the advanced part and all now uh yes if you liked it guys then please hit the like button please please mention in the comment section which part you like most if you have any doubt any question you can mention inside the comment section definitely I will look into it so thank you guys soon I'll be coming up with with uh many more uh different different courses and all so yeah uh please uh like subscribe the channel hit the like button uh if you liked it then yeah please share with your friends and whoever is required this type of tutorial and all so thank you bye-bye guys take care

Channel: Krish Naik

Views: 30,593

Keywords: yt:cc=on, generative ai tutorial, generative ai with Azure open ai services, azure open ai crash course

Id: 3SRh2nzN2DM

Length: 183min 16sec (10996 seconds)

Published: Thu May 30 2024