n8n, AI Agents and Workflow Automation training

1. Introduction: Hi, everyone. My name is Marko Bukowski, and I want to thank you for choosing this training. If you are curious about AI workflow automations, if you wonder how can they save you time by reducing some repetitive work you might have there, or even how can you build and sell AI solutions to other businesses, then you are in the right place. Don't worry. Theres no previous experience required. We will learn everything step by step here. You don't need any software development background or, in fact, any IT related experience to start learning ATN and automations today. Believe me, you are in the perfect place. The only thing you need really is a laptop, or PC, and some spare time. This training, you will learn how to run N ATM as a native Cloud solution or locally on your laptop, how to build automations, how to build AI agents all step by step, as I said, how can you connect your favorite apps and services together that can save you hours and hours of time? You will also see how to build AI solutions that you can even sell to other businesses. And you might be thinking now, Mark, how can I sell anything to other businesses if I don't even know what workflow automation is or how to build an AI agent? But believe me, you can. I'm really glad that you made that decision to start educating yourself in that direction by choosing this course. You have to just give yourself a chance. That's what it is. All you need to do really is start today and remember, the sooner you begin, the better. You'll already be ahead of 99.9% of people who have never even heard of NATM or workflow automation. Might have heard Nvidia CEO say that AI will create more millionaires in the next five years than the entire Internet did in over 20 years and there are many reasons to believe that is true. This training will give you all the skills that you need to understand how to build and run advanced AI solutions in NA ten and also how to start selling your work and your solutions to other businesses. But always remember, you don't have to take it too seriously. Treat this as an after hours fun. Maybe just start it as a little side hustle creating some automations for yourself and then maybe selling some of those solutions to your local businesses, just to see if this could be a career you'd like to pursue. You can't go wrong really. At the very least, you'll stay up to date with the most recent technology. But who knows? If you start making 1,000 or $2,000 a month as a side hustle, something you do after work, maybe at that point, you'll decide to make it your main source of income and do it full time. Who knows? That, of course, it's your choice. With NA ten support for AI agents, this really is the platform to go. It's no surprise its popularity has exploded recently, and I'm not surprised at all that you want to learn more about it. By the way, if you want to learn more about some related topics like AWS, Linux, and Docker, computer networking, and many moe, then check out our Automation Avenue platform where you will find hours and hours of very useful IT related materials. But going back to this training, we will have a project to complete and as your product work, you will build your own AI agent with its own large language model, its own memory, and instructions. You will see that completing this project is actually much easier than you might imagine.'s waste no time then. I will see you in the next episode where we'll prepare and run ITN software together. 2. 3 ways to self-host n8n AI Automation software: Before we dive in, please remember to hit the follow button. This is the best way to ensure you never miss new classes or important updates. Since the technology constantly evolves, I frequently add new material to keep you completely up to date. Don, are you sure? Okay, great. Let's start the training. In this video, we will just focus on the installation part. NA ten can be run either free of charge or as a paid service. In this video, you will learn three different ways to run NA ten on your local machine so you can use it free of charge. But first, I want to just make you aware of that paid option. If you go to that pricing tab, those paid options are shown here, and those are the prices of using NA ten in their cloud. With the Cloud version, you don't have to worry about installation, servers or configuration. Everything is there ready for you to use. The downside is that it costs money each month, you want to use this platform. But the great thing about NA ten is that it's open source product, which means you can pull entire code and install it on your machine instead of using their paid Cloud solution. You can see that Github icon in the top right corner. If you click on it, you can basically see entire code that is used to run NAM software. And because it's an open source, there are many ways you can install it yourself. Running it on your server is called self hosting, and that means you can use NA ten free of charge. So let's do that. Let's get started. We will go through three different methods to get it installed, and you only need to pick one of them. First two methods are Docker container based, and they are very similar to first method is where we run a simple Docker Run command and the second way, we will build a Docker Compose Yamofle and we will run it using Docker Compose app command. And the third method, we will install it on Proxmox server with just one command using so called community script or helperscript, it's also called. You can use official documentation from NAten. You can find a lot of information about installation and configuration here on the docenten dot IO. Here is the Docker, for example, and we will use some of the commands from here to have it up and running. All right, so first method using Locker Run, this one. This is my Ubuntu system, so it's a Linux based. Let me make it bigger. And you can run Docker containers on any operating system really. Here on Linux, you only need to install Docker, for example. But if you want to run them on Windows or Mac, they also let you run Docker containers, but you need to install additional software like Docker Desktop. So you can simply Google how to install Docker Desktop or MAC or how to install desktop on Windows, because it's not covered here, but just wanted to let you know that you can run Docker containers on Windows Or MAC as well. Here on my Ubuntu, I don't need Docker Desktop. Linux can run Docker containers natively. I just need to install Docker. But before I do, I run that command. It's a Sudo UG update and pudo UGT upgrade. And then I add Y. It's to simply answer yes to any questions that might come up during this process. So I press Enter, I need Sudo password. And that command is not Docker related, but you should always run it to have your system up to date before you do anything. But if I want to run Docker, I say Docker. You can see Command Docker is not found, but it can be installed with any of those, and I'm not really a fan of Snap, so I will just use this command. So do up to install docker dot IO. Just copy paste. And that's it. Well, I didn't put Y, so I have to answer yes. I might take a while depends on your Internet speed. But now we should have Docker up and running. So if I run system CTL status Docker, I can see that it's indeed up and running for 12 seconds. Now, next command is optional, but if I don't run it, I would have to run Docker with Sudo, always, like sudo Docker, blah, blah, blah. I don't want to do that, so I can run this command instead. I run sudo user mode, AG it's lowercase A and G then Docker. And dollar sign user. That means current user, whoever I am will be added to Docker group. I press Enter, and then to make it live, I run a new GRP Docker. It refreshes this group like you can also log off and log on, but this is easier option. Just run that new GRP Docker. Now if I run Docker command, you can see it gives me all options. I can run Docker Tag, Docker Stop, docker RM, et cetera. Maybe let me clear that to test Docker, there is a really nice command. It's Docker run Hello World. If I run that command and I get the output, that means everything works correctly. Docker can pull the images and all process works as expected. So let's press Enter. It says unable to find image locally. That's true because I never run this command before, but it was able to pull this image. We can see it here, pull complete, and it says, Hello from Docker. That means I can be sure that Docker works as expected. This is a really good test. That's basically it. We can now go to NAN and follow There's actually just two commands we have to run. They say, we have to create a volume first, that Nate N data because we will use that Docker managed volume. So that's fine. I will copy that. Maybe let me clear first. I will paste it. And the volume should be created. I can confirm running Docker volume LS command. And indeed, I can see that local volume called NAN data has been created. Next, we just run this long command, Docker run ITRM name. Basically what it is. You run it interactively. This is about how to stop the Docker. This is the name of the container port it's running on, and the volume we're using is that NTN data volume that we have just created. And if we go further, this is the image it will pull from doer dotn dot website. All right. So let's just copy. Copy, Paste. Again, unable to find locally, of course, because I've never run this command. This is fresh operating system. Nothing is installed here, and again, it might take a while. The process is now completed. You can see you can press O to open in browser or you can simply copy this link, local host on port 5678. So I just copy it. I will paste it in my browser and go. And that's it. That's my NA ten. That's the process Well nearly completed. We just have to set up the account. I mean, that's not the part of the installation, but maybe let's do that because I created like a test email. You can see NA ten test email at protonmil.com, basically just for that. So we can use it here. Merck put your name or whatever. You create password, and you just click Next. I will say it what describes my company, a business owner or myself. Doesn't really matter, get started. And now you have that little button. Send me a free license key. I mean, you can skip that, but you will have some extra options added if you put that license key. So let's do that. I will click that. Your license key is on the way. So I go back to my inbox and it should be sharply hopefully here. All right, that my free NATM license key. So I can activate it. You can do it like that, or you can simply copy this key. Go to your NATM, go to that usage and plan, and you can paste it here, Enter activation key. Activate license activated. You registered community addition has been successfully activated. And you can see it here. You are on community addition and you are registered. Well, that was like extra information, not part of the installation, yes. If we go back to terminal, you can see a new message, license successfully activated. And if I control C here, it's a stopping NA ten, and now my website basically doesn't work anymore because I stopped the service. Running it like that, I don't really like it the Docker Run command. This is not doesn't feel right for me. I like to run Docker images at least in Docker Compose file. Let's create it will be the second version, second way of running NA ten. We will create a Docker compose file. I'm currently in my home directory, home I can use any text editor I want. You can use even Notepad on Windows or text edit on Mac. Here on Linux, I've got, for example, Nano and I create file called Docker Compose dot yaml. 1 second. Let me just clear that. Nano Docker Compose Yamal. I can use that information from their website. I can use basically it. I can use this command and convert it to Yamel. We have some ports, we've got the volume, we've got the Docker image. But to save some time, let me just show you something, I will remove it. And this is it. This is what it looks like the Docker Compose yam. I don't want to both with the details how to translate that Docker Run to compose. But basically, that's it. That's my Docker compose file. I press Control O, enter Control X, at least on the MC to save this file. If I run cat Docker Compose, you can see that's what it looks like. Now, the only important thing, I have to be in the same folder where this Docker compose file is located. So if I run LSL, I can see I am here in my home directory. This is my home directory, and this is where I have my Docker compose file. There is one more thing. If I run that cut command again, the volume, I wanted to match it to what we already had previously. We run that Docker Run command, yes? And we created already this volume. It's called NA ten data. I mean, I can run this command again, it doesn't help. So basically, I can run it now, you should run it before you run Docker Compose really because these volumes NA ten data, this external kind of expects this volume to be already created. So you need to run this command if you haven't already and make sure running Docker volume S that indeed this volume is already here. And that's cool. However, if I run Docker Compose, it says it's not found because Docker compose, it's a separate component. It's not part of Docker. It's like separate component, but we know it can be installed with psudoU to install Docker Compose. Let's copy that. I will paste it. Again, psudoPassor. Yes. And now Docker Compose has been installed, which means now, let me maybe clear that again. I can simply run Docker Compose up the IpressEner and it created some network. It created the NA ten container. And if I run Docker PS, I can see that it indeed has been created. It's running for 12 seconds again, and I can connect to it on this first port, which is 5678. I mean, let me just cut it first. Basically, whatever you have here, I mean, you can't change this port. This is simply what application runs on. But if you want, you can change this port on the left side. So whatever I have here, I simply connect to it on the local host. So I go back again, HTTP, Local host 5678 and defect decanter. You can see that it takes me straight to this landing page. It doesn't take me to this account creation. Why does it do that? Because I matched this volume. This volume was already created. That means it doesn't matter if I run Docker Run or if I run Docker Compose D, we write simply to the same volume to the same storage. Whatever I saved when I run Docker Run, I can also read from that location when I run Docker Compose a D because I use the same volume. I hope that makes sense. And if I want to stop it, I can say, Docker compose. Stop. If you run Docker PSA, it will show us this container, but its status is exited, and then I can also remove it by running. Let me clear Docker compose down. The good thing, though, is it will not remove this volume. As you can see, this storage is still there, and all my configuration, if I refresh here, well, we stopped the process, yes, but I can simply run, again, that Docker compose up DRD, And if I go again to the local host, it will get me straight back to where I was before. That's why I like it. All right, but let's just stop it. Docker Compose down. Whoops. Down. It's now stopped. And let's go to the third option, how to install it the NA ten platform as an LAX container on Proxmox. This is my Proxmox and the fact is that the Ubuntu we worked on is also running on that Proxmox. It was this Ubuntu server with idea of two oh six, where we were running that Docker commands. But now I want to run it as standalone instance here on the Proxmox server. So what I can do, I can simply Google Proxmox helper scripts, and just click on the very top link, and we can see that view script button. And here, I can just search for Naten you can see it in Internet of Things and Smart Home section. So let's click on that. That's the NATN and all you have to do is to copy this link. You get this little message. Be careful when copying scripts from the Internet. Always remember to check the source. That's a good advice, and you can check the source by clicking this button, star on Github. It will open new tab and you will actually see what is the code and what it does. But I have already checked that, so I can just copy this, and then I go to my Px Mx, to PVE or whatever your node is called, to the shell, and I simply paste this command here. And when you presenter, it will take you through that interactive process. The simplest possible way is to just use that default settings. It's really nothing to show here. If I press Enter, it will just go through this process. But we will go through it again. I don't like some of the options that are chosen here, so I will show you how it's fully automated. I mean, it's running right now, but then I will show you how to change some of the options during the installation if you want. Like, for example, here, disk size 6 gigabytes or RM size 2,048 megabytes. It's maybe okay if you just start with NA ten. But if you want to store some additional files or maybe, let's say, run some additional databases that you want to connect to this NA ten service and you want to run it on the same Alexi container, then maybe you want to change those values, and I will do it later on. Right now, I will just wait until this setup finishes, and I will show you how to change those values if that's what you want to do. And if you wonder what are those little dots here, they are so called tags, and that's also not what I want to have. For example, this yellow is this sort of blue is automation, this green is community script tag. This is not important. It's just if you're curious. Let's wait for this process to complete. And now this fully automated process is completed. We can see the URL, it's different IP address it doesn't show us local host because it's a standalone Linux container, like C container. We can copy that. We paste it in our browser, and we should still be able to access NA ten. But as you can see, now, it takes you to that account creation page because simply that LCC has its own storage. It's a different storage than we used before. That's why you would have to go through the process from the beginning. You can see the resources here, memory, swap, course, and root disc is different. But I want to show, let me just remove it. Let's just shut down this container first and here in more, drop down when you remove because I don't need this container. I will remove everything. And let's go back to that shell. Do I have it still in the clipboard? That's past? No, no. Is the IP address. So let's copy this long comment again from the helper scripts. Go back. I will paste it again, run it. But this time, I will choose advanced settings. Number three, option number three, I press Enter, and now it tells me that to make a selection, Spacebar, okay? That's fine. Press Enter. Unprivileged container. That's what I want. Yes. Root password. Yes, I need that to be able to access this container, verify. Basically type it again. Container ID. I want maybe 215, let's say, cost name NA ten, that's fine. Disk size, not six, but maybe 100 might be overkill, but yeah, that's okay. CPU course, two courses is enough, yes, I would say. Ram, little bit more, 4,096. Okay, network bridge default, VMBR zero. That's fine. And the static IP, I want to use my own 192.168.1.215. Let's enter. Sorry. You have to type 192 dot one, 68 dot one, 215, forward slash 24. You have to provide subdt mask as well. It's fine. Gateway IP is the IP address of my router 192.168.1.1. I mean, these are my changes. You could leave it as DHTP and it will work fine. It will generate DIP for you like router will assign IP to this container, but I want to have it manually created. That's why I typed IP address and then default gateway as well. App casher leave blank, disable IPV six. Yes, MTU default 1,500 is fine. Search domain blank and DNS server IP, I will use 1111 because I use manual configuration, I don't rely on the DHCP, so that's why I typed this. Macaddress leave blank, V and blank, custom tags, these are these tags. Remember those two dots that were here, blue and green. I don't need those tags. I will remove them. You can use tab to move between those fields, okay? Now press Enter. SSH key for root, I could paste it here, but I already have password. It's good enough. Enable root SSH access? Yes. Enable fuse support? Yes, because I might want to use Arklon for that as well to synchronize some data. So I say yes. And enable verbose mode, I say, yes, this is for the installation process. It will show me more information. What is going on in the background. I press, ready to create? Yes. And the conflict file, no, I'm not interested. And now, the installation process starts again, but you can see those options here are now different. The container ID is 215, my IP address, these last digits, they match the container ID because that's what I actually not for containers only, but for all of those virtual machines and containers, if I see this virtual machine has container ID of two oh six, I know the IP address for that is 192.168.1.206. This is my way of doing things. I don't say it's better or worse. It's whatever you choose. But now I have disk size of 100 gig and I have four gig of Ram. And this is that verbos information I was talking about. So you get more info what is currently being pulled during this installation process. But we just have to wait again anyways, so maybe simply maybe it makes it less boring. And it's now also completed, but we can see the IP is the one that I specified and if I just copy that, if I go to the URL, we are indeed again on the account creation website. But if I go to the specs of this NATNGaxy container, we can see I have now four gig of memory, I've got 100 gigs of storage, et cetera, which means if I want to manually overwrite some settings, I can simply run that option number three at the very beginning. 