Transcription
Hello everyone. It's Misha, Superninja from Gatova. Today we'll talk about the automation system N or N8N, as it's called by the people. 3 weeks ago, I released a fork, or a local version hosted on our server, for everyone for free. After that, I saw that there is huge interest in this automation system from the public. In the first 24 hours, over 500 people registered, and 1,500 automations were launched in the very first days. I haven't checked since, I don't know, but it's clearly significantly more. It's evident that people are interested. And, you know, at that moment, I was discovering this platform for myself. I hadn't worked with it before. I was writing something of my own, something similar. Of course, simpler. M, but here I tried working with N8N, deploying it, figured it out. As a result, it all grew into a much more serious solution. I will tell you about it now, I will show you. The only thing is, during the demonstration, please don't judge too strictly, because I've only been working with the system for 3 weeks. I first encountered this automation system 3 weeks ago. I had heard about it before. You asked in the comments in our Telegram group. And so I got to it and decided it was time to master it all. The first thing I did was deploy it locally. The second thing I did was make it available for free, without subscriptions, without any conditions. You log into your account, get your credentials, and use it. How does it look in the DVR system? Well, for those who don't know yet, to use it, you still need to go through a quick, simple registration in DVR. DVR.com is a platform for it, it's a multi-tool for various AI solutions. As soon as you register, you'll land on a page like this, your QR code and all your various AI solutions. To start working with N8N, with N, click on Tools. In the tools, at the bottom, you'll find N8N, N8N. This is how it looks inside. Pay attention, if you haven't used the system yet, there's a special link. It's created automatically. You are also automatically given a login and password. They are generated automatically. You must use them. When you follow this link, as soon as you follow it and enter these details, you can immediately start using the NIT automation system. Here, on this page that you will see, I have also made some important things that will be useful for you in the future. This is process monitoring. Here you will see your workflows and can stop them if one of them gets stuck, frozen, or something is not working correctly. You can control them. In fact, this is your load analytics system. Also, I will tell you about this right now. But immediately, since we are on this page, you can install your own nodes. I figured it out, read the documentation, wrote my own nodes so that DVR can be tightly integrated into N8N. You can install them via this link. We click, we land on a familiar page. And here is this installation command, if in the terminal, you can download it, or you can simply enter this name through the search within N, and it will find it, install it, and you'll have everything. I will show you how it will look. I'll just show you right away that you have all the access, passwords here. Also here, I have prepared several ready-to-use workflows for different tasks. You can download them and use them immediately. There are also generators here that will allow you to generate ready-made workflows according to your requests. You fill in the fields, click generate, it generates, and you use it immediately in your automation. Okay, that's all, now let's get to work. I'll show you now. Look, when we enter the system, the somewhat unfamiliar appearance might initially confuse us. I'll switch back to the familiar one now. Look, usually N8N looks like this. Here are our pages, our automations, our folders, our workflows. It looks like this. But, there are two themes: light, dark. I initially kept the dark theme, but then I realized that, well, I don't like how it looks. Too many thin gray outlines on a light background. In my opinion, it's just awful. It just burns your eyes. Therefore, the first thing I did was rewrite the design. I rewrote the entire design. Well, and since I started rewriting the design, I removed all unnecessary sections that are superfluous, redundant, which cannot be used because the system creators ask us to pay them money. So I'm back now. Therefore, you have this little brush here. If we click it, everything will turn into this style. Here, as you can see, I've removed everything unnecessary. There's nothing superfluous, only what you need. And you can already do everything, start working without delay. Now I'll show you what's available now. For those who already work with DVR, know that DVR is a complex system that has assistants and various solutions for different AI tasks. So, the time has come when we can apply all this to our automation systems. Now I'll show you how it all works. Look, when we get to the constructor, I'll probably even show you in the familiar color scheme so that it's more familiar to you. Many of you watch YouTube videos where you learn all these things, automation, this whole system. So, well, you all know that when we go here, we do two things. We can create a working workflow, some specific automation, or we can create a specific, important, useful thing like a credential connection. And that's where we should start. When I click Credentials, I can connect. Many of you know, for example, how to create a Telegram bot. I enter Telegram here, then I enter its token, which I get when creating the bot. And everything works for me. It's not very difficult to create bots later. And now I present my own