AUTO-GPT Setup | STABLE Release

hello and welcome back to another video on autogpt this video is an update to my auto GPT in under 5 Minutes video since posting that video the process for setting up Auto GPT has changed quite a bit one major change in the setup process since releasing that video is that the auto GPT team themselves recommends to use the stable branch and not the master Branch for setup also since releasing my previous video autogpt also introduced Pinecone which seemed to cause some issues as well especially if you're trying to run auto GPT in continuous mode although Pinecone is great for long-term memory it comes at a massive cost if we go to the pricing model of pine cone you will notice that the standard package for pine cone is around 70 dollars a month but I do have some good news for you one major benefit of using the stable release for otgpt is that pine cone is now optional and we can now decide to use a local memory cache instead which is free in this video I will show show you the correct way to set up the stable version of Auto GPT as for the prerequisites it goes without saying that the GitHub page for auto GPT is a valuable source of information since the project is changing all the time I highly recommend referencing the documentation if you get stuck as for the requirements we need python 3.10 or later installed I will leave a link to the python website in the description below but after clicking on that link you will see this page if you scroll down to the bottom you can go ahead and download the installer file that matches your operating system for me I'll install the windows 64-bit installer after downloading this file we can execute it and you should be prompted with a screen like this before clicking on install now make sure that you check this box that says add Python 3 to path if you don't do this you could run into all sorts of issues when trying to execute Python scripts or run perp install commands after checking this box go ahead and click on install now and run through the installation after installation completes you might be asked to restart your machine I also recommend installing git on your machine you can find git from this URL which will also leave in the description below from this page download the file that corresponds to your operating system and install git and then lastly you also need an open AI account so for this I recommend going to platform.openii.com and then creating your account creating an account with openai is free and you should be given some free credit of the registration note that calling apis to GPT is a paid service but you should be able to make some good progress with your free credit next we need to download otgpt to our local machine you basically have two options and this is something with the auto GPT documentation is not very clear on every video that I've seen on this topic simply tells you to copy this URL and to clone it to your local machine however there is an issue with this simply cloning this URL will make a copy of the master branch of this repository in other words the files found in this master Branch over a year the issue with this though is that the master branch is extremely unstable and this could be due to the open source Community constantly pushing changes to the master Branch so although this page is telling us to make a clone of this master Branch we can see this comment at the top of the repo page saying use stable and not the master branch and it also says the master Branch might often be in a broken State and this seems to be the cause of many issues that we run into so the first option is to click on this link I will also link to it in the description we can then scroll down to the bottom of this page to assets and you can then download any one of these two files I'll download the zip file you can now move this file to a folder on your machine and extract the contents after the fall was extracted you can go ahead and remove this Arc archive or zip file you can then drill into the folders that were created to access the otgpt files for the second option we can clone the repository and then use git to switch over to the stable Branch so what we can do is on the master Branch click on code and copy this URL then create a folder on your machine and then type in Powershell or CMD to open the command prompt in the command window type git clone and then that URL that we copied press enter Then you can CD into this newly created folder and you will notice that this folder now contains the files for auto GPT however this is based on the master branch and we need to switch over to the stable version instead so in the command prompt ensure that you are in this directory then type git switch and stable and press enter after doing that you will notice the files changing in this folder and these files are now reflecting the stable version of Auto GPT either one of these options will work now that we have downloaded the otgpt files we can now start with setup in the auto GPT folder open Powershell or the command prompt by typing CMD or Powershell then in order to install the python dependencies type in pep install slash R requirements.txt and press enter if you have experience in Python you are more than welcome to create a virtual environment first before running installation but we will not cover that as part of this video of installing the requirements we can now go back to our folder and we now need to make a copy of the dot EnV dot template file we