How to Setup AutoGen Studio: Unleash Powerful AI Agents for Automation

hello everyone and welcome to the RPA Champions channel the place where we dive deep into Automation and explore how it can make our lives easier and our work more efficient today we are going to be looking at autogen Studio from Microsoft this is a amazing technology that is going to help us save countless hours of time by automating different gpts for us to do some amazing automations and perform some really cool and as a final note this is a very important skill to have in this coming times as AI is becoming more and more of a necessity for everybody to use and also a necessity for many businesses to have inside of their Solutions so knowing how to automate and automate different gpts and be ahead of the curve it is going to provide you with a strong competitive Advantage therefore keep this in mind let's jump and let's start with installing autogen studio autogen is a library that is released by Microsoft it is a library that is going to help you have different or multi-agent conversations so it is going to help you orchestrate the conversation between different gpts that are talking to each other with somebody that is managing them that is another GPT it also easily integrates with different applications and also allows you to use different models for different kinds of tasks and to build your own skills we're going to see all of this however this is available on GitHub it is python based it is a library that you can write code however on top of this there is a no code low code solution meaning that we're going to be using a studio to Rapid prototype the different solutions that we want to create also before diving into autogen it is good to get the understanding with a no code low code or a studio solution that is going to help us visualize these different functions with different buttons and allow us to create quickly different solutions also if we don't know how to code now out autogen Studio it is very very simple to set up to set up autogen on your computer it is going to be very easy the first thing that you have to do is set up a couple of dependencies so you need to set up python or you need to install python it is as simple as going to the python web page and downloading the installation file on for Windows or for Mac or whatever system you are on the next thing that you are going to need and that is going to make your life much easier it is not a necessity but it is going to make everything a lot easier is Visual Studio code this is going to allow you to edit the little bit of code that we are going to write and need to input while setting this up and also from configuring it later in the future this is just going to allow you to read and visualize the code nicer so I do advise it however it is not a necessity to set up you can go also forward without setting up visual studio code that is all of the different Necessities or requirements that we need and the other thing that we are going to do is create a virtual environment why do we need to create a virtual environment and what is a virtual environment now a virtual environment since this aogen studio is going to be installing different applications on our computer based on the different requirements so for example if it needs to install a library to run a python script it is going to install that inside of our virtual environment so we don't want it to install on our machine that's why we're setting up a virtual environment this step you can also skip it but I strongly suggest that you set up a virtual environment here they show us how to set up a cond virtual environment however we are not going to use cond we're just going to use a python virtual environment now I have asked chat GPT to set up a virtual environment for me and this is the kind of code that it has given me so this is the command that you need to run after installing python inside of the directory where you're going to be installing autogen so the first thing that you have to do is create empty folder in this case I have created the folder called autogen this folder was completely empty now once you have created this folder the next thing that you're going to do is you're going to open this folder from Visual Studio you're going to go to visual studio and open the autogen folder that you have created and after that you are going to run and visualize the terminal so we're going to be inserting different commands inside of the terminal and this is going to be the terminal where we're going to be setting up our environment and everything else now the first thing that we need to do is set up a python en environment a new virtual environment we're going to do this with this code right here now if this code doesn't work remove the Python 3 or check how it has been installed on your system it can be just py python or with Python 3 after that after you run this command you need to activate your virtual environment meaning that you need to start and you're going to be using this code right here once you use this code right here you're going to have a visualize something like this meaning that you are currently working in your virtual environment and this is where we are going to install autogen to install autogen there is a just a simple command which is PIP install aogen and that's pretty much it all the installation process is done after that to start autogen we will just run autogen UI inside of our terminal right here and the process would start there is one more thing that we have to do there are some agents that are going to require a internal API key to function so we are going to set up a API key for open AI now you need an open AI open AI API key to use aogen because it's going to be connecting to the open AI GPT 4 however you can also use local models like Ama or any other local models that you would like to connect to autogen another comment that I would like to mention is that for open AI we would need to set up a command that is like this and copy and paste this inside of the terminal so that we can have the uh API key Set uh set up and working now once all of this is done we can do autogen UI and launch the autogen studio that we have installed on our computer now to access autogen we would click on this link that is appearing here in the terminal or in the browser just visit this link right here and this is what it is going to look like it is going to look like without with just a few skills and a few things this are all different skills that I have created but it's going to be structured like this we are going to have a build a playground and a gallery and skills model agents and workflows how does all of this work and how is all of this going to allow us to automate and save countless hours of times this is super exciting and I'm really excited about this technology because the potential of this and currently also the existing use cases are amazing so in the scale sections this is where we're going to create python functions that generate or solve different tasks now for example like create a flight search so