credentials that I made. We just enter DVR here, and we have two types of connections. For those who watched my previous video, they understand that these things are based on API. API is a query language that allows dynamic, fine-grained work with artificial intelligence, with any software solutions. So, and in order for me to show one, I'll show the second. So, let's create a test assistant. Look. DVR Assistant API. I select it. Continue. What does it ask for and require? A key, as usual, right, common, and a database ID. This is something new. Many of you haven't encountered this. In DVR, I'll explain again for those who don't know. The DVR system is a large system with many AI solutions. One of the most popular solutions is the assistant constructor. The second is our own ChatGPT, which has all the models. I will also talk about it in this video because it is also involved here. So, the assistant constructor, if we go into it, we will find ourselves inside an office like this, where we can create our own assistants. And we can create our own assistants from scratch, give them a name, settings, prompts, description, choose models, or we can choose ready-made ones. Well, for me to be faster and easier, I'll choose ready-made ones. This is Groks. This is a somewhat humorous, original assistant. In principle, that's all. That is, I created an assistant with one click in front of you. We saw it on the left side. And it has an API. Here it has an API. We click on it and look. Here is its key. I take it and copy it, paste it. It also has a unique database identifier where it stores its knowledge. I also copy, paste, save. Everything, all excellent, everything worked out. And we can give it a name, because this is our test, let it be a test assistant. We'll call it GRKS. That's all. So, it's very fast. Here, everything is already created. Now, how to use it? We create a workflow. Click. An empty field appears. We add, click on the plus sign. We enter DVR again and immediately see several things. Well, first of all, there are more than needed. Actually, I use only three objects. Three objects. And here are the three. Assistant, Chat Model, and DVR LLM Request. There is duplication here because I am experimenting with libraries and I deployed it under Docker as a separate dependency, so there is duplication. Don't pay attention, these are identical names, identical actions. We created an assistant, right, so I choose the assistant. That's all, we don't need anything else. We can immediately write a prompt here, or a message, what we want to tell it. Well, maybe we'll make a trigger now, like a chat. I'll remove this chat trigger for now. Hello. We see that it's silent because nothing is configured yet. We see that it worked. Hello. Sent. We open our assistant. We see that here is our session ID. Here is our sent message. Now I'll switch to the diagram view. All we need. It's clear that we need to take the word itself and give it to our bot. It also accepts images, so it has digital vision. We can use it. We've already created credentials, so, well, we just created a new test assistant GrKS. We can check it. Now this message is being sent to it, and it should answer us. Here it answered us. It knows who I am because I configured it. This is my personal assistant, created just now. We see the model that answered us, which is inside. Here we see the answer, we see which model answered us, we see our balance, we see the cost for the current operation. In principle, everything is working. Except that we have instantly created a fully working solution, but unlike what everyone else does, we have created a system with an agent, in fact, with two nodes. Moreover, one node is just sending messages. Well, here it answers us. Everything is great. I'll switch to dark mode because I really dislike how the classic standard mode looks. This is how it all works. Everything works perfectly, everything works very well. And what is the difference between the DVR solution and the standard ways of working with artificial intelligence? The difference is that you don't need to, if you were creating it in the classic way, for example, creating an agent, right, what would you need to do for it? You would need to connect a model that would respond, that would be the brain of the agent, add memory, connect tools so that it could find something somewhere, search, work with some databases, or something else. In the case of such a solution, you don't deal with all these things. You just don't need it, because everything is already done according to the MCP principle. Everything is done on the side of a separate server solution. That is, all these procedures, everything happens within here. Here you add your own, I'll show you now for those who haven't encountered it yet, databases, tables, documents, website links, choose models, adjust its creativity and everything else. Everything is done here. And here you just turn it on, and it works, and that's it. That is, the system is only needed here so that you can create a correct, proper automation. Note that in the bottom right corner, there is also a similar head. Why is it needed, this head? This is your personal assistant. It is also built only into my solution, because the classic design does not provide for any inclusions. What does it do? It fully sees our canvas. I'll remove everything unnecessary now. We can verify this by clicking on this eye. Here it shows us, in fact, the current workflow that we have. And this is an assistant that sees the canvas and understands everything, knows everything, and we can ask it questions. It can help us, especially in the early stages when we are still creating something, when