3. AI Agents and Workflow Automations explained: What does the term workflow mean exactly? Or what is automation or an AI agent? These recently became very popular, but they are also very confusing terms. You might even hear something called AI workflow automation that combines all three of those concepts. The good thing is that they are actually very easy to explain and you will understand all of that in the next few minutes. We will talk about each of them, and then we'll see how the automation platforms like make Zapier or very popular recently ATM platform can help us build those AI workflow automations. Let's start with the term workflow and automation. Fact is that workflows and automations are not even the IT specific things. They've been with us for ages and workflow workflow is simply a sequence of steps that you have to follow one after another to get a particular task done. It might be surprising, but you already use some workflows in your daily life. Think about your morning routine, I don't know, brushing your teeth. Yes? What is the workflow for that task? What is the list of steps that you have to follow to have your teeth clean? Well, you need to first grab a toothbrush. You need to squeeze some toothpaste on it. Then you brush the teeth, spit, rinse, and then you put the toothbrush back. That's the workflow for this particular task. These are the steps that you have to follow in the exact order to have your teeth clean. Let's say maybe you drive to workers, what's the workflow for that task to get done? First, you need to grab the car keys, then you have to go and get into your car. Then you drive to work and you get out of the car. That's the workflow for drive to work task. With just another list of steps that you have to follow to get this task completed. A workflow is simply a list or a sequence of steps that have to be followed to complete a specific task. That's it. What about automation then? If you think about it, there are many workflows that are very easy to automate. Again, it does not need to be about IT industry. Think about all those manufacturing and automotive workflows that have been automated over time just because they were so repetitive. For example, if a CNC machine needs to produce 10,000 identical parts, let's say, it is worth to spend some time maybe at the very beginning and create a program that will follow automatically, it will follow these steps required to make that one part and that CNC machine then can be pre loaded with the tools and we keep repeating that sequence 10,000 times until the entire job until this entire batch of 10,000 elements is completed. Basically that machine will go on and on and on without or with very little supervision and intervention needed from a human being. So again, what is that automation then? The automation is that single workflow, that's a single task that is done automatic way, the way that needs none or very little human supervision. Again, as you can see, automation is nothing new. Automation has been with us for ages. But to automate any task, you first need to understand what the workflow looks like for this particular task. What is the sequence of steps you have to follow to get it completed? That's why you often hear that phrase workflow automation because that's exactly what it is. A workflow automation is that workflow, that sequence of steps that is executed in an automated way. Now let's maybe talk about IT industry because there are many workflows that are very easy to automate. What can be automated Loads of things, you might run some automated backup scripts every day or there are some scripts running every day on the servers. Even at home, you might have some automated way of backing up your photo library, let's say, if you work in IT, you might actually get paid to build some automation. Say, I don't know, your boss says that every time they get email, when the email arrives, they want to get some kind of notification on their Whatsapp application that they have installed on their mobile phone. As always, to get this job completed, you first need to understand and define what the workflow for that would look like because only then you can start automating it. What that workflow would look like, those steps that you have to follow. First, you have to know if there is a new email in the inbox or not. The first step is some kind of check that is able to connect to the boss's email inbox and check if there is an email or no. If there is no email, then there is nothing to do. But if there is an email, maybe you want to grab the subject of that email and that subject of the email will be then forwarded to Whatsapp application on their mobile phone. To make that simple automation, you have some platforms like NA ten or M or Zapier, and in fact, in previous video, we saw how we can install NA ten platform so we can self host it and use it free of charge. There you can choose a trigger because we always need some kind of trigger and this trigger will do some checks every minute. Let's say every minute, this automation will check given inbox to see if there is a new email or no, this step would do nothing if there is no new email, but if there is a new email, we would need to extract the subject line of that email and extracting that email subject field would be our second step of our workflow. Then once we have that email subject, we would need another third step to send this subject. This kind of processed information already and that processed information has to be sent to Was application. We can use so called web hooks, for example, or other whatsapp integrations to be able to send that message. And if you can run that workflow in automated way, you have workflow automation. By now, I hope it's clear what is a workflow and what is the automation. But what about AI agents? What is the AI agent and what exactly is that AI workflow automation? The AI workflow automation is a type of automation that includes a human like reasoning step in the workflow, and this step is very often handled by so called AI agent. But before we talk about those AI agents, let me give you an example, simple example where we would need that AI agent to be used as part of our workflow automation process. Let's say we still work on that same request from our boss regarding the email forwarding to their WhatsAp application. But this time, instead of just forwarding the subject line, bos asked us to send the summary of that email to WhatsAp application. Now the workflow will look slightly different. Well, the first step can be left the same because we still need to check if there is a new email or no. That trigger that checks email inbox every minute can be left as it was. But then once we have new email, we need to read that email and create a summary of what this email is about. With standard automation, we can maybe only pull that subject line of the email or we could forward entire email, but our workflow currently can't understand and summarize what that email says. Because to do that, we a human like reasoning element. We need that AI agent backed up by some LLM, large language model, I mean, that is able to read that email and generate a summary of it. That is where that human like reasoning comes into the workflow and that's what it turns it into that AI powered workflow automation. Basically, AI agent is that component in our workflow that can handle some more complex tasks. When I say more complex, I mean where human reasoning is required. Those AI agents effectively replace that human part of the process. If we create our workflow that way, we are also able to keep that workflow automated because it does not require human intervention. It still can be done automated way. That platform like NA ten, we can connect that AI agent there to any large language model that we want, GPT, cloud, deep Sk, blah, blah, whatever you can then also give them some memory so this AI agent can remember past steps or past tasks, and you can equip them also with some tools like calculators like APIs, and many other tools. These AI agents can be extremely versatile and they can be used for many tasks that are way beyond our current very simple example. Even in this example, I mean, it wouldn't make much sense to employ a human being to just read the email when it arrives and then forward the summary to our boss. Our boss might not even be happy with that, that somebody reads their emails. But it makes much more sense to employ that programmatic AI agent inside of our workflow. It becomes one step in our workflow, and that AI agent can be then used to create that summary every time a new email arrives. And those AI agents are becoming extremely popular, really. They're already being used in some stock trading. They are already making business decisions or they read CT scans or X ray they can do, much more. That's why I think it's super useful to know how you can employ those AI agents to help you with any workflows or maybe build workflows for companies to automate their tasks. 4. Your first AI Agent in 20 minutes in n8n!: If you have no idea what an AI agent is or how to build one, then you are in the right place. We are going to build our first AI agent from scratch, and you will see that it's very easy actually. You don't need any previous experience to start building and playing with that technology. You might be wondering first what an AI agent actually is, and I will explain that shortly as well. But to show you what it is and what it does, you have to first understand that an AI agent is a system that has to run somewhere. Need some kind of automation platform where you can host and run those AI agents. For our first AI agent, we are going to use NATN automation platform. But please note that NATN platform has both free and paid options for paid option, you simply need to sign on to the NATN platform and you can run your AI agents just there. Well, it will cost you money each month that you want to play with that NA ten platform in the Cloud. The great thing about NAT N platform though, is that you can just download it and you can run it locally on your laptop, PC, mini PC, if you do it this way, then NATen is completely free for you to use. NAN is a very powerful tool that lets you build all types of automations and AI agents and what NA ten is exactly and how to run it free of charge. We already covered that in the previous video. You might want to watch that one to get your platform first set up because in this material, we will focus only on that AI agent component available in NA ten. I mean, it's not NA ten specific thing. There are other platforms that will let you create AI agents, but we will concentrate today how to build AI agents in NA ten. This is my NA ten platform. It's running on my mini PC at home, and here you can see what the main page of that NA ten software looks like. Even before we start building anything here, if you, for example, went for paid option, hosted by NATen in the Cloud, you will go to NA ten is and you can actually here, see what this AI agent looks like. And each AI agent consists of four main components. You can see that chat model here, and that chat model can be a large language model like chat GPT, Cloud, Gemini, et cetera. This is simply the model that you can attach to your AI agent that will act as its brain. Then you can attach memory, and there are multiple types of memories, and then you can see those tools. But there is also one hidden component, it's the system prop. Which provides instructions to the AI agent. But let's go back. What is that AI agent then? This AI agent is a human like reasoning system that uses large language model like that HGPT I mentioned or Claude, and that model will power the AI agent so it can make, for example, some decisions or it can understand some stuff, and it can also remember because it has that memory, and then it can take some actions based on all that information it. This memory for our agent, it can remember previous conversations or tasks, so the AI agent can take the decision based on that history that is kept in this memory. When we talk about tools, these tools expand the agent's abilities even further because tools let the agent interact with external services and you connect to those services using so called APIs, and we will also talk about it, so don't worry about it. At this stage, you don't have to worry about any of that. In fact, we will configure such API access to a service called Open weather Map and to connect to that service, you need something called API key. I will also show you how to get that API. Want to know what an API key is, it's like a very long password. It's simply a long string of characters like letters and numbers, et cetera, and this string can be used by the application. This AI agent then can access other services or other external applications. Bear in mind that for some services, you might have to pay to get that API key. But there are other services that can give you that API key free of charge. We will get the free one actually because openweather will provide API key free of charge, don't worry about it. Simply think of an API key as a user and password combined. It's one long string of numbers and letters, and that string is used to securely access other applications. But that's enough about API keys. Let's talk about that fourth component, the system prompt. System prompt is a set of clear instructions for our AI agent, and it can tell our agent, for example, what is its role here, what kind of output we expect, what it should and what it should not do for us. And sometimes you can add even more information like some default values or some limitations or some further instructions. You will see it shortly, but I think it will be easier to explain all of that during the agent build process than just talking about it. So let's build that AI agent, yes. First thing, I will go back, I can close this. I can go back to my NA ten platform and I just create workflow. Each workflow workflow means some steps that you have to follow to get the task completed simply. Each workflow has to start with so called trigger. I mean, if you click that plus, you'll have some triggers to choose from. The trigger I'm going to use now is the chat message. It's really nice trigger because you will get input and output directly here in NA ten. You will see what I mean. If we go back to Canvas here, we can see that trigger now. Let's move it may be there, and then the next component can be already our AI agent. I click that plus. I say AI and I will choose that generic AI agent service. You can see it says connected chat trigger node. The left side is that trigger, and it can already recognize that it's connected to chat trigger node, and the prompt is user message. This is a JSON code. You will see what it is for later on. If I expand it, you can see right now it's just JSON chat input, but you will see what it is for when we run. I will close it. I will go back to Canvas. That's basically it. This is our AI agent. But we can also see it's not happy, it's red because it doesn't have any chat model attached. It doesn't have memory attached or tools attached. Memory and tools are actually optional, but chat model, you need to have something attached because currently it doesn't have a brain, the most important component. That chat model, that large language model I was talking about, that's the first external service that we are going to connect to. If I click that plus, it will give me, there are many preconfigured models I can choose from, you might recognize some of them like grog, Deepsk, Google Gemini. But if you just start your journey with A ten platform, I would suggest you go for that open Router chat model. If you configure that open outer, you will basically have access to all other models. Let me show you. Let's click that Open Router and then you have to do, you have to click that dropdown menu, or you will already see that create new credential. That credential is that API key I was talking about. But how to get the API key? You have to go to Openouter website. It's hdpsopouter.ai. And once you configure your account, you would have to go to credit Although I believe there are some free options as well, but to be able to access all available tools, you have to add some credit. As you can see, I added $8 here five days ago, and they still have $6.93 available. Will charge you per 1 million tokens. It's not like it doesn't cost much really. I used it quite a lot and in five days, I used $1. Once you've got that credit, I think $5 is the lowest you can go. I'm not sure. But you can basically then go to API keys and you just create API key. I will call mine NATN and you can also configure credit limit, but I will just leave blank. I will create new key, and that's my key. I'll just copy it, go back here, and I will paste it there. And this key is for open Router account, so I save it. You should get that green response, connection tested successfully. That means the API key works as expected. I can now close it, and from now on, I can access all available LLMs in that open router, which is a lot of them. You have open AI like DPT. We've got perplexity. We've got some Q or Rakai I've never even heard of. But you can see it's loads and loads of them, yes. By default, it's GPT for 1 minute, which is fine. We can leave it as it is. And this is the brain of your AI agent. So you can go back to Canvas. We can move things around here. You can just have a better view, but we already have something like a chat bot. Because this when chat message received, this is nice trigger because here below, that's where you can type your prompt. I will say just Hi, how are you? You can see agent works connected to open Router chat model, you can see the input input was from this trigger, and in the trigger, you can see that chat input. Hi, how are you? That was my question. And the output is the AI agent saying, Hello, I'm doing great. Thanks for asking. How can I assist you today? That's how you can use any LLM model you want because you can change it here. Sorry, not here. You have to go to open Router yes and you can change LLM and test other responses from other models. But I want to show you something else. Right now, if I tell it, hi, I'm Mark. You can see the input, higher Mark, and the output. Hello Mark. How can I assist you today? If I ask it, what is my name? I'm not sure what your name is. Why does it say? It doesn't remember anything. Why doesn't it remember? Because we have to add some memory so it can remember the previous conversation. By the way, you can see here the tokens used. You can see the response time, which is 1.9 second and you can see how many tokens were used for this query. As I said, they usually charge you per 1 million tokens, let's say $5 per million tokens. The longer the queries, the more tokens you will be using. Yes. This is pretty important field if you want to use it a lot. That's it. But I was talking about memory. Let's click that little plus then. I will add maybe that simple memory is the simplest as it can be because it's stored directly here on an 810 platform, and it asks you for content window length, which means how many past iterations the model receives as context. Or in other words, how many questions it will remember. Yes, five is more than enough for now, I can go back now. Now I say hi, I am Mark. Response is the same. Hi, Mark. How can I assist you? Now when I ask, what is my name, The output is now your name is Mark. How can I help you further? You can also see that it stores two iterations already in that simple memory. And it referred to that previous conversation. I had to look up that history and it found it, and it knows now that my name is Mark. Simple memory is just this. It is very simple to configure, and it remember numbers of previous interactions. If for some reason, it doesn't work, it doesn't refer to the memory, you might want to hover over this node. They are called nodes. Yes, all of them are called nodes. But you can hover over. You've got three little dots here. You just click on them. You might have yours deactivated. So if I had it deactivated, it won't be able to access it. So I would have to activate, yes. But that's just in case it doesn't work as expected. You might want to check this. And the AI agent became yellow because we changed something, so we would have to it doesn't really matter. You know, Yellow means there were some outside changes. So we've got now large language model as a brain. We've got some memory. Now, add a tool. I thought about the tool to use here. I was thinking, if I want to have a conversation with this agent, maybe I want to find out what's the current weather somewhere. I can use that external tool, which gives me the accurate weather forecast. If we search for weather, we can see that open weather map tool. Let's choose it and here you will also have to create new credential. Ignore this. This is my previous account. It's shown here. For you, it will probably show that create new credential. You would click on you also have to paste your access token. How do you get access token? You go to openweathermap.org. You create your account here, and I believe the first API key is already included in the welcome email. But even if you didn't get one, you can simply generate another one here. I will call it NATM, generate and maybe's remove that one. This is my API key. I will just copy it, go back to my Natan, and I paste it here. My account is called Open weather map account. I will add access token, I will save it. Now tool description, we can live as it is. Operation current weather, that's fine. Format, you can change if you use different format like Imperial or scientific for me, metric is fine. Location selection, I don't change anything here, but in the city, you've got that little icon here. If you hover over, it says let the model define this parameter. That's what I need. When I click on that, you can see change to defined automatically by the model. And the model, it means that LLM model, large language model. What that basically means, if I ask my AI agent, what's the weather like in London, the AI agent will pass the city name, which is London, will pass it to this tool, and weather specifically for London will be retrieved from here and returned back to my AI agent. So now I can go back to Canvas. And now we've got this tool configured. But there is one more thing I mentioned, and it's that system instruction, that prompt for AI agent because it doesn't really understand well what it's supposed to do here. I mean, if I ask a basic question, yes, it will respond, let's say, what is the capital of Portugal? As you can see, it's connected to LAM. I checked the memory, but it probably didn't use it because it didn't need to. I didn't see why it would connect to open weathermp because there was no question about weather yes, it just answers the question. The capital of Portugal is Lisbon. But I want to tell my agent that if it's a generic question, this is fine. But if it's a question about the weather, I wanted to contact this tool, open weather map. I want to pull the most recent information about the weather for particular location, and I just simply want to see the output here. How do I do that? I will just double click on that agent. Now we will change the source for prompt. If you click that dropdown menu, we will change it from here to define below. But even before we do that, just make note of what it takes as an input now is the JSON chat input. So let me change it to the define below. Then let me make it bigger. I already have something. Let me just copy paste. Copy past it, and then this little icon, I will make it full screen. And this is what I pasted. The instruction for AI agent is, you are a helpful assistant that can chat with users and also check the weather when needed. If the user asks something casual, like, Hi, how are you? Respond naturally. But if the user asks about the weather in a particular location, then call the Open weather map tool with that city name and only call that tool if the weather information is clearly requested. Yes. We don't want to call it if it's not needed. But what it's missing here is another line. I have to instruct it to take whatever is also in that chat message. Whatever the user input because currently it will not know what comes in from the previous step from that trigger. So I will have to instruct it, take the input from user, and now I will use that curly bracket. It says Anything inside those curly brackets is JavaScript. That's called JSON JavaScript Object notation. If I do that, I can now take that chat input. Previously, it was what is the capital of Portugal is. Whatever I type in is in this object here, chat input. I drag it here so basically what I'm saying is take the input from user, whatever was typed in in the chat message, take that in and then my AI agent will evaluate that input, and it will take a decision based on that. And if I ask for something casual, it will just respond to me. But if in that input, I asked about weather, it will use, hopefully, open Weathermp. Let's see if it works. I can close it now. I can see it here. And if I go back to Canvas, let's see. What's the weather like in London? Presenter. You can see it contacts Salam, check the memory, but there was nothing to find, and it responds. The weather in London is currently mild with scattered clouds, which means it understands the prompt. If I ask it basic question, it will not use this open weather map tool because, you know, there is no need for that. But if I ask about weather, that's when it finds this tool handy and it will use it as per instruction. Just wanted to add if for some reason that open weather map doesn't work for you, it not necessarily means that you did something wrong. Because when you create new API or new account, it can take up to 2 hours before your API key is ready to use. If you have some authorization problems or response like server is not available, then you might simply want to wait until your API key is active. It's not only for open weather maap, it's for other services as well. Bear it in mind. This is one of the simplest AI agents that you can build, but think what you can do next. Maybe you don't need weather information. Maybe you want an AI agent that keeps checking a stock market for you. Maybe you could change this message trigger to something else. For example, we can use schedule trigger instead of data message chat. It can run every minute, let's say, and then you can connect your AI agent using API key to a service that provides current stock exchange information. You can then add another tool or even continue working with your workflow. Maybe you want to send a WhatsApp message or an email to yours when a given share price drops below a certain level. Or maybe you want an NI agent that summarizes emails for you. You can set a trigger for every email that you get, and then you tell AI agent to send the summary of that email to your Whatsapp, so you only open the mailbox when you receive something important. What I mean is that the possibilities are endless here. You can start building more advanced automation even today. And as you can see, there are no particular skills needed. You just need to set up the platform and you can start building your AI solution. 