then need to rename this file to dot EnV we then need to open this file in an editor so I'll right click this file and you're welcome to edit it in notepad but for the purpose of this video I'll use vs code I think the environment file is just a bit more readable in vs code in the environment video available fall there are quite a few parameters that can be set but the only variable that is mandatory is the open AI underscore API underscore key please cover it and populate our openai key in this field so what we need to do is go back to platform.openai.com and then click on personal then click on view API keys and from this page click on create new secret key you can give your key a name I will call it auto GPT and then click on create secret key then go ahead and copy this key and go back to your environment variable file in this file replace this text with your key and save we do not have to change anything else in this file but we will go through some of these options after this step but for now go back to your auto GPT folder and let's try to run our otgpt instance let's open up the command prompt again from the command prompt type in Python Dash M followed by Auto GPT to run this just press enter we will be prompted to provide a name for our AI this can be anything that you want I will call mine Auto GPT bot as for the role of the bot I will say an AI bot designed to teach me all there is to know about Auto GPD and I'll press enter we can now enter up to five goals for the AI to execute in sequence after the goals have been completed or the GPT will close automatically for the goals I will say for goal one go online and search or to GPT for gold 2 find the guitar page for auto GPD and figure out what the project is about then for goal 3 explain what author GPT is in a file called Auto gpt.txt then finally once all of these goals have been made we will just tell it to terminate the session I will press enter we do not have a goal of five so we can just leave it blank and press enter again so now it will try to execute the first goal and we can see a few things coming back from the AI so for its thoughts it's saying that it thinks that the first step is to search for auto GPT on Google to get a better understanding of what it is it's also giving us the reasoning for these thoughts it's giving us a plan of action as well as some criticisms so for speak if we used the speak argument this would convert the speak text to audio we will have a look at that in a few minutes but in order to execute this step it's asking us to confirm if this information is correct so I will type yes and in there so author GPT did a Google search has got some information for us and how it wants to write this file to our local system and it's asking us to proceed I will just say yes and enter now it's saying it's written the file successfully and it's saying that because it's reached its goals it's going to terminate the session and again it's asking us to approve this manually I'll say yes and enter as you can see it's saying that all the goals have been met and therefore it's shut down we can see the output of this run by going back to the folder and we can go into the the auto GPT workspace folder and we got this file here otgpt.txt which we asked it to create and this file also contains the information that we asked for so now that we have our basic setup working I do want to take you through some of the other arguments that are possible when running otgpt the first argument that I would like to cover is to speak command we can get Auto GPT to respond to us with a human sounding voice in one of two ways when we enter the command to run otgpt we can add an additional argument with Double Dash and we can then add speak let's run this welcome back darks would you like me to return to being Auto GPT bot as you might have heard what the GPT returned audio to us as well however we are getting a few error messages as well and the audio sounded extremely robotic so it's recommended to make use of a service like 11 labs for the text-to-speech so what we can do is go to the website beta.11labs.io I will leave a link in the description and go ahead and sign up for an account after signing in click on your profile image then click on profile on this screen click on this little eyeball to view the key and then go ahead and copy your key then back in your project files open up the EnV file and then scroll down until you see the section TTS provider or text-to-speech provider so in this list we need to provide our 11 Labs API key also we need to provide it with voice IDs if I go back to 11 labs for a second in your dashboard you will see the sample form where you can test out the different voices under settings this is drop down with all the different voices you can even sample them by clicking on play if you do what you've always done you'll get what you've always gotten the voice ID refers to the unique ID linked to each of these voices you can find the IDS on the auto GPT Gita bripper by scrolling down to the speech mode section here you can see a list of all of those name IDs so if we wanted the ID for say Josh we can copy this ID and in the EnV file replace this text with that ID you can also provide a voice to as a fallback so I will just grab any one of these keys and replace this with the name ID you can then save this file so back in our folder we can run this command Again by typing python slash M Auto GPT dash dash speak after running this The Voice should be coming