we are beginners, like me now, I am a beginner, even though I have rewritten everything, I am still using many nodes for the first time and discovering all this for myself with great pleasure. For example, we ask for a node to send an email. I just send it to it, it thinks, searches itself, and generates it for me. Let's wait and see what node it offers us. So, what's the idea? The idea is that we don't necessarily have to search for information somewhere now, browse YouTube, read documentation. We have an assistant who helps us, even seeing our current situation, even the current workflow. We can describe our tasks, what we are doing, why, and it will solve these tasks. Here it answered. We see that it has already created a ready-made node and offers to insert it. See, right? Beautifully done. So, I just asked, and it already created it, found it. And you can even ask, for example, to configure something or to create a whole workflow or a whole automation, it will do it. Therefore, among the advantages of my solution, my fork that I am making now, is the ability to work with the canvas through artificial intelligence. Like your assistant sees everything that is happening here. It has a history, it understands what we are doing, what we are engaged in, understands what is happening here, sees it, and can correct it, update it, add to it. This is the beauty we get. We have gone through the assistant path, but that's not enough. Now I will show you path number two. Path number two. We will create one more. I'll save it, let it be. Add. We create one more credential. One more connection. This time, we enter DVR again. Only this time we choose not the assistant, but the language model itself. Here the connection is even simpler. In my previous video, I showed how the API of language models works. We just write test. Test DVR connection. And now I'll show you where we get it. We go to DVR, go to all sections. And here is the API to all models. I talked about this in my previous video. If you haven't seen it, you can watch it. All we need to do is take this key. Each user has their own. Copy, copy to clipboard. I will change it after the video, so I can show it to you without worry. We paste it here, save it. We get a notification that everything is fine, everything is green, everything is created. Test connection with DVR API created. These are already general models. We have a list of models. Here they are. Parameters that we can work with, examples of requests, and a list of supported models. These are all advanced models that are currently popular. Plus DVR's free models. And here they are, 2.5, GEMA 327, they have digital vision and are free from DVR. That is, I deployed them on my own servers and give them to you for free for testing, for experiments, for work, for automation. So, we've made a test connection, and now we can create workflows, but we've already created one, so I'll continue here again. Here we have our existing workflow that we made for connecting with the assistant. Now we will create a connection with the model. Again, we enter DVR here and choose not the assistant, but we choose DVR LLM Request. This is a pure model now. So, if we open it now, we will see that here we have created our account just now, selected it. We can write a prompt here if needed, if we use them as a starting point. And we choose models. The listed models that are there in the list, they are all available. We choose any model that is convenient for us, that suits us. In principle, that's all. So, we can now do the following operation again. Write. Hello. We see that everything worked, sent. Now, as usual, we do, we just find the cell where the message was sent, drag it here. Well, in this case, we are sending a request to a pure model. We can write a prompt here, some task, and leave it like, imagine you are a specialist and you consult and explain complex concepts. So. The user asks: and leave, for example, that cell. You can do that. So, everything that is happening now is sent to our GPT5 Nan model, which we chose. And here it has already answered us: "Glad to help, everything works perfectly, everything is excellent." So, our automation is now fully working. Only here the difference is that here we worked with an assistant, meaning inside it there is a database, vector databases, tables, documents, and here we have a pure model. Why do we need a pure model? For more subtle, more specific scenarios. And this same module is also present in another form. I'll show you now. Again, we enter DVR here and find DVR Chat Model. Why is this chat model needed? It's the same. This node, this same block, only capable of connecting to automation blocks. For example, AI. Let's take some agent. So, we decided to build an agent ourselves. For this, we break everything here again, dismantle it, connect it to this agent. And here, this brain is our system. What's good about our DVR system? is that we don't need to create anything, register anywhere, pay money to different providers in different offices, and use different APIs. You can also do this without problems. You can directly add the option to insert an API provider, for example, GPT, and use it completely for free, paying in the provider's office, like OpenAI, without problems. Well, this is how it works. And in this way, in such a combination, in such a scheme, you will be able to create much more sophisticated solutions. You can create, add search systems, knowledge bases. Vector Simple Memory. Simple memory. We can add memory, history, something else. Well, for those who work with this constantly, they know what it is and how to deal with it. The only question is that a node