5. Please follow me :): One last quick reminder. If you found this training helpful, please click that follow button right now. This is the only way to ensure that you never miss future classes or important updates I add to this course. Plus, it lets you easily track your questions and discussions. Thank you and I hope to see you in the next class. 6. Build n8n Telegram bot to access all LLM's (step-by-step): This is a telegram bot that is running on my phone that can access any large language model I want. It connects to my NA ten platform, and let me show you how it works. If I type, for example, use Deepsk, I will get an answer from Deepsk. If I say Use Groc, I will get answer from Grock and I can get answer from GPD five, Gemini, whatever I want. This bot has basically access to any available language model that you could think of, I will show you how to build that all and don't worry. You don't need any previous experience. With NA ten, it's easier than you think. Let's get started then. But first, we need that telegram app install this Telegram is a chat program, something like a Whatsapp or messenger that you can install on your phone, and it is bound to your telephone number. You just go to Google Play or Appstore and you install telegram on your phone, just like any other app. And once you have it installed on your phone, you can also authenticate this app on other devices like your laptop or PC, which means that you can run it on any device that you want. Once we have that telegram up and running, we need to create a chat bot in that telegram application so we can use it to connect to our AI agents. You have to go to boot father. It's at Bt father, you have to type, and when this window opens, you have several options here and you need that new boat. You can see, create a new boat. That's exactly what we need. I click on that and it says, New boat. How do you want to call it? You can call it anything you want. I will call it Automation Avenue boat. That's it. Now, you have to choose the username for your boat and it must end with both. It's either bot or underscore both as per the examples. I will try to copy paste it, but note that it has to be not in use currently by anybody else. Let's check if that one is available. Oh, yes, it is. That's it. That's my bot created already. What's the most important here is your API token. You've got here the API token that we will be able to use in NA ten to bind NA ten with this telegram application. Note that this API key is very important. You don't want to share it with anybody. I mean, I am showing this to you, but in fact, I will remove this bot as soon as I complete this video. But remember, this is something like a password for you and you need to keep it safe. But honestly, that's all we need here. We're already done with telegram. Next step is to run our NA ten automation platform somewhere. And one of the options to run NATen is to use their own NA ten Cloud solution, but that cost quite a lot of money every month, and to be honest, it has quite a few limitations. That's why you might consider using Hostinger instead. Hostinger has recently become a very popular option for self hosting NATen because they offer a VPs that comes preconfigured with everything you need to start creating your automations. Just go to hostinger.com slash Automation Avenue. And it will take you directly to the landing page for that self hosted NA ten virtual private server. You simply scroll down a little bit, and you will see all available VPS plans. This is where you choose the size of your server. And again, the KVM two is one of the most popular ones because this is really more than enough for NA ten to run, but it also has a very good price. It has two CPUs. It has eight gigram hundred gig SSD drive. That's really plenty to just host NA ten platform. Choose how long you will need this server for, you can pick one, 12 or 24 months. The longer the period, the lower the price. For two years, it's a 699 a month. And they gave me a code that you can use, and the code is automation Avenue. Once you apply that code, you will get 10% extra off that price. Whatever the price is, you will get 10% extra. You just continue and you have to register. If you don't have an account already, I mean, I will create a new account and then it will get you to this page where you choose a location for your server. By default, the closest location will be chosen. But if you want to build a server on the other side of the world, that's up to you. You can change it here. But I will stick to United Kingdom. It's fine. So I click next. Here, you can choose either plain OS. You can choose Santos, Debian, et cetera. But what we are interested in is the application. You can choose here, for example, vanila NA ten. That's actually what I'm interested in. But you can maybe want to go for the one that has already applied 100 workflows or maybe so called NA ten in Q mode. But as I said, I'm fine with just NA ten. That's what I need. I select it. I create a password, so I can log on to my server if I want. It's for SSH connection, and then you can choose if you want to have Malwascammer included. I will leave it as it is, which means I can click Finish setup now. And it says it will take about 5 minutes, but to be honest, it didn't take even 5 minutes. It took much less than that. But you just wait for that process to complete, or you can read the documentation. Bottom in the meantime if you want. But that's basically it. Now you can click that manage VPS which is Virtual Private server, and this is it. This is your server up and running. And if you click that manage Up button, it will take you straight to your NA ten application. It couldn't get any easier. But what I wanted to show you is a very important thing because what Hostinger gives you as well, is so called fully qualified domain name. And that's very important. Especially when you want to connect external resources like Telegram or Google or any other resource that want to contact your instance, your Natan instance from outside. That is usually problematic if you host it locally on your laptop or PC. You'd have to use CloudFlare tunnel or Aroc tunnel, but here you don't have to do that because Hostinger not only gave you public IP address, but they also gave you that fully qualified domain name, which can be used to interact with other services. And in fact, you can use that domain to connect to your service. Let's say you don't want to click that Manage. Maybe you are on another PC and you still want to connect to your NA ten application. So you just copy that address. You open a browser and you type HDDPs forwards forwards ten dot, and you past that address, you just copy it. Presenter, and that also takes you to your NA ten application. Because it's hosted in the Cloud in Hostinger Cloud, you can access it from anywhere you want. Because I open it for the first time, it will ask me to set up owner account. So you type email address, your name, last name, and you create a password. Just click Next, fill in a form, whatever, that takes you straight to your application. This is your NA ten platform up and running. Here you can create workflows. That's basically it. Personally, I don't like that light theme, so you can go to settings to personal and you can change that theme to dark theme. I mean, that is at least my preference. And we are ready to build our solution now. These are all three tabs opened in another browser. This is Opera browser, but we are at the same stage. That's my telegram, that's my hostinger, that's my NA ten, and let's connect the telegram to my NA ten. I need to click that plus. The first step is always so called trigger, and I need that telegram to be my trigger. I click the telegram and I say on message. If you check that web hook URL, you can see that subdomain is used for this connection. That's what I meant. HTTPS, N Aten, SRV. I mean, you don't have to do anything here, it's just to point out. This is very useful feature that you get with Hostinger server. And the credentials to connect with, we don't have any yet. I need to create new credential to connect Aten to my telegram, but the botfaer already created that HDDP API I can use. I just copy that, I copy selected text and paste this access token here. And I save it, and it should give me or that's what we need. Credential successfully created and connection tested successfully. And it's optional, but let me change this credential. I will call it I don't know telegram NA ten, maybe. I know that's the API that connects telegram to NA ten, and I save it. That's all I need here. I can close it now. And if I want to test this step, I can run that execute step. I can see the telegram is listening for test event telegram trigger, I mean, which means I can go back to my telegram. I can now use my bot. It says you will find it at blah, blah, blah. You can click that little link. No messages here yet, of course, because it's a new bot, I can start it and now even I don't have to type anything yet, this already should trigger something that forward slash start, starting the bot. If I go back, that's what it is. I can see it's a bot command and the text inserted was forward slash start. But I can test it further. I will click that execute step again. The telegram trigger within ANA ten now listens to any messages coming to the telegram itself, which means if I go to that telegram and say, I don't know, hi, how are you? And if we go back, now we can see the text here changed to hi how are you? The other important thing is that chat ID. We will use it later on. And if you're curious, what the heck mark is that date? This is something called epoch time. If you copy that, you can run something like epoch time translation. I don't know. EPO converter O. Placed it here timestamp to human date. It is indeed today's date and the hour up to the second. But it's called epoch time. Doesn't really matter because we will not use this one. It's just if you're curious why it's in that weird format. But that means our telegram trigger works as expected, which means also we can go back to Canvas and we can continue our build. Let me maybe move it to the left. And now let's configure the first AI agent. I click on that plus and I say AI agent and see it here. We will first build a simple solution with just one large language model, and once we've got that up and running, we will add more and more large language models. We have to simply understand what is going on under the hood. And the source for prompt, that's the first thing we want to change it here because it's not connected chat trigger node. It's chosen by default. We want define below. We want to define it ourselves. It's simply the information for a agent, what it is supposed to do here in this workflow. I will write a simple prompt, maybe something like take the output from and from where? Well, on the left side, we have the previous node. It's a telegram trigger, and we can see its output. And honestly, the only output I'm interested in is that text. Hi, how are you? Because we are going to send some text in our telegram message, and I'm not interested in anything else. I just want to pass that message to my AI agent. Whatever is on the left is the previous. Here is the current node, the AI agent, and on the right, you've got output, which is the processed data, nothing here because there is nothing processed. But once AI agent processes the data, it will go here on the right as an output. I want to grab that text from the telegram. I will just grab it here. Take the output from this field. Whatever is in this field, if we switch to JSOM, it's JSOM under message, under text. Whatever is here, let's go back to Schema, maybe. Whatever is in that box, in that field, and respond back. Couldn't get any simpler really. The full instruction is to just grab this text and respond back to that text. That's all we need to set here. I will now go back to Canvas, but we can see the AA agent is not happy because it doesn't have any chat model attached. Let's attach the chat model then. I click on that plus, and here you can use different anthropic, Azure, deep Sik, many LLM models, but the easiest way to access all models, not specific ones like Grock or Gemini, you can go to open Router hat Model, and open Router provides you API access to nearly every single LLM model that is available in the world, really. But that means you need an account with open Router. You just have to go to openouter.ai, create your account, and you immediately have access to some freely accessible LLM models, but not all models are available for free, of course. If you want to use the paid ones as well, then you have to go to your account to credit, and you have to add some money to your account. As you can see, I've got $6.50, and in two months, I used only $1.50. It's really a great way to access many, many LLM models from one single source. But what I need to do here, I have to go to API key. And they have to create new API key. I just call it let's say telegram to NA ten, maybe. You can set the limit for that API key. Let's say if you reach $2, that API key will stop working, but I will not do that, but it's up to you. I will just create it. And once you've got that API created, you have to copy it immediately because it will not show up again. You would have to create yet another API key. But now I've got that API key, so I can go back. And in this open outer chat model, I click that credential to connect with, I create new credential, and I can paste my API key. I will change the name again, maybe telegram also to NA ten, so I know what it is about, and I save it. I can see connection tested successfully, that's all we need. I can close it now. Now you can choose any model you want. If you have some credits, if you don't have credits, you can still choose Olama or there are models available free of charge, but with some credits available, you can access all of them. And have a look at that. The list is a nerve ending really, Deeps Google Gemini, infirmatic. You've got meta, ma, Microsoft, Mistral AI, Mistral is really good by the way. Got morph nur Open AI, GPT models. You've got multiple models. By default, 4.1 mini is chosen, but if you want, you can also use GPT five or GPT five chat or maybe GPTive Mini. Let's have a look further. Q and blah, blah, hundreds. I don't know how many of them are here. It's loads and loads of them. I will use GPT five Mini, but that's basically it. That's all we need here. I can go back now to Canvas. Now the agent is happy because it has now open router chat model. Well, one more thing, I will double click. Maybe I will change the name to GPT five maybe, so I know what I chose. Now we go back. I know it's open router, but we chose GPT five. One more thing we can add here is a memory. If I click on that little plus here next to memory, I can choose a simple memory. This is simply for my AI agent to remember the previous conversation. Maybe I want to refer to the previous conversation and ask some further questions. This way, using simple memory, I will be able to do so. Default is five previous conversations, but this JSON session ID, it doesn't exist here. I want to change it to define below. Basically, that's all I need to remember, you know, just the text window. You will simply remember those fields, what I was talking about previously, or what questions I asked previously. That's all it needs, and it's green now, which means it's valid. We can go back now. That's basically it. That's all we need for our agent to work with a single large language model. Let's test it again then. I say execute workflow. My telegram trigger now listens to new messages. I will go back to the telegram then and I will ask maybe this time, Hi who are you? A presenter. I'll go back quickly here and you can already see AA agent works with GPT five and they are working on the response. They used simple memory, which means any previous conversation was already looked up, and now I can open this log. At the bottom, you've got those logs, yes. I can use that up arrow, and this is the output. PD five asks, do you mean the assistant previously reply to, who are you or do you have some specific test? I mean, it works and doesn't work because we've got output, but we've got this output only here in this field. If I go back to my telegram, I can't see this output here yes. That's because we have one very important component missing. This AI agent now has to send this output back to our telegram. Currently, I can only see it here locally within the AI agent node. Let's add that telegram output then. I click on that little plus here, and I search for telegram again. But now because I want to create that return path, let's say, I will search for send a text message. Here it is. And here, credentials can stay the same. That's a good thing. It might take a little bit longer to configure it first, but then you don't have to configure the credentials again. You can reuse previous credentials. You can create new one if you want, but you can simply reuse the previous ones. Preserve message is correct. Operation is send message, which is correct. But in fact, maybe you want to change it to send photo or maybe you want to send video, maybe you want to work with different AI models that can generate some photos or videos, et cetera. That's where you would personalize here, the response. But for me, it's a simple send message. But what I need here is chat ID. That's what I mentioned. Well we can see the output is from the previous node and previous node was AI agent itself, we still can see the telegram trigger as well. We can see the information from the first node. The telegram trigger tells us what the chat ID was actually used and chat ID is here and we can respond to the same chat ID, and I can simply grab and put it here. Whatever chat ID is, I will use it here in this field. The text note that I'm no longer interested in this original text. I am interested now in the output from my AI agent, so I drag this instead. This is the response of my AA agent to my original question, and that's what I need to put back to the telegram channel. And I can simply execute step here. I will reuse those details. If I execute the step, bam, node execute it successfully. If I go back to my bot, have a look. Now I can see the response here in my telegram application as well. The one thing that I don't like is that this bottom line, this message was sent automatically with an item. This is not what I want to see. I can change it easily. I can go back to the workflow. And here there is this little button add field, and the very first choice, append NA ten attribution. You click on that, you disable it, and now if I execute this step, it says execute it successfully. If I go back, I indeed have the same information again, but with no, this message was sent automatically now just the message I want to see. Let's test it one more time maybe, but let's test full workflow. We'll go back. I say execute workflow. My telegram is listening for new messages, telegram trigger. So I go to telegram application and I say, maybe, Hi, what ALA model are you? Enter. Go back here. It works with GPT five. It used the simple memory. It was very quick. I didn't even see when it happened, and it sent the message back to the telegram. If I go back to my telegram application, Hi, I'm hat GPT. Allowed language model, blah, blah, blah. So it doesn't say ready, it's a GPT five. It just says it's chat GPT, but never mind. You can see that works as expected. How do I add more LLM models then? Currently, I only have GPT five. How can I add more models? Maybe I want to have Deep Sk, maybe I want to have grog, et cetera. It's actually very easy. Let's rearrange it a little bit. Let me move it up. Or you can use this rectangle. No, I don't like it. I want to move it somewhere here. To add other LLM models, it's actually very, very easy. You've got those tools here. You click that little plus, and now you search for AI, I will show you AI agent tool. That's exactly what we want to use. I click on that, and now my main AI agent, which operates with GPT five, will now be able to access other AI agents, and I can use different large language models for those other AI agents. Let's say this one I will want to use as a rock. That will be