what is this so this is a API call to Skyscanner to get the uh cheapest flights in US dollars for a specific destination let's check out something else so for example I get YouTube transcript this is a script that gets the transcript of a YouTube video so what are basically skills skills are different python functions in this syntax right here that are going to help us solve different tasks now I know said I said low code no code and this seems like code however what we can do is we can take this code right here and go to GPT and ask it to create different functions for us I have already uh done it a couple of times uh and create a few different functions and it seems to work with a little bit of fine tuning therefore so this is for example one of the functions that I have created it needed a little bit of fine-tuning but it is a great start and what I mean by fine tuning is just pasting the message back to GPT and getting the response so once we create a scale after that we can use our scale inside of our agents now our agents what are the agents the agents are different agents or different employees or different users that are tasks with specific actions so for example we have a strategic advisor here a strategic advisor we have a system message that puts this into context that puts this strategic advisor into context and tells it what it can do we can also add different skills to this advisor so for example like you're a strategic advisor that you're specialized in doing certain things and you can also use different skills that we have created all of the different skills that we have created right here in this example right here it doesn't use a skill however it what it does it passes information between the different agents inside of a workflow to get the best answer for you so let's see another another for example I get flight prices so this is a AI assistant that uses a skill that is create flight search and it gets the data from the flight information so that it can pass it to another adviser that is going to create a custom travel plan for you we can also see a viral content creator this is the latest one that I have used this is using two different skills it is using a fetch trending topic skill that is making a API call to Google Trends and then is taking the most trending keyword and is creating an outline for that keyword before passing it to the copywriter Pro the copywriter Pro is going to take this sample outline and then with a very strong and detailed prompt it is going to create a blog outline an article outline just the way that it has been described here all of these assistants we can see that they are using different models we can add whatever kind of model that we want in this example I'm using gp4 preview because it is the simplest one to get started with if you go to the models part you can add a new model in this example I have the default model is this and the only thing that you're going to need is the API key of your model once you paste your API key you should test the model and make sure that it's working after that all of your agents will can use that model that you have created or different models that could be specific for different tasks or for different information sources that you analyzing there are also a couple of uh default agents so for example if you create a new agent a default sample assistant is going to be created this sample assistant has a very good prompt that allows it to use and install code on the virtual environment that we are working on allowing it to create and use different applications with python once we have created a new agent and once we have added the skills and the model that it needs set up the system message and also added the human input mode meaning never meaning that it never is going to communicate with the user that means that it is not going to ask for input from the user it is just going to communicate with other chatbots once we have configured all of this now this is when the magic happens this is our workflows our workflows are like chat group where we can add all of the different agents in a conversation and have them talk to each other to try to figure out what makes the most sense and what we should do for to resolve the user prompt or the task that the user has provided now remember with ch GPT is all about prompting and asking for different information or inside here it is all about tasks and completing different activities for us so let's see one of these that has been already created so General agent workflow this is a default workflow we can see that that to configure the workflow we just need the workflow name the description and this summary method last none or LM we also have the most important part which is the sender and the receiver so who is sending the message who is receiving the message so in the sender for example we have this robot right here that called a user proxy and this is just a simple robot that doesn't have any model or any skill that all is doing is just passing the message from the user and it's passing it to the primary assistant now the primary assistant what it does it has its prompt right here it is able to do all of these things it is taking from the user proxy The Prompt and the information and it's trying to uh solve it using the gp4 model and using these different skills that have been added by default so this skill right here it is a skill to find pap papers on uh ax I and the other one is to generate images now you can't use these skills without replicating this uh this assistant the primary assistant because they're not in your skills but if you create a primary assistant you can use the skills and also add your own skills that you have created so that you can amplify its functionality So currently this General agent workflow has one one robot here which is a primary assistant with two skills and already with this it can do a lot of things let's put it to the test and see what we can do with just the general agent workflow that is a default workflow that is already provided after installing it so how would we test all of the different agents that we create we would go into the playground and create a new session from the new session we would select the different workflow that we want so General agent workflow and let's click on create let's try this prompt right here I want to create a presentation in PDF format about the five sweetest fruits from the Amazon and let's run this now keep in mind that once you run this it is going to take a little bit of time to get your answer why because different agents are interacting do you want to see what is happening behind the screen in that case you would open up your Microsoft Visual Studio and in here in the visual studio you are going to see what is going to be happening so right now we send this prompt right here to the uh primary assistant the primary assistant to the user proxy to create a presentation about five Suites from the Amazon we will need the following these steps research the identity the five uh sweetest fruits from the Amazon generate images for each