appears that allows you to use any neural networks in a convenient combined way, with control, also with support for your own API keys. So you can use it like this, or like this. Either the DVR solution itself, or the DVR solution as a proxy, as a proxy system with its own API. That's all. So, from here on, you can experiment as you wish, do whatever you want. At the same time, DVR here handles the main task of accessing the language model. So, now we have our own Credentials, our own connections, and our own nodes. Here is the assistant node, the LLM node, and the node that performs the function of a central brain when connecting to AI agent-like blocks. Well, this is how it classically looks in the classic design. In principle, that's all. So, about the assistant, I mentioned it a bit. Now I'll show it again, because at the time I was showing it, the video was covering, uh, covering my video with, it was covering this element. So, we have our own design, our own nodes, our own credentials, connections. And we also have this little thing. This is our assistant, who knows everything, helps with everything. It sees everything that is happening on this canvas, and gives us advice. We can consult with it, ask it various questions. It sees, to see what it sees on this canvas, we click this eye, and it shows the entire workflow that is currently running. So, with every interaction with it, it sees everything that is happening here. It also allows you to create various nodes, insert them. For this, I just asked for an email node, and it immediately typed it out, gave it to me, I clicked, and inserted it here. So, unlike the standard N8N solution, here in the solution that I am offering you, meaning my own fork that I made, it contains your personal assistant, built into your workflows. It's right inside, and it sees everything. This is your real, actual assistant who will consult you, help you, create for you. At the same time, everything it gives you, these code blocks, there is a green button next to the code. And this button, it is clickable. You click it, and the block you asked for appears on your workflow. You can work with it immediately, use it. If something is unclear, write to it, it will answer you. Well, and for dessert, I'll show you one more thing. Very, uh, peculiar, specific. I'll clear all this now. Look, when we see various automation systems, many YouTube videos, almost all of them are related to the same thing, it's getting a bit tiresome, honestly. Almost all automations. If someone talks about MCP, it's always a solution from Anthropic, which is on our computer, and also some Obsidian. If we work with automations in N8N, then almost every automation, every second, almost every first one, is Telegram input, output, posting messages to a group, and so on, and so on. But in reality, uh, I think a little differently. Let me show you as an example, let's recall what artificial intelligence started with, what its development began with. The company OpenAI created a chat. All that they did, they connected a regular primitive funny chat, a field for communication with artificial intelligence. And this caused a sensation, that is, this is exactly what launched artificial intelligence into the masses. This is truly a merit that belongs to OpenAI. And as we see, it's a chat. It's just a chat. And when we say ChatGPT, we say it's a chat with Generative Pretrained Transformers with a mathematical model and AI. And other solutions look exactly the same. This is our classic OpenAI GPT. This is our DeepSeek. I'll open it now. I'll open them. This is our DeepSeek. This is our Anthropic Claude. Here is its field. This is Google Gemini. So, we see the same interface everywhere. Therefore, in one of my videos, I showed that within DVR, I created my own GPT. Here it is, the second in order. And this GPT works exactly the same, only it supports all the neural networks in the world at once. Well, everything that exists, everything is here. Also, I give a lot, again, as I said, for free. Here it is all. You can use it immediately, write, everything works. Good. There are many possibilities and connections here, and we can upload documents, and we can prompt various data and search the internet. I have implemented all this, done everything. But it's interesting to create an automation specifically related to this format, because this format is truly convenient. I start a lot of my work by first discussing the concept with artificial intelligence, choosing a model, and discussing it, and then I create some prototype, some VP solution. So, what am I missing? I am again missing some flexibility in this system, yes, there are models, there is prompting, but what if we want to add this our ChatGPT to the automation system? How does it work? Look, we need to create a Webhook node. Here is our webhook trigger. As everyone knows, if we use a webhook node element, it automatically generates its own internal unique identifier. Using this identifier, we make requests to the server. We can influence this node. The only thing that needs to be chosen here is to change the method from GET to POST, because GET is an external request. I use a hidden method. And here also change the setting, we will use "Respond to Webhook Node". This mode allows us to pass all received data to the next node in an unchanged form. That's, in principle, all, we've created it, right? In principle, all. Also, I haven't shown it yet. Here is this identifier. We take it, go to DVR, select the very first N8N integration. We paste our identifier, which I just took from here. Here it is. I took it, pasted it here. While we are testing, I select test, test hook. Let's remember