my grog. Change the name and now here in the prompt, I don't want to type anything. I will just use this star like thing icon. Let the model define this parameter. That's perfectly fine. I just click on that because that parent node is AI agent and this like a child node is also AI agent, they will be able to communicate fine. I don't need to tell them what to do. I just change the name to Grock. Prompt will be defined by the model, I can go back to Canvas now. The only thing I have to do here is to connect the Grock chat model for this tool. With here. I will click on chat Model, and you might be tempted to use that Grochat model. But remember, we can use open Router and open Router. We already have credentials with open Router and within this solution, I can search for Grock. Maybe I want to use Grog three Mini. I changed the name to grog here as well. And that's as easy as it is. I already have credentials. I just choose different model. I go back to Canvas, and that's it. This tool is called Grock. Let's add another tool. Again, I use AI agent tool. I just click that stars, so the AI agents will talk to each other. I change this name to I don't know what should we choose. Maybe Deep Sk. That's all I need here. I go back. Okay, I messed up the name. Deepss not that important, but, you know, here, the important bit is chat Model. Again, don't search for Deepsk. It's already here, but we need open Router. That's the old beauty, you know. You just need one account with open Router, and that gives you access to every other large language model. And here with an open router, I search for Dipsk. Which one shall you choose? Maybe one, the original one. I will change it to Deepsk as well, the name of the model. Let's go back. That's cool. But how our main AA agent knows when to use them? When to use Grock or when to use Dipsik? Well, it doesn't, in fact. We have to give it instruction when to use this model, when to use that model, or maybe when to use just the default one and do not refer to any other tools because they are tools, yes. We can do that by double clicking on our main AI agent, and we just change the prompt. I can click this little icon here to make it bigger, and now I want to change my so called prompt, which is instruction for AI agent. I will just maybe copy paste. I say, take the output from Jason message text, which is still here because I still want to take the same message from telegram. If it includes something like use grog or Grock at the beginning of the message, maybe, I will add. Then use Grock tool to generate response. If it includes something like use Deep seek or Deep seek at the beginning of the message, then use Deep Sik tool to generate response. Once you have that response back, forward it back to telegram. If there is no specific instruction which tool to use, then just use default chat model and answer straight back to telegram. Pretty straightforward, isn't it? That should work in theory. Let's see. Let's go back to Canvas and now I will execute the entire workflow again. I go back to my telegram and I say, use Grock. Who are you? Let's enter, let's go back quickly. A agent still will use first GPT, and now we can already see it uses grog. It understands this instruction. It knows it expects response from the grog. Then it will prepare the message and send it back to telegram, hopefully. And bom, this is it. Workflow executed successfully. Let's see what it responded with. I'm grog, A assistant, blah, blah, blah, from X. A assistant built by X. That's enough to prove that Grock was used. We actually saw it running in the background. What about the Deep sk? Let's try that. Use Deep Sk. I will just ask the same, who are you? Let's enter. Let's go back quickly. Oh, I can execute work flow quickly, it should work. Yeah. If you do it pretty quickly, it will still work. But look at this now. It uses Deep Sik goes back, and it should send the message to the telegram. Bam. Workflow executed successfully. Greetings. I'm Deep seek R one, artificial intelligence. All right. But what if I don't say anything? At the beginning? I mean, I don't want to use Grock or Deepsk. I just say or ask, who are you? Let's enter. Let's go back. Again, execute workflow. Forgot to click first, but never mind. Let's see what happens. As you can see, it didn't go to Grock. It didn't go to Deepsk. I responded straight to a telegram without using any of those tools. Let's see what it responded with. Might not be super clear, but we can see it is response from GPT five, but it says, I don't have the who are you output. Please paste it here. It simply tells us what it expects as an input. Let's just delete the chart and start again, maybe. If I just start but see what happens. It refers to the instructions, maybe I will ask. Explain very shortly what is an A ten for. Enter execute workflow. We can see it doesn't refer to Grock or Dipsk. Let's see what's there as a response. All right, NA ten is an open source, note based workflow automation. That works. Let's compare it to response from Dipsk. I say use Deep Seek, explain very shortly what is an A ten for. Enter, go here, execute workflow. It hopefully now use Deep Seek as a tool and it does. Now it will generate the response. And let's have a look what Deep Sik version looks like. It's pretty similar. It's even shorter than coming from GPT, and let's compare the Grock. Explain very sharply what is an ADN for. Let's enter, go here, execute work flow quickly, and now it should go to Grock it does. Let's see. All three are very similar but different at the same time. But basically, once you have it configured and everything works as expected, now you can make this workflow active. You can just click active here in the top right corner. You say, got it and now you don't have to keep clicking that execute workflow. Every time you paste something in your telegram, this workflow will work automatically and we'll keep responding to all the messages that you post in that channel. That's all I wanted to show you today, thank you for watching, Mark. 7. Build Mistral OCR & AI Agent that reads documents: In this video, we will build a fully working NATN AI automation from scratch that can read PDF documents. I mean, even ones of very poor quality or with barely visible text. We will use a tool called Mistral OCR, and then we will pass that document to an AI agent and that AI agent can do anything we want with that document. For example, it can summarize it for us, I can translate it to another language, or it can answer questions about its content. As I said, we will go through the entire process step by step and by the end, you'll have fully functional solution that you can use at home or at work, or who knows? Maybe even customize it a little bit further and sell as a service to other companies. It's my personal automation and I've been using it quite a lot recently, as I had to review some legal documents after purchasing a property. These documents usually look like that and they are written in some legal language that I don't really understand. Instead of trying to read them as they are, I just use my AI automation instead and upload the document to Mr. OCR and then AI agent will process that data for. Maybe I will simply show how it works, okay? This is a PDF document that we are going to work with, and you can see it's very poor quality scan of my lease document from 1980s or 1990s, and it's not fun to read at all. Plus, you can see some letters are barely visible here, and also the text takes just a quarter of every page. But then most of the PDFs look like that, to be honest. And I think this will be a good example. And we can see it's 44 pages document, extremely boring, which means instead of reading all, what I usually do. You probably know by now that I'm lazy, I use my automation instead. I just do execute workflow. I browse for that document. This is this list document. I select it and submit. Now we can see the workflow is running. The document is being uploaded to Mistral we then get signed URL. Now the OCR happens, optical character recognition. This is another AI from Mistral that deals with this document. It tries to read what is in that document. Then we've got a markdown node, and now AI agent creates a summary for me of that document. I use Open AI, I think GPD for dot one, if I remember right. But we are going to build that from scratch again anyways. If we check the output, if I click the agent. All right, something I didn't think of actually at the beginning, there are some personal details, but you can see, certainly, here is a little summary of the document you provided, which is a lease agreement and associated deeds related to the property I mentioned. Let me scroll down maybe. It tells me that this document includes images such as architectural role plans and company seals, blah, blah, blah. But now I've got one page and my instruction is very basic. If we check my instructions only take all information from this side. And create summary of that document, and this is the output. This is summary of this document. That's all the AI does and how to set it up, how to customize the instructions for the AI agent and everything else, you will see during the build process. So maybe let's get started. Let me close this one. I didn't really mean to change anything, so I can live without saving. But to get you started, you need one big prerequisite, though, and this is the NA ten platform. And NA ten is a free to use automation platform, but it's free to use if you self host it, which means you run it locally on your PC laptop or maybe mini PC, like I do, or maybe some server in the Cloud. And what NATen is and how to run it locally, we had a dedicated video about it. So if you do that first, if you want to follow each step, you have to watch that previous video which includes step by step guide on how to run this NA ten locally, which means you can run it free of charge. There was also another video where we built our first AI agent, and that one is not really a prerequisite, but it might make a lot of sense to watch that one too. But as I said, it's optional, so it's your choice if you want to watch that one or not. Once you have NA ten up and running, we can start creating our workflow. By default, you will see nothing here, no projects, but you will have that create workflow button. That's what you want to click. And that's where you start building your workflows in ANATN. The first step we want, first step is always so called trigger. Something has to trigger that entire workflow. Let me click that plus then, and I will search I think it's a form something. Yes, NATNFm. And I need that trigger on a new NA ten form event. We are building that first bottom where I could upload my file, my last document, in that case. We will then be able to send that document up to Mistral OCR. And there is not much here for us to do. Form title, maybe I will call it document. You can add description, but it's optional, but what we want to do here is add form element. And field name, you will see what it is. Don't worry about it. Form element, we will call it document again. But the element type, we want to choose file because there are many elements that you can submit using this node, but we want to submit file. And we want to submit single file. So let me disable that multiple files. We are interested in single file, and now you can also limit it further to maybe just PDFs. But the fact is mistrl OCR can read more than PDF. It can read word documents, power point presentations, so I will just leave it as it is. I will switch that to a required field because we have to attach a document. We have to work with something. That's it. That's all for this node. If we now go back to Canvas, you can see we've got our first node because they are all called nodes here. It's also a trigger, so called trigger. Let me maybe rearrange that and I can already test it. If I say execute workflow, now you can see those fields. First field we call document. Second field, we also called document. Sorry. This is that second field, and this is this window which let us select the file and submit it. If I browse for my file, that's the least document, I can now select it and I can also submit. And we've got workflow executed successfully. That's what we want to see. That's basically it. Our first node is up and running, which means now we can create a second node, which will upload our document up to Mistral AI. But before we do that, we have to have an account on Mistral OCR or Mistral AI. Because if you go there, the website is mistral.ai. That's what we need. Mistral OCR is currently the best software available on the market to read information from PDFs, images, and other documents. The coolest thing about Mistral OCR is that it's currently free of charge for you to use. Have to open so called experiment plan or experimental plan, something like that, and it will not require any card details or any other payment methods. Simply when you sign up, it will ask you what type of account it is. You just say experiment, and that's it. And what it means, you will be able to click that try the API. And this is where you will get free API keys. You will be able to create API keys. Like here, I can create new key. I already have some, I mean, but I can create another one. And if you go to that doc in the top right corner, and then on the left, we choose document AI, I think. Yes, that's it. And we choose that basic OCR. It's not really basic. It's very sophisticated. But you can see the options OCR with PDF, OCR with uploaded PDF, and OCR with image. OCR stands for optical character recognition. What we need from this website, we have to scroll down and these are our options. I'm already on curl, but I believe by default there on Python, you will see something like that. If you have your solution written in Python, this is the code you can use. But for us, we will need that curl. You will see shortly why. Simply click that curl option, and all other options we should switch to that curl as well, and you might guess it, we will need that upload a file and we need that curl code. Why do we need it? You will see. There is this gray little icon that will let you copy everything. Let's use it. I will just click on that to copy everything. I go back to my workflow now, and now I will click on this little plus to add another node, and I will search for HTTP. And this is exactly what I need. We will need actually three of them, and this is the first one. These requests can simply communicate between our PC or Mini PC or server, whatever you run your NATanm and it's the way of communication with Mistral AI remote site. I clicked on that, and now I can click that Import curl. And now we simply paste this code here. I will paste it, click Import, and what it does, it simply populates all those fields for me. And this is all cool. But the first thing I need here is the API key and how to get the API key. Once you have your account, you have to go to that mistrl either here API or maybe from the main side, you click that try the API, you will basically land on this page and you click that API keys. And you click Create New key. You give it a name, Mark Mark key. That's fine. I can set expiration date. That's really good choice here. For this video, I will choose tomorrow's date, which means tomorrow this key will already be expired. That way, I don't have to blur it or something because API key is basically your password, so how to keep it safe. But because mine will expire tomorrow, I will not create this video till tomorrow. It will be already expired when you watch the video. I click Create New key. And this API key is only shown once, so you copy the key and either save it immediately somewhere, or I will simply go back to our workflow and I will paste it here, and I want to paste it exactly here after that barrier information. Which means this is a header. This is a HTTP header that will go to Mistral and it will show Mistral that it's you. You can think of that as your user name and password for that Mistral service. This is what API key basically is. Maybe I will change one more thing because we need three of those nodes, and they will all be called HTTP request. Let me change that name to maybe upload to Mistral. That's cool. And we've got the authorization, and below also was automatically populated. The purpose is OCR, but there is one more thing because here you've got parameter type, and the type is NA ten binary file. That's correct. The name can be left as file. It doesn't really matter, but what matters here is that input data field name. And what it is, the binary file, if you switch because you've got different views here, here, we've got Schema, we've got table, JSON, and the binary. So make sure you are here on the binary, and we need the name of our file, which is here called a document. Simply whatever here is on the top in red, you want to copy paste it here in that field. And once we have it, I believe that's all we need. If I execute the step, okay, I asked me to run the whole workflow. That's fine. I will go back. Let's run this workflow, I will browse, I will choose my document, select, submit, and let's see what happens. Two things can actually happen. I can work or cannot because I will tell you, all right, workflow executed successfully. That's success, but you might get an error at this stage because these API keys, when you just create them, sometimes they don't work immediately. Actually, it can take up to 2 hours before an API key works correctly. Was lucky one because mine worked within 1 minute, but it might not be the case for you. If it doesn't work, don't panic, don't worry, and don't try to change anything. Simply leave it as it is and try again in a while. Maybe your API key is simply not ready yet, but for us, it works already as expected, so we can go further. What I can also do, I can double click on that and I can pin that information. What that pin does, it says exactly node will always output current data instead of executing. I will click that, I will go back, and now it's like this purple color. That means it's pinned. And for testing purposes, it's a good option because it means when you execute the workflow, this data is not constantly sent to mistrl back and forth. It will simply remember the current data and we'll just use this. It's especially useful if you have API that is limited to maybe ten requests a minute, let's say. You don't want to push the same data over and over again, it's better to pin it and it's stored here locally for now. All right, but let's build another node, and another node will get us signed URL. Let me click that plus. It will be again HTTP request. That's fine. And we want to also import it as curl command. So I will go back here to this page with our curl commands, and this is the one I need. Get signed URL. But what even it is about? I will maybe first copy it and then I will explain it, right. So I copied this, get signed URL. I go back to my node, Import curl. It's not the one. I have a new one, it's this command, and I say Import, and I change the node name to get URL. And now let me explain what it is at all. You can see here on the left, it's input from the previous node from that node that uploaded the file to Mistral. It uploaded it to Mistral and Mistral gave us that ID. This is the identifier for our file, but already there in that remote location in their cloud. And what we can do now, we can create a signed URL signed URL is a secure temporary link that can access your file in that mistral cloud because our file is there in Mistral and it now just waits to be processed, but it's not processed yet. This signed URL is a secure link that we can use to locate our file in that mistral cloud. I need this ID. I want to move it here. In this URL, I want to replace. You can see dollar sign ID. I want to remove that and what I really need, I want to move this container. I want to move this JSON object. You can even swap to JSON. You can see ID, yes. Is the JSON view. If I switch to schema, that's the schema view. I can now hover over that and drag this ID here, to that URL. Make sure this filled up changed to expression. You can click first expression and only then drag this container. This is JSON object, and whatever is in that object, we are not interested about this value right now because this is static value for this particular file. We are interested in this container, in this object. It doesn't matter what current value is. We want to have anything that is currently in this object, in this JSON object. That's why I drag this portion here, and I say, whatever is in that JSON ID, currently it's this. It's whatever is in that object. I want to have it in this URL. If I click here, result, the result shows this actual value that is currently there, this is simply taken out from this container. Okay, I hope that makes sense. Below, you can also change the expiry. By default, it's 24 hours, but honestly, we will need just maybe 1 hour or even note. But on the other hand, it's safe to leave 24 hours because they are pretty safe. The signed URL is really secure link. It can be valid for 24 hours. That's not a problem. And now let's scroll further. I entirely sure if I need this header, let's just remove it. But we need the API key again. Maybe let me remove this value. For now, I will go back and copy it from the previous node because I already don't have it in case. I copied something else in the meantime. So I'll go back now to get URL, again, authorization, and this is better, and this is my API key because it stays the same. And that I believe is everything we need here. Let's go back and we need one more HTDPRquest node. So search again, HTDP. That's the one we need. And now we need a node that will actually process the data. Go back to basic OCR to this website I mean. And we search for it's at the top, OCR with PDF. So basically the very first one here on this website right now. That's the curl command we need. I copy this one, I go back to my workflow, and again, import curl. I remove that and I paste this one. I say Import, all fields are populated, and I will need APIKey again. So maybe I will first change the name again to let's say OCR results. I will go back again and copy my API key again from this node this time. That's the API key. I go back to my OCR results node, and I paste it there. But if we scroll further, we will also change. We will also have to change this document URL. For the time being, it's like some random, I don't know what it is, even, but that's not really what we need. What we need is actually our document URL, and our document URL is that signed URL that we created in previous node. And unfortunately, I can't see it here because I forgot to run it and pin the data. What we have to do I have to go back to Canvas and maybe let's run entire workflow, right? So I will browse for my lease document again. I will select it, I will submit it. This node will just go through. Get URL seems to be working fine because it's also green. I will double click on that. I will pin this data as well because I know everything seems to be fine and I can even see problem in node OCR results. Well, I know about it. That's because