AutoGPT tutorial: Build your personal assistant WITHOUT code (Via Relevance AI)

Key Takeaways

how do you create your own Auto GPT app that can do things for you without coding if you don't know what Auto GPD is it's basically a Next Level AI assistant that is as smart as GPT but can have access to a wide range of different tools like Google external docs and many many different things I just give a tutorial about das.dt which is no code platform that allow you to create auto GPD type of apps and if you haven't watched it highly recommend but today I discovered another platform called relevance AI which is a no code platform that allows you to build powerful Auto GPT apps it has a lot of unique features like allow you to upload a PDF file send HTTP requests and a bit more visual no code building experience so I'm going to teach you how can you build an auto GPT app with relevance AI let's get it let's say I want to write a code outrage email to Mark Cuban for Android investing the way I would do that is firstly I will want to go to LinkedIn take a quick look about Mark Cuban's background and then our Google about the latest news about Mark Cuban find the article that can help me understand his latest perspective and thesis about certain area so that I can use this to write a warm intro email to do this in relevance AI is fairly straightforward once you get in you will say starting page like this they do provide a few use case out of the box for example allow you to upload PDF and then Auto structure database or summarize insights from certain interview transfers but what we need is build AI chains so let's click on this and then we will add a new chain for our own use case I want to firstly start searching about Mark Cuban and his latest news our click on search web for LM context so this is the main interface of reference Ai and there are two parts on the right side is where you can build a simple web app interface where you can ask users to put in certain information in our case it will be the name of the person and then the goal of code and on the left side where you review the actual AI chain firstly we will use a block called make API request to search on Google about Mark Cuban and return the latest news and we will use a service called serper which give high quality API results for Google search in server you can use a playground to configure the ones that you need so for example in our case we only want to get the news and on the right side I can see the parameters I will need to pass on so I'll just copy those things here and in the body this is where we were adding those parameters so that it should return the results we want we are using double Curry brackets to refer to param stock query which is referring to this fields in this case are putting Mark Cuban and the inside header that's where you will replace with your own API key you can get from the server by clicking copy and come back here to replace once you did this let's click on run notebook button on the right top corner to see if we got results okay so in here you can see it actually start returning good results I can start seeing a few options but not all the results here are relevant I actually want to just search about Mark Cuban relevant to AI so I might add a new field here on the right side called topic and so this is where I can put in specific type of news I want to search about Market but on the left side I were adding this field error.top so it will search for both smart Cuban and Ai and let's run this again okay now you can see the results is much more relevant and what we want to do is we just want to extract the news list here so that we can pass on to the GPT and let it choose which article is the most relevant one our insert a JS code transformation which is a block for me to run JS code and in here I just wanted to return the news which is this one and remove all the other irrelevant information so what I would do is create a constant Snips steps dot API call output response body which is through results of the first step step API call steps.api Core as it is referring to the output.response body this is the content that's why it's quite long and what I want is just news so I will just dot news and should just return me the list of news let's see okay great here we can see in this field returns both title and also link which we can use to scrap the data and I want to pass on this list of news to gbt and let's choose the best news teaching can help me understand Mark Cuban's perspective about AI so I will give it a prompt firstly I will insert the results from the last step which is steps to our news list and the output.transformed so you can basically follow the naming convention like this is the name of the object and this is for you to choose output this should insert a list of news here and then above the list of news about parents doctor people which refer to Mark Cuban here and please help me choose the best news that can help me understand Mark Cuban's perspective and thesis about premise.topic which is AI please return only the link of the news do not include title source or anything else only the link because I just wanted to return me this link so that I can use it to scrap data so let me run this notebook again but before you do this you also need to put in your open AI API key and you will do this by clicking on the button on the left side where you can add in the API key thank you all right remember to click on the save button guys this app that do not Auto say okay I'm going to do this again from scratch all right let's run again great now if you click on output or answer this is a response GPT give us which is exactly just a link of a article called Mark Cuban on impact of AI and GPD let's create the next step we will need to do is script the information about this website and what you can do is you will add a make API request and we will use a browser last to scrape the data and if you don't know how to do the script you can actually create some examples so here example all I need to do just to copy the URL and I will need to use browser last which is a service that can help you scrape the data of certain websites or go to my account change the token to be my own it should a post and inside the body it will pass to elements once URL which is URL of news that we want to script and also element selector body this means we try to get a content of the body which is normally the part where they contain the news we need to refer back to the last step and we can change this to be the news link so it's a bit easier I want you to see the results so we need to use steps.use Link and output dot answer perhaps the news link dot output dot answer and yeah this should work um and but on the other side for this example you can see they actually provide you a way to round multiple times which means if I want to script the data of all the news in return I can also do that but for the purpose of this demo I just want to script one of the news and let's run this again okay it Returns the results very quickly inside the text this is where it will return the majority part of the content and what we want is we want to just grab text here and pass it on to GPT asking it to summarize for us okay it's kind of annoying that every time I change the name it will erase the content from the last patch so I have to run again because I couldn't remember the structure we will need to use a JS code transformation again to write JS code that can extract just this part and it's pretty straightforward I just create a constant text equal to step.use.text Output response body and then our to this to extract the text itself next we're gonna pass the news content to GPT and ask it to summarize for us so our putting the steps.news.com10 output transform which is a news content that we extracted and give it a prompt above the article where Mark Cuban is included and I want to understand Mark Cuban's perspective about AI through this article above please summarize it to help me understand Mark Cuban's perspective and thesis about AI the last step is I want to pass on the summary of this news to GPT and ask it to write a outrage email for me based on those context and this is prompt I got you are some of the best code email marketing expert in the world it's super helpful to give GPD a row so it will behave like that way I want to write an average email to Mark Cuban and below is the article where Mark Cuban is included I will insert the article summary that we just extracted and below is a really good prompt about writing emails and I didn't come up with this it's something I learned from another YouTuber called Jordan and he has some great video about sales please help me write an auto Rich email to Mark Cuban about the goal I have which I need to put in here as I will try to get Angel investment from Mark Cuban about a AI startup called Jasper I need to get a quick summary about what Jasper is this is actually what another tool is quite useful also by dust called xp1 it allow me to use GPT for the content on the specific tabs I'm opening so in this case I will ask Eva nice and good about aspect once I get this our back here so the goal is again Joe that's for my AI stockhold Jasper low is okay all right I think we have everything now please help me write an outrage email to Mark Cuban about the goal and please ensure to follow all the rules listed don't skip any of them this is a good mechanism you can use to making sure PPT follows specific requirements you have so rule one write a cash flow not to ensue that's the compliment about them it needs to be specific can only be set to Market root 2. don't be cringe and write in the voice of some of the best code email marketing expert in the world and try to include numbers and metrics that can grab his attention rule 4 keep it short keep it short keep it short and Rule five again don't be cringe not overly enthusiastic let's try this okay we got our email I hope this email found you well I recently came across your Source on AI chatbot and their potential impact on the job market your insights on the importance of ethical consideration in the department of use AI resonate with me this is really good it's very contextual and referred to the things that he just recently talked about now we have our app and one unique thing that relevancy I provide is you can actually deploy in two ways you can use it like normal API but they also provide a shareable link so I don't need to worry about deployment I can just run this app like normal I have this URL so people can just start using app like this I have attached this link of code email gbt in the description you can try it out if you want it will generate email below so this is a quick intro about relevance AI as I mentioned they have some unique features like upload PDF or make HTTP requests which are Super handy they also have a list of features about Vector search which can be quite powerful when you deal with large amount of content that we haven't explored today so go and create some interesting AI apps and comment below about use case you are exploring