from 11 labs and it should sound very different and better than the the standard voice let's press enter welcome back would you like me to return to being Auto GPT bot as you can hear the 11 laps API sounds so much more natural I will search for auto GPT on Google to learn more about it and as you can hear the voice is just so much better but at the moment we still have to press Y and confirm each of these steps manually so in order to get Auto GPT to really run autonomously we need to do the following we can cancel the execution of this bot by pressing Ctrl and C alternatively just close the prompt window and open it back up again if we want to run auto GPT in autonomous mode we can run python Dash in Auto GPT dash dash continuous if you wanted to add speak as well you can do that by giving a space between that argument and in typing dash dash speak but for this demo I'll just remove speak let's press enter when running in continuous mode we do get this warning which is understandable there is a cost involved in running these GPT API calls and running Auto GPT in continuous mode especially if it's not supervised it could really add to your bill or use up all your free credit so just be careful so it's asking us if we want to run this script again I will just say yes alternatively you could say no and just specify a brand new bot with a new role and new goals but for this demo I'll just say yes after pressing yes Auto GPT should run continuously without asking us to authorize any requests I also want to spend a few minutes just looking at the values in this dot EnV file and these are values that I think will resolve many of the issues that you've been experiencing in the fall we've got this value here called execute local commands I found this to be extremely valuable for projects where Auto GPT needs the ability to install packages by itself so if you run into issues with that just set this value to True another important value is the memory section so in order for audit GPT to keep track of its progress and its store state it needs to store the information in a database like Pinecone or readers but with a stable version of autogpt they've introduced a local memory cache basically what this does is it stores the progress and the state in a Json file if we go back to the folder for a second we can see this file here called otgpt This is a Json file if you open up this file in an editor again I'll just use Visual Studio code you can see all of this gibberish here and basically this is all the information and prompts that were generated by otgpt and this file is massive and for larger prompts with multiple goals this file gets extremely big and for the AI to fetch information from this file can become quite slow therefore for more extreme edge cases it's better to use an AI database like Pinecone or something like redis if we wanted to switch over to a point database we can simply replace local with something like Pinecone and we then need to provide the Pinecone API key as well as the point can region this is not something that we'll deal with in depth in this video but you can create your Pinecone account by going to pinecone.io and you can sign up for free after logging into pine cone you might be informed that you've been added to a waiting list so you just have to wait until your free account has been approved and then you can grab your API key and region details and then add them to this dot EnV file I will just change this back to local if you are using a redis database you will need to provide this information over here if for some reason you are having issues with auto GPT using Google to search for Content online you can create a Google API key and then replace this value with your Google API key in order to get a Google API key you need to go to console.cloud.google.com or leave a link in the description from here you can click on the available projects and then click on new project we can now give our project a name I will call it auto GPT and we can then click on create after creating the project go back to this drop down and click on auto GPT once auto GPT the project is selected click on credentials and then click on create credentials and select API key this will now generate the API key which you can copy and then back in the search provider section replace this text with your Google API key this should now resolve any Google search related issues that you've had otgpt should now be able to use the Google API to perform searches online and one more thing if you do not have access to gpt4 yet you can run auto GPT using the GPT 3.5 model as well you can do this by running python Dash M Auto GPT dash dash gpt3 only you should now see a message saying GPT 3.5 only mode is enabled this will work pretty well although it is a bit more restricted than gpt4 I really hope you found this video valuable and that you are and that you are now able to run your own instance of Auto GPT locally if you would like to see more content on auto GPT or AI related topics then please subscribe to my channel and like this video I'll see you in the next one bye bye

Channel: Leon van Zyl

Views: 24,677

Keywords: auto gpt, autonomous ai, gpt 4, artificial intelligence, chatgpt, ai, gpt 3, gpt 4 demo, chat gpt explained, auto gpt setup, auto gpt tutorial, auto gpt howto, agi, auto gtp agi, open ai, open ai chat gpt, pinecone, auto gpt github, pinecode setup, gpt4, gpt-4, auto gpt4

Id: sfpaj4DTcWo

Length: 17min 2sec (1022 seconds)

Published: Mon Apr 17 2023