we need that URL. Nod OCR results is the next node this. Here, I want to only pin the data. I go back and we go to OCR results. We just double click on that, we scroll down, and now we have that URL. This is this long URL, it's signed URL, which allows us to pull the file back to our PC or wherever this NATA is running. I just drag this. It's again, JSON object that says URL, and I drag it there and paste it between those double quotes. And now we've got JSON URL. That's exactly what we need. That's basically it. And the last thing here, this node will try to output everything as a binary file. You can see file here, response format. So you can either change it to JSOMO what you can also do. You can simply remove that entire response field. And maybe I will use this option. I will simply remove. You only have this send body, and there is nothing else. Let's now go back to Canvas and let's try to execute that workflow again. Let's see what happens. Select, submit. It will go through these because they are pinned, only this will be processed. Will it works as expected this time? Yes, we can see workflow executed successfully. Let me pin this data before I forget by clicking this pin data icon. We can actually see already what was previously barely visible text we had in that PDF. We now have it in clear format, maybe let's move a few pages, no personal information is further down this document. You can see now, it clearly says what it is, demised premises, the lesson making good damage the fire blah, blah. Whatever was in that poorly scanned PDF document, now I have it here and it's formatted properly. Every single page of that document is shown here as a single markdown field. If I scroll further, there is another markdown. It's another index. Index means index seven means page eight really because computer starts counting from zero. I can scroll further and further each page was translated by mistrl OCR and it's here. It doesn't change the fact that it's still the same boring document I don't want to read. Maybe, first of all, I want to get rid of all that redundant data. I don't need that height weight dimensions, blah, blah. I want to have it formatted properly and maybe not markdown, but maybe as a text document or word document or maybe let's say HTML document that I can simply open in this browser. If you want to switch that view to HTML, it's actually very simple because you go back and you add another node which is called markdown. That's the one. We want to switch markdown to HTML code. There is one little issue, though, because if all your documents had only one page, then you would have to go here. Page zero is basically first page. You would want to drag this markdown simply first change to expression, yes. And now I can drag this field here. But the problem with that is this is only working for page zero, which is this first page. Or if I want maybe to translate just page one, I would remove this and I would drag that, yes. This will work only for page one, but that's not what I need. My document has multiple pages, and I need all of them to be translated to HTML, and maybe I want to also preserve the page numbers. So my query here has to be a little bit more sophisticated. We can simply ask hat GPT. We can copy this output here, maybe JSON, maybe in JSON format. I will just copy portion of that and I can ask hat GPT what JSON query I should use to preserve all the pages and to combine information only from Markdown field. I basically need each of them to be on a separate page. That will be again separate page, and that again. And that's exactly what I did. I asked hatDPT and hatDPT came up with several solutions because there is not one solution that work, but it told me to use, for example, that, and this indeed works as expected, because I already tested it. And the destination key, we also want to change it as HTML. So if I now go back to Canvas, let's run this workflow again. It's a workflow executed successfully. So let's have a look. Let me pin that data before I forget. The input is that from OCR results. It's processed here in the middle and the output is on my right. If I slide this view here to the right, this is HTML code that can be displayed by my browser. Again, let me scroll down so there is no personal data. But you can see it's split into pages and each markdown is now on separate page. You could simply save it, and maybe that's what you need. Maybe you want to display it in your browser. But that's not what I want to do. I don't want to read this document at all. I want an AI agent to have a look on that and either give me a summary of that or maybe translate it to other language, or maybe I only need information about some fixtures, like here or external walls. Maybe I want to ask AI a very specific question. It will go through this document and will simply answer instead of me going through 44 pages trying to find this information. The good thing is the HTML code is perfectly fine for AI. It doesn't need specific format as long as it's formatted properly here, it's one big blob it can work with. For AI, it doesn't really matter if it's HTML, if it's text document, if it's Word document. This is basically something we can pass further to AI if that's what we need. I'll go back to Canvas and maybe let's rearrange that a bit because we need a space for our AI agent. I just tick that plus. I say AI, AI agent. Let's switch the view to table. Now on the right, I mean, on the left side on the input, but if we slide to the right, we've got this HTML code, and maybe let's scroll it down a little bit, so there is no personal information, only some generic one. And all I have to do here is to create a prompt for my AA agent. What should be the source of the prompt? I say, I will define it below. I click that and I simply tell it, take all the information from this HTML. Which is another JSON object that stores all these data, yes. So again, change this to expression here now and move this HTML field there. So take all the information from this field, basically everything that is there and create a summary of that document. That's very basic prompt. You can get more sophisticated one or maybe you want to ask specific question or maybe you want to translate it to other language. That's where you can specify the instruction for AI agent, what to do with it. I will just tell it to create a summary, but maybe that's not what you want. Maybe you want to do something else. That's where you specify that information. So I will go back to Canvas because that's all we need here. And our agent is red right now because it doesn't have any LLM model attached. And to attach LLM model, you need one more thing. You have to sign up to one of the LLMs, which means you can click that plus, but you also will need an account on one of those providers. I have, for example, OpenAI, which has Chat GPT and GPT 3.5 or GPD for one models. You can use Olama which I think has some free of charge models for you to use. But what I would really recommend is to get an account on that open Router. Why? Let me show you. I click that open router. And what open outer allows me to do, I can choose any LLM model I want. Look, I've got Open AI here included. Let me scroll down maybe. We've got Grock from X, yes. We've got some Tod DM, I don't know what it is. The drummer, no idea as well. Soso, blah, blah, blah, QN. And we also have perplexity. You know, you have one tool that has access to most of the LLMs, or maybe all of the LLMs, because most of them I've never even heard of. And we've got also Open AI included here. Like Open AI GPT 41 Mini is selected by default. But the thing is, if you want to use Open Router and Open AI model, then you have to first create an account open router, and then you have to add some credits to your account. You have to top up, for example, $5, and then you can use those credits when you use the API keys. You would go then to API keys. You would create your API key, and once you have that API key, you go back. Let me maybe duplicate this tab. Basically, go back to the main page, you create credential and you create credential for open Router. And you paste your API key here, but I already have one, so I don't have to do it. But if you think it's worth it, and I really recommend it because then you have access to all LLMs. You can switch between various different large language models. I chose for one mini or maybe for one. Let's change it to for one, not for one mini. That's basically all I need. I can go back to Canvas. Now the agent is happy, it's not read anymore, at least. So now, maybe let's save get of work. It's worth to save every now and then. And let's execute that workflow again. Let's see what happens. My least document, select, submit. Went through because all the information is pinned, but AI agent now works on a summary of that document for us. It took a while, but it says workflow executed successfully. Available click on that, that's what I have currently. Here is a summary of the key information in the document provided. It's much shorter version, for example, alterations, no structural alterations or exterior changes without written consent. Now I have one page to go through rather than 44. Or if I have specific question about this document, I can simply change it here. And that's where we are now. But you can develop it even further. What you want to do with this document? Maybe you want to send it to Whatsapp, maybe to telegram, maybe you want to email it somewhere. Or maybe this is just first part of your longer automation. Maybe you want to add even more nodes, maybe you want to process it even further. I don't know. You can add memory, you can add tools. You can continue that workflow, develop it even further. You can do anything you want with it, and probably every person will have different needs and will want to do different maybe send it somewhere else. But for me, this state is completed because our document is processed and it's been summarized. And that was my goal here today. The video was helpful and clear enough for you to build this solution yourself. Thank you for watching and see you later. 8. Access n8n workflows from the Internet: In this video, I want to share with you a tool I built for myself in NA ten that I use basically every single day. It's just a bunch of AI agents, but each AI agent has different LLM model attached, and I can access those AI agents or automations from anywhere in the world. So this is what it looks like in my browser, and I have that drop down menu that I can use, and I can choose here any model I want to currently use. For example, GPT, Cloud, Gemini, deepsk QN perplexity, you can configure as many as you want. We will use OpenRouter for that, which has access to loads and loads of different LLMs. You will see shortly. If I choose Gemini flash, let's say, I can ask a question here and if I want to attach a file, for example, a screenshot I took or the photo or text file, whatever I want, I can attach here as well using that choose files bottom. But let's maybe run it as it is. And we can see we get the answer in no time. I can access that web page from any PC, laptop, phone, or even from my TV. No, I'm not joking. Here it is. It's me accessing my AI agents from my TV. But that means I always have access to that automation from any device I want. We will build all of that today from scratch. If you look at the actual automation running in the background, it might look a little bit complicated, but what it really is is just a single AI agent that I copied eight times and I just change the LLM model. You will see shortly what I mean. It's very basic automation, believe me, and you don't need any previous experience to follow this video. You can start building it right now with me. But anyways, first, we need that NA ten platform. And if you don't have your own NA ten yet, then there are multiple options you can choose from. You can even self host your NA ten platform on your laptop or PC and we had a video about it, but then attaching a domain might be a little bit challenging because you need something called Angroc or Cloud fer tunnel to make it working. What you can use instead, you can use, for example, AWS platform where you can run a server for free for six months, or even longer if you include credit. Or you can use other Cloud providers like Hostinger which I'm using right now. I think it's the easiest option which also automatically provides you a domain so you don't have to worry about it. Hostinger has recently become a very popular option for self hosting NATen because they offer a VPs that comes preconfigured with everything you need to start creating your automations. Just go to hostinger.com slash automation Avenue, and it will take you directly to the landing page for that self hosted NA ten virtual private server. You simply scroll down a little bit. And you will see all available VPS plans. This is where you choose the size of your server. Then you choose how long you will need this server for. Hostinger gave me a code that you can use, and the code is automation Avenue. Once you apply that code, you will get 10% extra off that price. Then you just continue and you have to register. If you don't have an account already, will create a new account and then it will get you to this page where you choose a location for your server. I will stick to United Kingdom. It's fine. So I click next. And here, you can choose either plain OS. You can choos Santos, Debian, et cetera. But what we are interested in is the application. And you can choose here, for example, vanilla NA ten. That's actually what I'm interested in. But you can maybe want to go for the one that has already applied 100 workflows or maybe so called NA ten in Q mode. But as I said, I'm fine with just NA ten. That's what I need. Then you can choose if you want to have Malware scammer included. I will leave it as it is, which means I can click Finish setup now. Now you can click that Manage VPS which is Virtual Private server, and this is it. This is your server up and running. And if you click that Manage Up button, it will take you straight to your NA ten application. It couldn't get any easier. Because I open it for the first time, it will ask me to set up owner account. So we type like email address, your name, the last name, and you create a password. Then you just click next, fill in a form, whatever, that takes you straight to your application. This is your NA ten platform up and running. That's what it looks like right now, and you can create your NA ten AI agents and workflow automations. Let's create another one then. I say create Workflow here in the top right corner. And we always have to start with the trigger. Let's create that so called trigger. I click that plus, and I'm looking for NA ten form. That's the one, and this is OnwFmevent. And at the very top, you've got very interesting thing. You've got that test URL and production URL. These are the locations, I mean, URLs that you can simply copy, and they will work over the Internet because this is fully qualified domain name. That's what I was talking about. If you are self hosting on laptop or PC, this part is a little bit challenging. But here with Hostinger, you don't even have to think about it. They will provide you with those URLs, which is great. For the time being, let's leave it. Next field this authentication. And yes, we will use the authentication. But for the time being, let's leave it as known. The form title, let's call it just chat. Form description, let's say multiple LLMs and form elements. Let's add some elements. Don't worry if you don't know what it is about. You will see shortly what we are doing here. Let's just add that form element, and I will say LLM to use. Element type, I won't drop down. And for the time being, let's call it just GPT, maybe. GPT 4.1, I think I had in original form. And I will make it required field. We have to choose this. What we can do now, we can test that URL. Let's check that test URL maybe. I can click here. It will copy that long URL, says form URL copy it. I will open new tab and I will just paste it. It will now not work because it says form triggering isn't listening yet. But if we go back and if we say execute you can now see it listening, and in fact, you don't even have to open a new tab. It's basically just displayed here in a new window. If it's not displayed, you have to search for a little warning that pop up window was blocked or something like that. You have to allow N ATN to display those pop up windows, right? This is basically this form we are working on. Form title is chat. The form description is multiple LLMs. The form element, first form element, we only have one so far is called LLM to use, and in the dropdown menu, we've got this GPT for one. Now I can submit that form and look at that node executed successfully. We can close it now. And yes, let's pin it. It says, You can pin this output instead of waiting for a test event. Let's pin it. This output simply will be saved temporarily here. You know what? Maybe not yet because we want to add some more elements. We've got that form element LLM to use. That will let us choose GPT for one, for example, but we also need a field for our instruction for the GPT four. I will add another form element here. I will call it instruction for LLM element type. This time, it can be text where I can ask my LLM some questions or provide some instructions. Yes. And in the place hurler, I will just say type here, and I will also make it required field. So now if I go, I will copy that test URL, I will execute this step now. It's listening again. If I open it in a new window, place it and go, we can see now we have a new option. First one is that LLM to use where I can choose my GPT for one, and we've got that new field where I can type my instruction like what is an A 104. I don't know. If I submit it now, it says form submitted, and if we go back to our automation, it's already disappeared, but it was saying it was submitted successfully. But in our form, we also had the field where we could attach file, yes. So I will add yet another form element, and I will say attach file. Element type, I will choose file. Let multiple files to be attached. That's not a problem. Maybe sometimes I want to ask AI question about multiple files, not about single file. Here you can filter what type of files you want to accept. Only JPx, only PNG, maybe only TXT, let's say. But if you leave it clear, it will simply accept all file types. But I will not make this field required. Not always, my question to my LLM model is about a picture or snapshot or something that I attach. Maybe I have a simple question like what is two plus two. Then I will send it with no picture or file attached. So we leave it not required we also can change here, respond when the workflow finishes, because we don't want to have it as soon as format is suben it. When we work on our workflow, we want to have a response when workflow is finished, complete workflow including AI agents. If I execute it now, again, past and go, now we have dropdown money with a single option, but we have it there. Instruction we can provide and we can also now attach a file. If I click this button, you can see I can browse for any files I might have and I can attach them if I want. What is two plus? Let's say, I will not attach the file at this time at this stage, and I will submit it. If I go back, it says node executed successfully. And now we can pin it. I will pin it, so it will be saved because that's all we need, really. There is one more thing we can change. In this chart, you can see form automated with NA ten. If you want to get rid of this, let's go back here and in options, you can add that append NA ten automation, attribution you can see it's enabled by default. If I disable it, it will display this form with no attribution, which is this form automated with Natan. But, you know, it's not really important. Anyways, we can now go back to Canvas. This form element is now completed, at least for the time being. Now we can add our AI agent. I click that little plus, I say AI, and I say AI agent. And on the left, you can see what comes to the AI agent from the previous node from this NA ten form. You can see all of our fields, LLM to use PT for one, instruction for LLM, attach file. Currently, there is no file attached, but they are basically all the elements that we had in our form. The middle section is for our AI agent, what it has to do with that stuff on the left. So I say source for prompt. I use that dropdown menu and I change it to define below. And here in this window, I just say, it's very basic. It will be user message. I will just drag this field. Instruction for Alam. This is what I said. What is to plastos? This is my question to my AAA LLM to my AI agent that will use AAA LLM model. I simply drag this field here. Ser message is whatever is in this field, instruction for LLM, and then user file. And maybe let's add optional because you will not always have user file, and user file will be here in this attached file field. Currently, there is no file, but it doesn't matter. I simply drag this field. Whenever there is a file, AI agent will know about it and will be able to read it. And that's it. It can be as simple as it is, these two fields. User message, user file optional. Now we click that add option and we choose system message. System message tells our agent what is its role and what is it supposed to do with all that stuff that comes from the user? It starts with your helpful assistant, but maybe let me clear that and I will copy paste my message because it's pretty long, I will paste it here. And it says, You are a helpful assistant that analyses documents and answers questions about them. No, that's not correct. Sorry. That responds to user messages, maybe. You can fiddle with that system message, whatever suits you best. Maybe you have specific instructions for the AI, so you can change it to whatever you want. So we're a helpful assistant that responds to user messages. When a user submits a file, carefully review its content and provides summary, direct answers, relevance. I will add also, please note the file is optional. So if there is no file, then just respond to what is in user message. Field. That's it. This is it, but now we have to attach hat model. This is the most important bit. Yes. Our AI agent is dumb if it doesn't have any chat model attached, like large language model. Yes. We click this little plus, or you can choose it here. But if you are here on this page, you can also click here this little plus next to chat Model. Next you do the same choice of models. Can choose anthropic specific ones. Google Gemini, Deep seek. But I would highly recommend to go to that OpenRouter chat model instead because this OpenRouter will give you access to all LLMs, even to the ones that you've never heard of like Kimi, pH, Mistral really good one. Minimax, Loads Lama from meta. You got loads and loads Google Gemini, look at that. How many of them just from Google. But to have access to OpenRouter API, you first need to create an account with OpenRouter. I already have account with OpenRouter. So if you don't have one yet, you simply sign