one create a presentation with the information and images of the fruits convert the presentation into a PDF format am amazing and so on and so on so it is creating a file uh file python file that is going to generate images and it is continuing in with the execution and with the processing of all of the different steps so the primary assistant has identified a work plan for the different assistant to follow and right now it is going to iter it through its plan completing all of the different steps that it finds we can fine-tune inside of the configuration we can fine-tune the number of messages and we can set different limits and all of the different configuration now keep in mind that the studio version meaning this graphical interface has some limitations as it is not directly Cod and you cannot F tune every single thing however for us to get started and to create already amazing workflows and test out this new technology it is great another observation that I wanted to pass is that all of the technologies that we are using are Microsoft Technologies they seem to be on top of things in terms of Automation and where things are going also on my channel I visit a lot or I create a lot of content about Microsoft power automate desktop and Cloud flow which are tools used for manual and repetitive automation but when we combine Ai and generative AI into the mix this whole process becomes much more scarier in terms that we can automate endtoend processes meaning that we can automate the cognitive part and we can automate the repetitive part as well so we can see that it's actually installing different libraries like python pptx it has installed and if we go into our libraries so this is the virtual environment that we have created and we see all of the different scripts so we go into the libraries and we see all of the different libraries that this has installed on our computer it is pretty amazing so it is installing the different libraries also from the previous prompts that I have sent it is trying to create a file it is having some issues but no matter what sometimes it resolves these issues by itself just by creating another file and iterating over it now we could also give it all kinds of commands I have created all kinds of different workflows for all kind of different things that give way way much better answers than if were just to talk with GPT I have done a experiment where I have previously created chpt to create me a brochure for uh to download for example like what are the five biggest fish or the biggest fish in the ocean and it has created me a PDF document however the output of the document doesn't nowhere resembles the quality of the document that comes out from a iteration or a back and forward for different agents now our agent has completed I believe and this is what it has created so it has created different files for us it has created a PowerPoint presentation and it has also generated an image for each one of the different fruits from the Amazon rainforest isn't that amazing we're able to accomplish all of this with just a simple prompt that was I want to create a presentation in PDF format about the five sweetest fruits from the Amazon now these eight results we can investigate them further but I want to show you that there has been 18 messages sent in here in three minutes of time so it starts off with the user proxy user passing this message to the primary assistant the primary assistant then creates a plan it starts creating a script that is going to do different things the user proxy then talks again with the primary assistant telling it that the code is having or the image was saved right here and continues going on with the different fruits one has installed all of different fruits after that it installs a library on my computer that allows me to create PowerPoint presentations after that it creates some information that is going to put inside of the PowerPoint and it outputs all of this information to me you can also try with the general agent workflow you can also try the different buttons right here that are going to give you some pre-built prompts that you can already try so for example paint a picture of a glass of Ethiopian coffee freshly brewed in a tall glass cup and a table right in front of a lash green forestry scenery now this again it is going to call the assistant user proxy and the other agent and the information is going to be passed between them however let's take a look once this is finished at a more complex workflow where we can create different skills and put the together those skills to have a workflow that is going to solve information for us this has already created a a image of a coffee in a beautiful C capup with a very Lush background we can also see the script that and prompt that was used to create this file now if we go to the gallery now once we have created one of these libraries or one of these sessions that we like we can choose to publish it so that we can see it and use it and share it and visualize it later for reference and to know what is working for the application that we are going to be building with this kind of technology or solution so in here I have already used this on different examples so for example right here we have a uh request that that is a request to go to Tokyo however I cannot see any more the details of the messages so as we can see uh it is taking my simple request message it is breaking it down it is doing different things such as identifying the departure airport IDE code for Milan and for Tokyo it's putting the dates accordingly it is searching on flight scanner it is creating a uh a gun for me to do and so on so the potential of these Technologies are countless also if you would like to see me create a step by step process or where we are going to create by building a skill by building a agent and a workflow please let me know the exact use case or example that you would like to create and I'll make sure that I created thank you so much for watching my video and if you've sticked together with me until the end I wanted to thank you and to highlight how important it is that you're learning the skills and being a Pioneer in generative AI in AI Engineering in flow Creation in automation but overall in intelligent automation so give yourself a pet on the back because you rule you really do you are going to be a new leader in this AI world so keep on learning these things and give yourself a high five thank you so much for watching and have a great day

Channel: RPA Champion

Views: 2,316

Keywords: AutoGen Studio setup, AI automation, productivity tools, AI agents tutorial, automation guide, tech tutorials, content automation, AutoGen Studio agents, software automation, programming automation, AI for productivity, AutoGen tutorial, how to use AutoGen, AutoGen Studio examples, automation software tutorial, AI technology, AutoGen Studio basics, learn AutoGen Studio, coding automation, AutoGen Studio walkthrough, streamline tasks with AI, autogen studio, autogen

Id: e8TWZSbXEAc

Length: 23min 28sec (1408 seconds)

Published: Wed Feb 14 2024