Use OpenAI's ChatGPT in Python

what is going on guys welcome back in this video today we're going to learn how to integrate and use chat GPT inside of python using the open AI package so let us get right into it [Music] all right so we're going to learn how to integrate and use chat gbt in Python today and this can be quite useful if you are aiming to build something more complex with the power of chat gbt so you want to use the chat completion you want to have this chat bot but you also maybe want to react to certain kinds of inputs or outputs you want to build an application around the power of chat gbt this is what this tutorial is for and the first thing we need to do is we need to install the open AI package which we're going to need to access the open AI API so we're going to open up the command line first and we're going to type pip install open AI now in addition to this package we're also going to need an API key so something that allows us to authenticate ourselves to the API so that we can send requests from our account and for this we're going to go to platform.openai.com I think I will have to move my camera here so that I can show you what I want to show you essentially if you have an account you just log in if you don't have an account you just create an account but if you have used chat gbt before you probably have an account and once you're here you just click on your profile in the upper right corner and then you click on view API keys and here if you have a secret key already you can just use it otherwise just create a new secret key copy it store it in a file and then you can load it into your script so that you can send requests to the API now one thing that's important to know is that the API requests cost money now it's very very cheap you can go to openai.com pricing and also before you click away because you don't want to spend money um I think that when you create a new account for openai you get some free credits uh in the beginning so if I click on chat here maybe uh what can I see that actually manage account or something like that there you go free trial usage I have 18 dollars for whatever reason I didn't spend any money on the API yet but you get some free credits maybe not 18 maybe five dollars or something like that but you will be able to send some requests without having to spend money um but you can see the prices here so for chat gbt which is not gpd4 it's this one chat and then you have GPT 3.5 turbo you spend 0.002 dollars per 1000 tokens uh meaning input and output um we also have the ability to use other language models but we're going to focus here on chat so we're going to use the chat completion which is also what you have here I also have a tutorial here a text-based tutorial um and as of right now this is the model that this is running on maybe quite soon you're going to also be able to use gpt4 but today we're going to have to use GPT 3.5 Turbo so uh let me just move back move my camera back up here what we're going to do now once we have the API key and once we have the open AI package installed is we're going to import open AI we're going to say API underscore key equals and you have to load the file that you're storing your key in so we're going to say open API key in my case because that's the file that I have here with my key inside of it I'm going to open it in reading mode and I'm going to read the key from that file now what we need to do now is we need to take that key and uh set it as the API key for open AI so we say open AI dot API underscore key equals API key and by doing that we now have our key our account associated with the package and when we send a request we send a request request through our profile so to get a very basic response to a simple prompt what we need to do is we need to say response equals and then we can just use openai dot chat completion and what we need to pass here is we need to pass two parameters the first one being the model that we want to use and we already talked about this the model is going to be GPT Dash 3.5 Dash Turbo and then we also need to have messages and messages doesn't have to be just one message it can be a message history and the message history has to have a certain format so what we do here is we say message equals and then we need to pass a list of dictionaries so I have this list here and inside of that list I have dictionaries and each dictionary is basically just a key value pair or contains two key value pairs namely the first one being the role so who said something and in this case we're going to be the user that says something and then we're going to have the second key value pair which is going to be the actual content and for example I could say um what is the difference between Celsius and Fahrenheit something like that and or maybe we could ask how do I convert from one to the other but that's now one message if I want to add a second message to the history I can say the same thing again roll and I think the roles are assistant user and system now user being us the the actual user who sends a request to chat GPT the assistant being chat GPT itself and the system is more like um the the system is more like context information like you are this and that so you can go actually to the tutorial here and you can see I think some examples um do we have some examples here system there you go so the system would say something like you are a helpful assistant that translates English to French and then the user would say translate the following sentence so system just gives the assistant uh information about their role um but for now we're just going to leave that one message here and in order to actually print the response we need to of course just print the response so there's