up to openrouter.ai. And then once you are on that website, you can see all the models available, and there are some models that are free of charge, and you can use it straightaway. But if you want to access all models, then you have to add some credits to your account. As you can see, I still have $14 and I can add more credits if I want. And once you add some credits, even $5 or whatever, then you will get access to all LLM models. And this is real beauty because you can access them all from one single place. And then you can go to API keys, and I've got a key, but I can create also new key. By default, you will have no key, yes. Then you create new key. You call it like NI ten. New key, maybe you can set credit limit, maybe $3. And once this limit is reached, this key stops working, and you just create it. And this is my API key that I can now copy it's copied, I go back to my NA ten and what I can do now in these credentials, I can create a new credential. You can ignore this one. This is my current API key, but you will have no keys at all. You have to click that Create New credential. Now I just call it NA ten new key, maybe underscores and I will paste my key. And they save it. Connection tested successfully, which means my API key works as expected. That's what it is. Credential to connect with my API key. I connects me to OpenRouter and the model. Yes, let's leave it as it is. GPT 4.1. It's chosen by default, and that's what we also chose in our drop down my knee. That's basically it. We can go back to Canvas and our AI agent should be completed now. I go back to my form then, I copy my test URL. Let's execute the step. Oh, yes, I can now unpin the data. Listening for form submission. Let me close the old ones. I will create new one, New Tab, paste and go, LLM, which was GPT 4.1. I ask some silly like what we already asked, what is 2.2. I submit it, but now this will go. If we go to agent, I can say execute step. Now it's executing the step and the answer is 2.2 equals four, which is great. We've got the answer, but we've got the answer only here. We want the answer back on our website. I want this answer here in my form. I don't want it locally here only. I want this answer to be fed back to my form. So what I do now, I'll go back to Canvas and we need another node. I click that little plus, and I say NA ten. Again, it's NA ten form, but now I say form ending. And the page type and NA ten form submission, they can be both left as they are. Completion title, I say maybe response from let's see what we have available here. These are all the fields I have available, and I want to say that it's response from my LLM that I chose. So I drag this field again. Response from whatever LLM was used. Completion message is the answer from my AI agent. I actually only have one field from AI agent, and it is output, but that's exactly what I need. I need the output to be here in my completion message, and that's basically it. If we execute step now, look at that. Another pop up window, it shows us what the response looks like, and I think it looks okay. It tells me what model was in use, and it gives me the response from my chat model. But what you will see, what you will notice, if we go back, if we save it, I keep forgetting to save sometimes and I lose part of the work. You have to save as often as possible. Basically, if we go back, if I copy the test URL, if I execute the step Paste and go, GPT, what is three plus three? Maybe submit. If you run it like that, it will show what you might be tempted to do. It will not show you the answer, yes. I mean, there are many things you can do, but you can actually make it production now. You say active. You can activate that workflow, and you can use your production, URL. You switch to production, you copy this. And now, if you paste it, let's check What is three plus three? Now you get this response, yes. So there is a little difference, I mean, between test and production links. You have to bear that in mind. But what I also wanted to show you is this field here is very narrow. If you have long output, because this is basic question three plus three. But what if you asking about the meaning of life or whatever? The answer might be very long, but this field will be very narrow, so the answer will be like in million different lines. We can make this answer wider, this entire window wider. Okay? Let me show you how to do that. Let's go back. In the last form, if you choose that options, you can see that custom form styling. And if we make it bigger maybe, you can see all the form styling. And what you need to find is dimensions. You can see them here. They are the dimensions of this little window, and we want to make it wider. What I tend to do is I tend to change from pixels like 448 pixels. You can change to either 90, let's say, FW W, its so called view port, or you can simply use percent percent of the width available, or maybe not even 90 but 98%. Let's use that 98% and if I close it, now if we run it again, save, remember to save it. Once it's saved, you can just open again, past and go, GPT, what maybe what is NA ten for. Let's choose this one. It will generate longer output. We submit it now. And now you can see this window is much wider. You can play with those settings. This is just one setting. You can change the spacing, you can change the fonts, et cetera. You can customize it any way you wish. I'm not that picky, so I left everything else as it was, but you can play with that if you want. There are many other settings you can change. And believe me or not, but we are nearly there. You will say a mark, but we only have one large language model, yes. Well, yes, that's true, but it's very easy now to add many more. We simply let's remove this link here, so we detach that form submission. I can simply now highlight everything. And I say, duplicate three nodes. This is it. Let's rearrange a bit. And here in the middle, we add another node called switch that will switch between different LLM models, and we create rules. First rule, I will drag LLM to use. That's the only difference between them. And I say when that LLM to use equals GPT for one, which is this value that's it for now. Let's go back to Canvas. You can see a new connection was created and when I choose the GPT 4.1, I want this switch to connect to this AI agent because it uses GPT 41 as my LLM model. Now let's go back to this agent, and in OpenRouter, I will change it to whatever. Maybe anthropic Cloud Sonnet 4.5. What I have to do now, I have to go to form submission. I have to add that option. I will call it just Cloud, maybe. That's it. Just remember, whatever you put here, you have to copy, it has to be exactly the same in the following nodes, like in the switch, I double click on switch. I add another routing rule. If this field LLM to use will equal two. I will paste my Cloud. We go back to Canvas. We can see another connection was created, and that is for Cloud. That simply will be attached to this AI agent, which uses Cloud as OpenRouter chat model. Maybe let's save it now and see if it works. I say, Execute workflow, we can test it here and we can see now two options. I will choose Cloud and I will ask it, what is an A ten for? Maybe let's move it a bit so we can see which AI agent it goes to. I submit it, and as you can see, it goes to clod. For a while. And we've got the response from Cloud A ten workload, Mon, blah, blah, blah. This is basically what you have to do. You have to do it for every single agent. If you want more of them, it's super easy. You copy everything again, you duplicate them, and you just one sec. For this one, let's say, I will use Deep seek, so I go back, starting from the form submission, I need to add another field option. I will call it Deep sik. Let's copy it. And on the switch, it has to be under the same name. I just add another routing rule when LLM to use will equal to Deep Sk, then I will attach it to this AI agent. Save. Don't forget to save because it won't work if you don't save. And now let's execute the workflow. We've got now Deep Sik. What is an ATN for submit you can see. Yeah, it uses Deep Sik now. Let's wait for the answer. And that's response from Lipsik. So it's a great tool to test many different LLM models. You can attach as many LLMs as you want to this switch. Just remember that on form submission, and on the switch, they have to be under the same name. So that switch understands what is coming, yes. The last thing you should think of is to add authentication because this URL, for example, you can see it here. I will remove the entire automation. This is just for testing. But if you overlooked somebody else's URL, you could simply start using DTS because there is no authentication at all. Let's start using authentication. I already have basic auth I created, but in your case, you will have no authentication method. You have to go back all the way to all the projects and you don't create workflow, you create credential. And you create basic Of. Say, continue, you can call it whatever my of my authentication. And let's say my user will be Marek and my password, whatever you want your password to be. I save this my of. Now I can go back to the workflows to my new workflow, to the form and for basic Of, I choose that credential that I have just created. I called it my Of. It only supports basic authentication for now. Maybe it will change in the future. But if you create complicated user name and password, I think it's good enough, you know? Yeah. So now, if we go back to production, if we save it first, yes. I mean, let me copy the production URL. So if I open it, now it will ask me for my credentials. It was Marek, and the password was whatever you configured it to. It's a simple authentication to get to your webpage. Hope that makes sense. I hope that video was helpful and thank you for watching. 9. Get 2000+ FREE n8n templates (DON'T buy 'premium' workflows!): Going to be a very short video. I simply want to show you how to obtain thousands and thousands of NA ten workflows completely free of charge. It's simply another day where I see people complaining that they were scammed because they were trying to buy premium NA ten templates or so called premium, workflows for like 20 or $30, and these templates either never arrived or maybe they even received some files, but they have no idea what to do with those files. And the fact is, if somebody is trying to sell 1,000 or 2000 workflows, then it means that these workflows were never theirs anyways. They simply try to resell some stuff that they found online themselves. I still don't really even understand why people would fall for that because I think it only really makes sense if you want to buy one very specific crafted for you, workflow that is designed to solve your particular problem. But if you are looking for thousands and thousands of templates, then they are already available in multiple places and they are completely free Well, the funny thing is that even on the NNA ten website, there's already over 4,500 workflows available. You just go to NATNFward Workflows, something like that, and you can simply copy paste all those workflows straight to your NATen. But honestly, these cameras didn't even bother to copy paste thousands of templates from ANA ten website. What they do, they are selling the workflows that are already available on Github, Gitlab, Bitbucket, or other remote repositories. So how to find ones on the NA ten website is very simple. But let me show you how you can easily find those other ones in remote repository. Just Google something like I don't know, Github and Aten workflows. I will just press Enter and I will pick I don't know, very first link. It already says, this one is 1,000 workflows. Here, I first link, we've got 2053 workflows. Okay, that's fine. And the workflows are here in this folder. But let's see what to do exactly. We can see read me file, we'll tell you, Okay, it's 2053 workflows, and there is even some documentation system. Looks like it's something that you can open in the browser. So let's have a look further. What does it say? You've got some categories, et cetera. All right, usage instruction, modern fast system recommended. Okay, let's follow these steps then. First, it says Git clone and repo URL. What I will do then, I will open my terminal, and they say, Git Clone. And the repo URL, we have to scroll up, and this is the Repo URL. You have to click that code, and this HDDPS GitHub, blah, blah, blah. That's what we have to copy. I will go back, paste it, and I will present. Those files are now being downloaded from Gitlab from Github, sorry. This is actually Ubunto server. So if you on LS Li, I don't need A, but I can see the folder NA ten workflows. It's just been created. That means I have those files now locally here on my Ubunto. Let's have a look then what we should do next. Git clone that's done CD into that folder. Okay? I see the NATen workflows, which means I change directory. I'm now in my home directory in NA ten workflows. What do we do next? It says to run PIP install requirements. Oh, I don't even have PIP installed here. Let me just copy this. Ubuntu tells me what to do. I paste the command. And now it says, start the documentation server. It's really interesting. Let's see what it is. I have no idea, to be honest. But this is my test server. Anyways, I can't break. I mean, even if I break anything here, it doesn't really matter. The good thing about Github or Gitlab or other remote repositories is that you can see exactly what is in those files before you download them. But here, this is my test server, so I will run this command. Python Run Pi. Oh, for me, it's actually Python three. Run Pi. It doesn't have missing dependencies. UCornFast IPI. Should be picked up previously, but let's run that manual. Probably something with the requirements that takes TFL. But let's run again that Python three Run Pi command. Python three, sorry. We can see this server is now running. I can simply copy this. I will paste it into my browser and let's see what it is. Seems like these are the templates available and we even have a search filled, let's say YouTube. What have we got for YouTube? Oh, there are multiple workflows specifically for YouTube. That looks good, and it says, if you want to stop this server running within on your server, you can simply press Control C, which means if I press Control C, this is now stopped. If I refresh, you can see unable to connect. If I go back to my terminal, so if I run it again, it should be now available again and it is. It even seems to have a dark mode. Oh, yeah, it has. Here you search and let's pick any of those, maybe post new youth videos to X. Let's click on that gives you some description. It says download JSON, view, and, there is a diagram as well. We can view the JSON code. That's cool. That means we can copy it here or we can download, but I just copy that, they should be able to paste it on my Anita platform. As I create workflow, there is Control V. That's it. Let's double click to have a look. Let's Zoom to fit. Looks like extremely basic one, but honestly, that's what you would get paying $20 anyways. But it looks like we indeed have loads and loads of templates here available, and that's how easy it is to get them. Just Google and download already available ones rather than paying money to some scammers online. That's all I wanted to show today already. So hope that helps. Thank you, Mark. 10. Bonus - Access ALL AI models free of charge! (ChatGPT Pro and Claude Pro for FREE!): What if I told you that there is a tool that lets you access loads and loads of AI models available today like GPT, grog, Cloud, Deep Seek, and many more, even some that you might have never heard of, and they're all in one place. But most importantly, you can access them all completely free of charge. This tool will also show you which AA models are currently the best and you will also be able to test them yourself. Regarding testing, there is one very interesting way to test A models within that tool where you, for example, ask one question, then you get multiple answers from various AA models, but you don't know which one you simply pick the answer, in your opinion, that you like the best simply, and you will find out only what AI model wrote that answer after you pick that favorite answer. In this video, you will see how to use that tool and why it is available for free. I think after my last few sentences, you might start getting the idea why you can access these AI models free of charge. But let's start from the beginning. The tool I'm talking about is called L M Arena, and you can just Google LM Arena and pick the top link. As you can imagine, the fact that you ask questions and pick what you think is the best answer makes you kind of AI tester for that website. I mean, you can click that what is it? Trophy like World Cup leaderboard icon, okay, in the top left corner, and you will see the score for each AI model displayed in real time. And there are different sections like which one is best for text, for web development. You can also click that top sections like it will show you only for text, yes. And we can see that Gemini 2.5 P is current leader in this section and text to image GPT Image one currently. And when you use this website, you simply participate in that ranking. You have a say which AI model is, in your opinion, better than the other. So basically, you could say that you are an AI tester for that LM arena. Before we enter anything here, before we start typing anything, there is one important thing to understand because it has some implications. You have to be aware that whatever you type here, it will be sent to multiple AI models or multiple organizations like Open AI, anthropic, Google X, et cetera. So you have to be careful what you type in and not to type some sensitive data or some company internal documents. I mean, it should be true even if you use a single AA model like let's say hatGPT. But here on this website, you have to be extra careful to not just send some important data because this data might land in multiple destinations at the same time. All right, but let's talk about this tool then. By the way, Lamarna is not a new tool. It's been there for a while, but it was previously meant to be configured and accessed via developers or software engineers. But now they have that new user interface that is much more user friendly. Everybody can access it and understand what is going on here. We already know that first icon, that leaderboard icon. We know that it will show us the results, which AI model is current leader for given section, but the next one is open New chart. Let's click on that. Now at the top, you have a dropdown menu and you can choose between battle side by side and direct chat. Battle is the one that I already mentioned. It's a battle between two anonymous models. You don't know which models participate, you simply pick the best answer you think or let's say the answer that you like most. So if I click on that, it's chosen by default. Anyways. Let's say, I don't know. Tell me a funny joke. And I think all those AA models do not seem to understand the term funny. So let's see. Why don't Skatonsever go on dates? Because they have no body to offer, right? I mean, you simply have some examples, yes. Let's say I like this joke more or maybe not the joke itself, but also how it's said, yes. So let's say I pick this one. I mean, that's not the only choice here. You can see left is better. It's a tie. Both are bad, or right is better. Maybe I would say left is better. Now, it will well, it prompts like, you know, I have to verify I am human. We can see the last one was coming from Jemma tree model and right one from Gemini 2.5 flash. That's the battle. But now if we choose, for example, the side by side, they are not anonymous anymore because I can simply choose a specific AI model and you can see minstrel, small, Q W Q, command A. There are many different models like Amazon Nova. Let's say cloud 37 Sonnet, right one maybe grog three. I will repeat that. Tell me a funny joke. Let's see what they will come up with. So it's kind of similar, but we know from the beginning which models participate in that, yes, and now we can make the same choice. Right is better, left is better or, you know. But what I want to mention as well, look at this drop down menu. You don't have to just type text here, yes. I mean, you can choose image creation, yes, create a new chat window. And now, well, it basically it's changed the models because for image creation, we have different models, yes. So if I choose side by side, we can see it's GPT image one. We've got flux context Max, flux Context Pro. And some other models like photon that are able to create images, and then you can describe the image you want to create. But basically, the idea is the same. So if I say, create a photo realistic image of an elephant riding a bike. You can see the time the timer starts, how long it will take them to generate that image. And let's see I have no idea what that bugle is, to be honest, but never mind. So photon was first. It took around 30 seconds for photon to create Well, it looks pretty good. I mean, I can't see the puddles, but it is an elephant on the bike. Let's see. Let's wait for that bugle to complete this task. So it took much longer, and I don't really like it because it doesn't make much sense, does it? The front legs are on the pedals. This one, at least the photon tried to hide this fact that it's very difficult to put an elephant on a bike. Never mind, you get the idea. Let's change it back to text to chat. Sorry. Yes, start new session because the models will change again. And now I just wanted to show you the third option, which is not side by side, but direct chat. But this is kind of self explanatory because you simply choose one and you just wear Hunan turbos. Okay, what tell me funny joke. What HunianTurbos will come up with. Let's see. All right, came up with something, but we can see we can't compare it to anything, but we can still use it. So maybe let's see how useful they really are. Be telling jokes is just not a proper task? Let's say, maybe, but maybe side by side, side by side, chat, and let's choose something more popular. Cloud, it's fine. Chat DPT. I want to create maybe HTML website, so maybe Grock doesn't really matter. Let's say, can you create a code for my website? And I mean code can include only HTML, CSS, and JavaScript. Basically, what I'm trying to do. I'm on Ubuntu server, so let's maybe install nginx quickly, and we will just change the index HDML file. That's my idea. But nginx can natively only work with HTML, CSS, and JavaScript. So let's stick to that and I will say I will I will say that I can keep as a single index HTML. And can be served by nginx alone with no back end. It can be any website. No, maybe the one that includes some funny jokes. All right. Presenter, let's have a go and let's see what they come up with. They start Sonnet seems to be a bit quicker. Gro is a bit slower here. Definitely, looked, how much code Cloud already generated. Sorry. Oh, it's completed. Nice. So let's wait for Grock to complete. The fact that Clote we finished first doesn't mean anything yet. Let's see what Grock comes up with. If it's good, maybe it's much better than Cloude. So yeah. In the meantime okay. We will leave it running. It doesn't really matter how long it generates this code. If it's good, I'm more than happy with that. So let's just open the terminal. Let's install because I don't have engine H I? No, I don't think. No, I don't have enginX. I say, sudo up to install nginx, maybe I will dash Y, I will that sudo password. S. That's cool. Let me clear that. My engine should be up and running. I can check with System CTL status nginx. And it is up and running for 13 seconds. Okay, that's cool. So if it's running, it already serves a website. If I go here, oh, look at that. Greg still creating that. No, it's created, but it creates really good instruction, it looks like. All right. So let's open another window. If I run HTDP local host, I should have, yes, that's the response from EnginX. If I go here, if I may be clear again, clear, Cd var HTML, LSL. So this is basically this text index I mean, what we can see here, welcome to Engine X. If you see full page, blah, blah, it's this Engine X, if you see this page. It's basically this single file, index dot nginx dot wb. It HDML. Now we can just copy. Alright, let's copy Clouds code first. Generated it much quicker. So copy the code, and let's VIM. We will change sorry, pseudo. We will change that file now. I don't have VM. I mean, I could use nano as well, but well, I'm already installing VIM, so never mind. Because I thought VIM is there by default, but it looks like it's not. It's a fresh server. There is nothing running on it. So again, so do VIM, D 99 d, it will remove everything. 99 lines it will remove. We will paste the code we got from what? From Cloud Sonnet. I press call on WQ to save this file. And now, if I go to HTTP Local host, oh, that's the website. Random joke generator. Tell me a joke. What do you call a fish wearing a bow tie? So fish decayed. All right. Tell me another joke. Well, never mind. You know, this is what it looks like. That's what this website looks like. It's not bad, is it? Because it took it like 10 seconds to generate it. Okay, I thought maybe it would be worse, but it's not bad. Let's see what Grock created for us. Whoa, it's a long code. Okay, let's see, copy, and we will now paste this code. I will say psudo again, the 9990 because there were many lines, so I will remove them all. I will paste this one. Call on WQ. And now let's refresh and see Whoa. Looks better. Okay, so that's what I was talking about. Oh, I took longer, but look at that. It looks much better. Oh, they're going up and down, those faces, those little faces. Let's see if the jokes are actually funny. Get a new joke. Why did the banana go to the doctor? It wasn't appearing well. All right. I've heard words. But generally, this website isn't bad, is it? It's very simple. Has two colors, blending in the middle, not bad at all. Okay. That's basically it. This website isn't very sophisticated, but you can see you can achieve a lot in very quick time and you can do it for free. 11. Hosting n8n in AWS intro: The next three videos are optional to watch. We will go through the process of running NA ten software in AWS Cloud, and I will show you how to run that NA ten server in AWS Cloud free of charge for a year. If you wonder why would you want to run your server in AWS? Well, this is simply more professional production ready, let's say, approach, where you run NATen services in public Cloud rather than on your local machine. It might be useful for you to know how to configure it all from scratch, if you or one of your customers need that more professional setup. You simply know how to configure and host that server for yourself or for other organizations. There is also additional benefit because running it this way makes all external connections to your NATN server possible because we are going to buy a domain and we will use that so called fully qualified domain name to run our service. I know that it might sound confusing at the moment, but don't worry about it. This is again, step by step tutorial, so we will learn all of that from scratch again. In the first video, we will create our own AWS account. We will also talk about something called a free tier, the thing that lets us run that server for a year that I mentioned for free. Then we will talk about multifactor authentication then in the next video in the second video regarding AWS, we will talk about so called AWS budget and that budget is used to monitor our expenses if we have any because hopefully we'll run everything for free, but maybe you want to run some paid services, and that's where AWS budget comes in handy. Then in third video, we will install and configure our NA ten server in AWS Cloud. Let's get started. 12. Create AWS Account (1st pre-requisite for n8n in AWS video): Today, I will show you entire step by step process of how to create new AWS account. This is new fresh operating system with absolutely nothing configured, and we'll be creating here new AWS account. Let me open the browser and we'll navigate to something like AWS three tier. And what free tier is, we'll talk about it later. But now, let me just explain shortly. It's simply AWS offering that lets you using loads and loads of resources for a year for free. When you create new account, you can use those services for free for one year. But we'll talk about it later when we actually create resources. Now we're creating the account itself. So as you can see here, I have create an AWS account. Let's click that because that's what we're interested in. It takes me to that web page with the create free account button. Let's click on that. And maybe let's accept those cookies. Now what I need is some kind of email address. It can be any email address use or you own, you just have to be sure you can access that email during the signup process. I will just use gmail.com. This is important bit. The one below, it's just an alias for your account, so it can be anything you want it to be called automation Avenue, but it's not that important really. The email is important here. I click Verify email address, and now I will have to just wait for the email to arrive. As you could probably hear, the email has just arrived, and I've received a verification code. You have to type it here and click Verify. Okay, now we have to create the password for root user. We'll talk about it a little bit later who is actually root user, but now let's just create a password. And as you can see, it has to include uppercase, lowercase, number, and non alpha numer character. Just make sure you've got that all. Even when I just start typing, let's say, capital letter, as you can see first box is ticked the lower letter, now digit, and now non alphanumeric. As you can see, all those boxes are ticked, so my password is good enough for them. Now I have to repeat. And that's it. We can continue. Just wanted to note here on the left, you have a link if you want to explore those three tier products. This is the link you can use. But never mind. At this stage, we'll just concentrate on the sign up process. I'll just click Continue. I will save the password. And now it asks me for contact information. It asks me if it's business account or personal account. I will treat it as personal account. Anyway, if it's business, it will ask you for organization name as well. Let's click personal and my name is. You have to give them name and address, and the phone number. And you agree that you've read the customer agreement. We can go further to step two. Now it asks you for billing information. That's why you have to provide your credit or debit card number because as I said, there are some free services you can use, but by signing up, you can use all services, the ones that are free, and the ones that you have to pay for. That's why you have to put the credit card details. But in later videos, I will show you how to create budget. This budget will inform you if you started using some services that you have to pay for. For the time being, we have to put that information in to be able to progress. So that's the credit card, and we can go further. As I used my vote card, I have to confirm that in the banking application, I click Confirm and that authorizes this step. Now it asks me for a telephone number to use to confirm the identity. I will use the same that they used previously on previous page. Now, the capture, cop I have it. If it's completely messy, you can also always refresh it. It will generate new one because some of them are generated in such a weird way that you can barely see what's there. Okay, now let's wait for the verification code sent to my phone number. And as you could hear, it's just arrived. And we can continue to step number four. Here it asks me again what type of account I really need. So we'll go, obviously with the free one and want to pay $29 a month or 100. The free one is more than we need really, and that's it. We click Complete, sign up. As you can see, it was quick and easy process, and it says, Oh. Okay, as you get here, AWS says, we're activating your account, but I believe it's just been activated. So yes, received the email, it said that I can start using my account. Well, I can click either this button or that button doesn't really matter. Let's maybe click this one. Go to AWS Management Console. And now it asks me if I want to sign in as a root user or the IAM user. We really only have root at this stage. So for the time being, as there is no choice. We only have root user. We'll put the email address we used for the signup process. So for me, it was gmail.com. We'll click next and the password, the one that you just created as well for this user. And sign in. Not bad. We are already in our AWS console. That's what it's called AWS Console, where we can do everything regarding this account or our resources. Thing I want to show you it's here that says automation Avenue. Remember, that's the alias we actually pasted during the signup process. So instead of some long weird digits, automation avenue will be used here. For you, it can be anything you want to name this account, us. The next thing is we can see stockhm. Well, you chosen, I don't know why, but that's what we call AWS region. This is where you want to create your resources. You can see you can create it all over the world. Let me just change it to London because that's the closest for me. What I want you to do is to activate multifactor authentication for the root user. That's very important. This account has to be very secure. What you can do here in this search field, you can put IM IAM. We'll click that and you can see you can see that even the Amazon itself, there is a security recommendation one, add MFA for root user. To be almost every single user you create should have MFA enabled, in my opinion, but root one is definitely one you need to create MFA authentication for. We can add MFA here, clicking this button. And now we have a choice what we want to use as a second separate authentication method. The easiest way, I think, is authenticator app, which is chosen by default anyways. You can call, for example, for me, it will be my Samsung phone. I can use that as an doesn't really matter. It's just for you, information for you. We can choose authentication app. Click Next. And now you can see, you'll have to install one of the applications on your foam, either Google authenticator, do a mobile or some other apps that the list of all applications, you can see here. But basically, you need just one of them. Google is really good one. I can recommend that one. And then once you've installed that application, you have to click that Show QR code and just scan it with your foam. Once you scan it, you will receive MFA code, which you will have to paste here twice. I mean, first code and second code. This way, you will add that device as authorized one to receive the MFA codes. Will not do that because this is just temporary account. I will remove it later on, but you definitely should do that. Okay, so I think at this stage, I don't want this video to be too long. So I think that's it for today. In next video, I will show you how to create that budget before you start actually creating any resources in AWS Cloud. But you've got your own AWS account now, so congratulations. See you next time. 13. AWS Budget (2nd pre-requisite for n8n in AWS video): In previous video, we created our new AWS account. We created and logo as a root user, and we added MFA authentication for that root user. What is next most important thing is that we should know how much money AWS services will cost us, if any, we can either try to not spend any money there just using free tier eligible services, but maybe we are okay to spend some money, but we don't want to exceed certain threshold like maybe $10 a month, let's say. But one way or another, we definitely want to be in control of our budget. And that's what AW budget service exactly is for. Being logged on to our console, still as a road user, we can type here in the services, we can search for budget. And we can see here budgets as of bidding and cost management feature, or we can simply click that bidding and cost management service, which budget is part of. Maybe let's click the top one. Doesn't really matter. Here, you usually see some summary how much last month it cost you, what's prediction for this month. But because this is new account, there is no data available yet. But it doesn't really matter because what we need is budgets here in the left down corner. So let's click that budget as you can see, we can create a new budget here by clicking this button. Now we have a choice of using a template or we can customize using advanced settings. Let's stick to just simplified version. Now, what type of template do you want to use? By default, you can see zero spent budget. Create a budget that notifies you once your spending exceeds $0.01, which is above the AWS free tier limit. That sounds good, doesn't it? Because that means if I spend any money on anything, I will get a notification. Also, that's important to remember, the AWS budget will not disable any resources for us. It's not meant to be. It will only notify us every time we exceed certain threshold. We can figure here. So the first threshold will be that $0.01. If we use any service that we have to pay for, it will send an email to the email address we specify here below. Maybe before we go there, there is a budget name as well. It's called My Zero spend Budget, which is okay, but let me just personalize it. Maybe Mark. Zero spend budget. Doesn't really matter. It's just a name. And here is where we enter the email address. And as you can see, it doesn't have to be one. You can put multiple email addresses here. You just have to separate them using commas. I will just use one, let's say, Mark at Avenue. That's it. If I want to another one, just use coma and blah, blah, blah. But we'll just use this one. That's really, I believe, because yes, we can leave everything else as it is, and you can see a confirmation. You will be notified via email when any spend above $0.01 is incurred. That's fine. That's what I want. We create budget. And that's it. We've got our budget, and don't ask me why it always shows $1 rather than $0.01. I don't know why it is like that, but it should be for $0.01. But that's not a big deal. It works as expected, actually, but maybe we want to create another budget. What if we want to sometimes use some services, but we don't want to exceed $10 a month. So we can create just another budget and set threshold to that. So let's click Create budget. We'll leave it as a template simplified, which is much easier. And now we will change from zero spend budget to maybe this one, monthly cost budget. What's interesting about it, as you can see, it notifies you if you exceed or are forecasted to exceed the budget amount. If you start a new month and on the first day you exceeded $1, this AWS budget can see you will actually exceed $10 by the end of the month. It will send you notification before you actually reach that $10 threshold because it's forecasted for you to exceed that by the end of the month. I hope that makes sense, because our ultimate goal is not to exceed $10 a month. That's exactly what we need. So now we can scroll down. Let's call it $10 budget, yeah. It's clear for us what it is about. Now, we just adjust this value to $10. Again, just list of emails. We want that notification to be sent too. And that's it. And here below, you've got a summary of how it works. You will be notified when. Your spend reaches 85%. So if you are at 8:50, let's say, you will get email, then it will be at 100% at $10. But the third option, if you're forecasted spent is expected to reach 100%. But that's very handy and useful for us. That's exactly what I need. Just create a budget. This way, you can create as many budgets as you want and get notified every time you reach any of those thresholds. 14. Run n8n in AWS Cloud! (FREE for a year!): If you want to run NA ten automation software 247 in your own AWS Cloud, then you are in the right place. I will show you step by step how to build and configure a virtual server in AWS and then how to set it up to run that NA ten software for us. What's the best part, I can run continuously for entire year within that AWS completely free of charge. And that's why I chose AWS as the provider for today's AWS is very reliable and it offers so called free tier. For the first tier and for the newly created accounts, you can run stuff free of charge. Even if you want to keep that server running beyond that first free, it only then cost four or $5 a month anyways, after that free, I mean, it's not the end of the world either. Note that we already had a video showing how to run NATN locally, but that was a very basic setup where you had to be on the same network as your NATN server to be able to run and create new workflow. That was very quick and easy to build setup that you could run on your laptop or PC, but you will have a problem using external web hooks or triggers or to be honest, any traffic that needs to be initiated from outside of your network. You'd have to configure some additional third party tools to make it happen. What we are building today is something different. It's a solution where the NAN instance is available over the Internet, so you can access this server from anywhere in the world, which means you don't have to be at home or on the same local network to be able to access your NATN. Also show you how to buy and set up a domain for our NATN service, so you can access it using something like hdpsnten dot automation avenue.com instead of typing the IP address and the port number. By the end of this video, you will have NATN platform up and running in AWS. So let's get started. The first thing you need is obviously an AWS account. It's very easy to create. I have a dedicated video just for that process. If you want to follow this video step by step from now on, maybe watch that previous video first, get your AWS account ready, and only then go back to this video. I would also say that there is another prerequisite. Although it's optional, I would suggest to set up so called budget in AWS. Because even though our virtual server will be running free of charge, you might be tempted or you might want to use some other paid features within AWS, and that AWS budget service is there just to make sure that you are notified of any charges detected by AWS account. It's very quick and easy process to get that AWS budget configured, so I would highly recommend that. The fact is, I will use one of those paid services today because I want to buy a domain for our NA ten platform because even though it's possible to get a free domain, free of charge, I prefer to get one directly from AWS for two or $3 a year and have a proper root domain rather than some free sub domain from some third parties. As you can see, I already have an AWS account. I already have budgets configured, even more than one, so we can now start building our server for the NA ten automation software. First thing I need to do is to go to that EC two service. It's Elastic Compute Cloud within AWS. That's where you create your virtual servers, and then you choose the region where you want your server to be built, it's your choice, but I will choose a region that is close to me, which is London, EU West region. AWS will switch me to that region, and now I can build my virtual server. I say Launch instance, this orange field. And now we are presented with that web page. We can choose a name for our server. I will just call it Aton. It doesn't really matter what you call it. It's just so you can recognize what it is. Then we scroll down and we can choose our operating system and they want to run it on Ubuntu operating system. And our method today will work for any Ubuntu and Debian operating systems, and with a little adjustments, it can work for any other systems, really. But you will see later on what I mean. I choose Ubuntu, and I will choose Ubuntu version 22 oh four. Even though 24 oh four is newer, I think 22 oh four is more stable. It says that some of the changes will be removed or blah, blah, but it doesn't really matter. You just confirm the changes, and this is our operating system now chosen for our server. So I scroll down. Here is the instance type, which is the size of my server. But now let me go back to that free tier. You can see here that the free tier in your first year of opening AWS account, you have 750 hours every single month of T two micro instance or T three Micro where T two isn't available, and have a look here. It says free tier eligible and that's what you want to choose. The one that you can run free of charge for a year. For me, it's T three Micro, but you have to make sure that your operating system is also free tier eligible and Ubuntu indeed is free tier eligible. I've got operating system, I've got server type and you can even see here. Even after that year, it costs just $0.01 per hour to run it. Not bad, isn't it? But let's scroll down keeper. Keeper is so called SSH key that you need to be able to log on to your server remotely. I don't have a keeper. I've got some, but it doesn't really matter. You will not have any keepers. What we need to click is that create New keeper. I'll click on that. We can choose a name for it. I will call mine NADM keeper. Something like that. Again, doesn't really matter what you call it. It's just a name. So I say, create a keeper. And look at that. I will download so called private key to your local server. So you will find it in downloads ana ten keeper dot p. We will work with it later, but just remember that this has been downloaded from AWS automatically to your local server. Whatever device you're using right now, I mean. That's the keeper, keeper sorted. We scroll down, and now we have network settings. And in network settings, we want to make sure the public IP address is enabled. We will need a public IP, and there are ways to run this server without public IP. You can, for example, configure Cloudflare tunnels or you can use AWS Cloud front service and keep this server behind those services, and