nothing fancy about that when you call this um I don't know if that's a function or a Constructor but when you call this thing it actually already sends the request and you get a response now this might take some time depending on when you run this now we didn't get did I do something wrong oh I mean to say messages this is plural here uh we still get an empty we still get an empty response let me just check why that is oh okay sorry we need to say open AI dot chat completion dot create so this is just uh I think this is the class right yeah this is the class and this is the function that we need to use to actually create uh the request so now it should work and probably it will take some time there you go so what we get here as a response is um Celsius and Fahrenheit are two different units of temperature measurement the Celsius scale is used worldwide and is based on the metric system it measures temperature using degrees Celsius with the temperature and so on and so forth I don't want to read everything now but you can see here that the response is also structured here as a dictionary as a sort of Json object where you have the choices you have the message and you have some information about how many tokens uh were used so you can see here completion tokens prompt tokens total tokens so we now spend money for 226 tokens remember 1 000 tokens cost 0.002 dollars so 0.2 cents essentially [Music] um yeah so this is quite cheap we can now go ahead and build basically a chat system in Python by just running a loop and constantly adding to the message log how would we do that let's remove all of this here and let's say we have a chat log which is an empty list and we say while true user message equals input whatever the user wants to put in if the user message dot lower equals quit we're going to just break out of the loop and otherwise we're going to say chat log dot append and we're going to append a key value pair which is role user and content user message so this is going to be our message object or messages object essentially and what we do now again is we say response equals open AI dot chatcompletion dot create we see again the model is going to be GPT 3.5 Turbo and the messages are going to be the Chat lock um and then basically if you want to get the actual response if you just want to get the message instead of the whole Json object what we do is we say assistant response equals response choices zero message and then we want to get the actual content from the message now what's important is of course since we're talking to chat GPT and chat GPT is able to recognize context we need to also add the assistant response to the chat log otherwise we just have our own messages we don't have the responses of chat GPT so what we want to do first we want to print of course chat GPT says something it says assistant response and we need to strip it from New Line characters otherwise it may look funny then we're also going to strip it from White spaces so we're going to just get a text and then we're going to say Chat lock dot append and we want to append another key value pair here or another message with two key value pairs the role being the assistant and the content being the assistant response um and of course here we need to also say script the new line character and the white spaces there you go so that should already be it and I think I need to just enter a message right so if I say here uh what could I ask um what is the difference between and interpret it and a compiled language and then we should get a response with an explanation then I can ask which one of the two is python for example so I don't have to specify again that I'm looking for the difference between interpreted and compiled languages I can just say okay and just give me the information what of the two is python so here I got now interpreted language is a programming language this code is interpreted into machine code on the Fly and stuff like that so let's go ahead and say great which one is python then question mark and then it should be able to recognize the context because if I just ask without context what python is it will tell me basically python is a programming language it's used for I don't know what and in this case it tells me python is an interpreted language because it recognizes the context so that's basic chat GPT and the good thing about this API now is that you can use it to automate things now you can also just use chat GPT in the web you don't need an application like this but you can take this now and you can build functionality around it you can say for example if certain tokens um occur in the response or if certain uh stuff like that right so if you say user message lower equals quit you have the functionality to quit uh the program you can also add other commands you can also say okay if the response is containing certain words I'm going to censor it I'm going to call cause an action because of it I'm going to do something because of it um and this is how you can use chat GPT in Python to build more complex chat Bots or more complex chattings systems all right so that's it for today's video I hope you enjoyed it and hope you learned something if so let me know by hitting the like button and leaving a comment in the comment section down below and of course don't forget to subscribe to this Channel and hit the notification Bell to not miss a single future video for free other than that thank you much for watching see you in the next video and bye thank you [Music]

Channel: NeuralNine

Views: 16,414

Keywords: chatgpt, openai, python, python chatgpt, gpt3, gpt4, gpt-3.5, gpt-3, gpt-4, python gpt, python chatbot openai

Id: YVFWBJ1WVF8

Length: 13min 17sec (797 seconds)

Published: Sun Mar 19 2023