it is considered safer, but the way we will set up our server today can also be considered safe solution. Why? Because allow SSH traffic, I will lock it down from anywhere, not that anybody can just log onto my server. I will change it to my IP. I will be only able to log onto my server and manage my server only from my current IP address, which is my home location. So I say, Allow SSH traffic from this IP. What's this IP address? If I go to the Internet and I ask, what is my IP? That's basically it. This is my IP address. We can see it finishes with 78. And indeed, it's here 1778. That's my IP, and only I can access this server. That makes it pretty safe. We will want to add another option though for our security group. We want to allow HTTPS traffic from the Internet to our server because that's how we are going to access our NA ten service over HTTPS protocol. So with those two settings, we are now fine. We also have public IP. We can scroll down. By the way, that public IPV four address that we configured that we left enabled, that's also free of charge, or I mean, 750 hours per month are free of charge. Here for storage, we can see that 30 gigabytes of IBS storage is also free of charge. By default, we have it here set to eight, so I will change it to 30 because that's still free of charge. I will change the type to GP three GP means general purpose, and you can see even here, free tier eligible customers can get up to 30 gig of IBS general purpose, SSD. P three is a newer version, so I chose GPT, and we have to click one more thing is this advanced tab and here that little arrow, and I want to have it encrypted, and it's not optional. You have to have this so called root volume. You have to have it encrypted. Otherwise, you will get error when creating a virtual server. It simply needs to be encrypted, and now that's all we really need. I can now click Launch Instance. I just do that and my instance is being created. It will only take few seconds. And already done. It's a successfully initiated launch of the instance, and this is my instance ID. If I click on that, it will take me straight to this instance, which means instance is a virtual server in the cloud, and that's it. I have all the details about my virtual server. I have two IP addresses allocated. This is so called private IP address, but I'm more interested about my public IP address. And my public IP address is 134020 eight.247. If I click on those instances here in top left corner, I can indeed see my virtual server. The state is running. However, the status check is still initializing. It might take a minute or two before it's fully up and running. But we don't have to wait for anything. Once it's ticked here, I can choose this connect, but I can click on that, go to that SSH client, and here we have a little instruction. It's instruction on how to connect to our server. It says open an SSH client, and for me because I'm also Ubunto, this is my laptop, local laptop. It's also running. It's MgBok but it's running Ubunto in parallels. Bit complicated, but never mind. Basically, if I go, first of all, if I run command PWD, that shows me my current location. My location, my folder I'm currently in is home parallels, so I can see the to downloads. And here I can run command LS L, and I can see that private key that was created during the process in my Downloads folder. I can also, I mean, you know, you can also click these files. And if I go to downloads, that's basically the same location, right? It's atenkeeper dot pm. Not sure what this file is, but we are concentrating on that one. And that NA ten keeper is a private SSH key I want to use to connect to my server. And it says that to run that if I want to connect to my server, I actually have to run this command first, change mode for 100 and then this file. I will copy this and being in the same folder where my file is, I will just paste this command. What it does, if I run LSL, it changes the permission. It now only I mean, this means that only the owner can read this file. Previously, it was read write permission. So it changes the permission for this file. And it says that now I can connect to my server, and it gives me this example, SSH. That's why I can copy it and paste it. But the fact is I will shorten it a bit here because I don't have to use DNS, I can use public IP address as well. So I can do this use just the public IP, and here is the location for this file. And full location is home parallels and this file. I know it's a bit confusing, but I can simply write something like Home parallels. This command, this full command, I can run it from anywhere. I don't even have to be in the same folder where this file is now because I specified full path. And if I run this command now, it should take me to my server. I should log on to my server using this command. That's presenter. I say yes, you have to say yes. It asks you if you are sure what you are connecting to. Yes, I know what I'm connecting to. Oh, it says permission denied. Oh, you know what? I'm stupid. And you are probably watching that. Mark, what are you doing? It's not this location, home parallels, but then our key is in downloads. Yes. Of course. Little mistake. Let's add that Downloads. Basically, if I go here, sorry about that. This is where my key is. And the full path, if I click here, is home parallels, downloads, and then this file. And now you can see a little log here as well. That means it's locked only. It can only be read by current owner, and that's good. That was added when we changed when we run that change mode. Command. All right, let me try it again. Anyways, press Enter. We are going straight to our server. This is our server. This is private IP address, 18 151. I can double check. Let's click on that instance. Private IP is 18 151. Correct. If I run DF H, I can see indeed it's 30 gig. It's shown as 29 gigabytes, but it's basically 30 gig disc that we configured during the configuration process, which means, yes, it's indeed my server in the cloud. And I can log onto it remotely from my home because I'm logged on from my laptop, from my MacBook, and my server is in AWS Cloud. Maybe I will show you one more thing. Let me exit that. The command is just exit. I'm back on my laptop. I will use up arrow. This was the long command I had to use to log onto my server. Yes. But what I can do, I can run command called alias. And then I let me make it bigger wider and I can call that alias something, maybe NA ten. Then I say equals and then a single quote. Then I have to go to the end and also add a single quote, which we will make all of that one long command. Now I press Enter and if I run now alias, that's basically it. That's my alias. What this alias does, I can now simply type NA ten. Instead of running SS one, two, blah, blah, blah, I can now simply say NA ten, and that NA ten will run this long command for me. Let me show you I press Enter. And bam, I'm straight back on my Cloud server. Let me clear that maybe, and now it's always considered a best practice if you run sudo UtgUdate and Sudo upto get upgrade. And then I add Y, which will answer yes to any possible questions. And what it does, it will simply update all packages and all software that this current server is running on. I just press Enter and I just wait. That upgrade is optional, I would say, but up to get update that first command is definitely something you want to run every time you build a new server. Now it asks me to restart some services, which is fine. I just use tab buttom to switch to that o, and now press Enter. And that process is completed. But which is another good practice, you want to reboot the server right now. I say, pudo reboot. Press Enter, you will be obviously disconnected from the server now, but you only have to wait 20, 30 seconds maybe, and then we can log on back to our server. And remember, now, I only have to run command eight N, which is our alias for that long command. Maybe 15 seconds past. Let me try again already. Press Enter. Oh, we're already back in. If I say uptime, our server is up 0 minutes because it was just rebooted. That's all for our server build. Now, we start configuring the NA ten service on this server. What I will do then, I will go to Google. We are done with AWS for now, but we will get back to this service. For now, I say NATen AWS, something like that. I just simply need official documentation from NATen. If I click on that first link, it will take me to server setups for Amazon web services, and you might think that this is what we need, but this is to run NA ten on another AWS service called Elastic Kubernatis service. Kuberntis is like a cluster that can autoscale. That's not what we need. What we will use today is a Docker Compose option. I will click on that Docker Compose, and basically, this is the option that you can use for any other provider as well. I configured the server in AWS, but you can use Docker Compose approach for any other Cloud solution. Like if you have your server in digital ocean or Hetsner or Azure, you can follow exactly the same steps like we are doing here. Let's have a look what they say here then, and they say, here, install Docker and Docker Compose. There are many commands, but what you can do, you can simply click on that square icon that says Copy to Clipboard. We will copy all those commands, and in fact, first command is also UgtUdateeven twice, you can see, what it will do, it will install everything we need on our server. I mean, Docker and Docker compose, at least. I go back to the terminal and I simply paste it. All comments are here, I just press Enter and white. Looks like they didn't add that Y. That's why you have to manually type, yes. Yes, I want to continue. Press Enter again. And that's job done. We go back to that website. We have installed Docker and Docker Compose. So scroll further. It asks to verify if the Docker and Docker Compose were installed properly. We can run Docker version. We can basically click that copy to clipboard again go back to my server, maybe clear first and paste it there. And we can see we have Docker Version 2833 and Docker Compose version 2391, which means they are both installed and ready for us to use. Now we can scroll further. It's optional non root user access. You know what? It's a good option. You will not have to use that studio every time when you run Docker commands. It's not a big deal, it's optional, but I will just copy to Clipboard. Go back again, paste, enter, and that's job done. We can go further and here we have DNS setup. It says to host NA ten online or on a network, create a dedicated subdomain pointed at your server, and I know it might sound scary, but it's actually pretty simple process. But what we need is a subdomain, and to get subdomain, we need to first buy a domain. As I said, there are ways to get a domain free of charge, but I'm not really fan of that approach. I'd rather pay two or $3 and have a proper domain from AWS. I will do then, I will go back to AWS. I will need different service now. The service is called Route 53. You can search here for the Route 53. You click on that, and you are presented with that web page. Let's click on that Get Started. And it says register a domain. You have other options, but now we exactly need this option, register a domain. So I again say get started. And domain is like google.com or youtube.com. I will search for NA ten. Let's see what's available because it says check availability for a domain. I will search, and I can see a ten.com is not available. Somebody already purchased that. But here below, we have some options, but I already can see they're all expensive. I'm not going to pay $15 for a domain. I could have it asnaten dot red for 23 but I'm not happy with these prices. I can take here those standard pricing, and here I can sort them according to price. And now you can see that domain that finishes with dot click is the cheapest. It costs only $3 a year. It's a price per year, so it makes it $0.25 per month, yes. Let's see if ten dot clique is available then. I will search for ten dot clique, and unfortunately it is not. Enten dot clue is not available. But n9n click is available and it costs $3. I know it might not be perfect pronunciation nninndt click. But I don't care. You can search further and further, personalize it more, but I will just pick this one. $3, I'm happy with that. I say select. It says the domain registration fee is $3, and I'm happy with that. I proceed to check out. And here you can choose if you want this domain to be automatically renewed every year. No, I'm not happy with that. This is basically just my test, so I will just run it for year. It will cost me $3, and close to the end of that year, I will get a notification that this domain will expire if I want to renew it or not, but I will not renew it, so I'm okay with this setting, and it can simply expire in one year. I just say next. And contact type, I say person and I put my personal details here. Fill that in. I say next. Now just a confirmation. I say I have read agreed to domain registration agreement and I submit. It says the request is in progress, and it might take several minutes, but usually it's pretty quick. I can refresh it here. I clicked Hosted Zones and it's already shown in Hosted zone. Because when you register domain, you will also have Hosted zone created and it looks like it's already here, if I click on my domain, this first part is already done. It might take few more minutes to be shown up in registered domains as well, but really hosted zone is something even more interested. You will see shortly why because if we go back to that website and follow the instructions further, what we now have to do is to create a subdomain and point it to server IP address as A type record. That's what they say, and all that means is that I have to go back to my Route 53 to hosted zones to my domain n9n click, and I create a record. And they said it needs to be A type record, which is chosen by default anyways. The value is the public IP address of our server. So if I go back to my server, this is my public IP address. And by the way, if you go to the server itself, if you run curl I configure M, that's basically the same information. That's your public IP address. I will copy it from here's dot 20 eight.247. If I go here, 20 eight.247. It's alternative way of getting your public IP address. And now I go to that Route 53 and they want to paste it here. And now this is my domain. It's called ninndt click. And whatever I put here is so called subdomain. You can add multiple records by adding something in front of your domain. I will call it maybe server one. So basically what I create now is server onennt click. And that record will point to this IP address. That's what they want. I create records, and this record has been created for me. We can see it here, which means now I can go back, do compose, configuration setup, and we can follow further steps. It says, create dot ENV file. It says, create a directory, make directory, and then CD into that directory. All right. I'll go back to my server, but I'm currently in home directory on the remote server, which is home Ubuntu. And it's up to you where you want to create that folder they are talking about, but it should be somewhere within that home directory, I would say. So I'll go back and I will copy both those commands. Place them here. This one created the folder and this one took me into that folder. If I run PWD now, I'm in home Ubuntu and I ten compose folder now. Let's go back and they say, inside the NATen compose directory, create dot NV file to customize your NATN instance details. What I want to do is copy this all of that. Again, I go back to my Cloud server and I say nano dot ENV. Nano is a text editor. It will create dot NV file for me. Enter, it will be empty though. So I can write click and paste what they asked us to paste. But what I have to change now my domain name. I bought a domain, yes, but the domain name was called nnn dot click. This was a domain, and server one was a subdomain. I have to go back and they say, subdomain is server one. And now the time zone, this is not correct for me, I am in Europe but in London, and that's where my server is created anyways. And as I sell email, I also want to change to the actual email I want to use for that. That's some testing email. I created for that purpose, and I say now Control O to write it out, now Enter and Control X to exit. This file is now created, and I can run Catt ENV just to check if everything is correct, but it looks like it's correct. Nninendt clqu is my domain. Server one was my subdomain. Time zone, you can look up. You can Google simply what are available time zones. I remember that mine is Europe, London, and email whatever email you want to use. That's it. So we go back. This is done now. Next step, they say, create local files directory. Inside your project directory, and we are still in that project directory. This is what they mean NATen compose. We created in previous step. We're still there, so I just copy that command, I paste it. If I run LSL now, we have created another folder inside NA ten Compose folder. Let's go back and let's go further. It asks us to create compose dot YamlFlePast the following in the file. I copy everything again. I go back here, I say nano compose dot Yamal. This will create file compose Yamal, and I will be able to edit it. I press Enter and I paste all the output again. It's pretty long output, and what it does really, it installs and runs two Docker containers, traffic and NA ten, and it uses the so called variables that we configured in previous file dot ENV. For example, this SSL email, this value will be read from the previous file. That's why it's so important that we have it configured correctly. And if we scroll down, you can see subdomain, domain name. This is also information. And here again, domain name, this is information that we also included in that previous dot ENV file. And what it also does, even though the NA ten itself runs on port 5678, this Docker compose we secure our traffic, and we can connect to our server via traffic on port 443, which is HTTPS, and it's considered secure. But this is all the extra information. You don't have to worry about it. You simply can run again, Control O, enter Control X. And if you had everything configured correctly in the previous file in this one, you don't have to study really what that composamel does. Your convenience, I also created a PDF document that you can download free of charge, of course, that document will include all steps and all commands that I run here today because I know it's quite a lot of information involved here in the process, but you can just grab that document using the link below, and I hope that might help you to follow all of these steps involved. We go back to these instructions, scroll further, and this is just little instruction, what it will do. But if we scroll even further, we can start Docker Compose now. It's a pseudo Docker Compose a D. I will copy it again. Let's go here, paste, press Enter, and everything should work. Honestly, we didn't even have to run that sudo because we run that previous command, which should run Docker compose with no sudo, but I'm lazy. I just pasted it, so it doesn't really matter. The images are being downloaded for traffic and for NATM and once we have them locally, the service should start. It might take several minutes even let's just wait. It took less than a minute actually for me, and it says, we have two containers running. We have two volumes created, and one network was created. And we can confirm it. We can say Docker ps. That shows us that we have two Docker containers running, indeed. One is NA ten and one is traffic. They're up for 41 seconds, and we can also run Docker volume. LS. And indeed, we can see that two Docker volumes were created. That makes this service up and running. And remember, we're still on this Cloud server, but now I should be able to access my service from anywhere in the world, not only from this laptop, but from any other laptop as well. And to access my NITM, I should simply go to this website. Let me copy it. Let's have a look. HDDPS Column forward slash forward slash, and I paste my server one dot ten dot click. Press Enter. Some browser will show you this information, some will not. You have to go to Advanced, accept Risk and continue, but it's really not risk. I will show you what it is. But you can see it works. This is my NA ten platform, and I can use it from anywhere just by going to this URL. And what that error was about, have that little log. Says connection not secure. It is secure. But if you click that more information, it says this website does not supply ownership information, and my domain has ownership information. It's simply how it's resolved here in that docker in that traffic and how let's encrypt gives us the SSL sert for this service. It's a bit complicated. But what I mean, some browsers will display that error, but some browsers will simply ignore that and you will not even notice anything. Believe me, this connection is secure because it says HTTPS. You have SSL ALS encryption for that traffic between you and your server in the Cloud. And that is our setup completed up and running. So I hope that's helpful. Thank you for watching. 15. What next ?: Might be wondering, what other videos or courses I should watch to progress further? The answer is you don't need to watch anything else. From now on, you should focus on running your own NATN software and building your own solutions. This is a very common mistake. People, sometimes including me, still, I have to admit, often spend too much time watching tutorials instead of taking their current knowledge and start practicing with real life solutions. The fact is you already know how to host your own NATN software. You know what nodes and workflows are, how to build an AI agent, and how to work with APIs or connect to other services. You are now ready to work on your own solutions. If you don't know what you should build, then head to NA ten website and search for workflow templates. You will see thousands and thousands of examples and ready to use templates there. You can sort them by popularity or customer reviews to get an idea what other people are building and what the current hot topics are. Of course, you will stumble upon some problems, but that is the most efficient way to learn and to become professional. You have to know how to resolve issues because there will always be challenges when creating more and more complex workflow. Remember to use official documentation from NITM or look up solutions to specific problems you might encounter. But from now on, the best way forward is to create your own workflows. Start by building something simple and then expand on it. You might also think about what automation other businesses might need. Maybe you even go and ask some people running local businesses what they need or what tools would help them to save some time. So yeah, that's all I wanted to pass in this training, and I'm really happy that you have completed it, and I hope you found it useful. Thank you for watching then, and I hope to see you next time, Mark.