that our node, this one, allows us to have two types. Test URL and URL for production, for actual work. For now, they differ little. They differ only in webhooktest and webhook without the word test. The identifier remains the same. Therefore, when we paste it here, we simply select the test hook, click save. Everything is saved. And now our node should listen to our our ChatGPT. Well, for example, I'll type it now. It's listening, and I'll write here: "Hello." Okay, something is wrong with it. Ah, it seems to have understood. Here, this Webhook node also needs a response webhook. Without it, it won't work. So, we take the responding webhook node and connect it. This is the minimum setup we need for everything to work. So, let's see, the system has worked. We see that we received this data from our GPT. And now I need to send it back. Let's check, write: "Hello." Sent. We see that it couldn't find a suitable field with text. We see that we received. We received the word "hello" in the request body. Well, and then it doesn't know what to do with it, because here it received it, here it is ready to send it, but where and why,
It does not know. Therefore, we create an intermediate node. We will do it, we will make it based on Dyvar. Based on the DVR node. We enter, we write DVR. We will take. We can take our assistant. Let's take the assistant, select our test assistant, and direct our greeting here. Let's check. Here he answered us. We just need to insert it not here, but here. Everything, we check. It does not answer. Need to turn it on. We will turn it on now. Check. We see, yes, our assistant, whom we chose, whom we created at the beginning, our Grok answered us and sent a reply. Thus, we can configure it in the assistant constructor. And here we can freely communicate with it. Well, it is not necessary to use it. We can use it again, for example, the DVR model. We can use some other nodes that are usually used for this. But since I work with DVR and it seems more promising and simpler to me, I use, for example, the DVR request node. The same thing. I exclude, for example, I disable the assistant for now. I connect to our model and send it here. Everything is the same, only here our assistant worked, and here our model works. Here is our connection. Here we choose any model we need. We add the same message here. We check its operability. It answers. In principle, everything is ready. We can turn it on, check. Hello again. Everything answered. Only this time the model itself answered us, that is, the clean model. Here is our answer. Everything came to us. We see everything, all parameters, everything is there. Thus, we can create our own chat GPT on any convenient model. That is, we can use any other nodes at all. That is, we can add some other options here, some of our parameters. That is, everything is the same, only on other blocks, the presence of these two nodes remains important and necessary. And so that you do not get confused, I have posted them on our page that is DVR. And there, remember the tools at the very bottom of N8N. Everything necessary is already there. Here, download the solution for connecting to GPT from DVR. Custom nodes themselves, this is it. That is, this entire system of ours is already pre-configured there. Thus, you can create quite complex systems with such a plan. You will be able to connect, for example, some agent yourself. Let's assemble it now. I'll remove this one, connect the agent. I'll connect the model to it. [music] Oops, chat model, for example. Connect it here and then direct it to the node. We can do it this way, or we can choose the Dyvar one. These round ones are models that connect to our agents. We can connect the same thing here. We can use several models simultaneously to increase system reliability. Like this. That is, for example, your model and a Dyvar backup, or vice versa, or several Dyvar ones, but with different selected models here. That is, you choose different models, and they back each other up. All we need to do now is to take the message here again. This will be our prompt. Everything is ready. And everything is already working for us. That is, everything continues, everything functions. All three of ours are working. Everything is fully working for us. I will disable these now, and we will check this one. We see that the request has gone, it is being processed. That is, I am now doing it through classic nodes. That is, our triggers are present, hook triggers that work with GPT, processing the classic model from OpenAI. Well, then it will send the request to the node, and we should get an answer. What do we have here? GPT mini, everything worked out. We received the answer. Everything, then it will answer faster, work. As soon as everything is configured, and we are ready to use our workflow, do not forget to enable activation, and here do not forget to remove the test hook so that everything starts working. Thus, we have obtained a powerful automation system. Moreover, we have obtained an excellent automation system, deployed on our own server, locally, allowing us to solve our tasks. And this is just one example of how we can use some external solutions for some of our tasks. That is, in this case, I did not show you an example of how we can create, for example, a multi-agent system for Telegram. I think that if you are interested, we will do it in future videos. We will do many useful things, integrations with various services, but everything in its own time. In the near future, we need to do high-quality automation focused on good, correct data extraction. That is, now I am working on this thing. A database that works much cooler than what currently exists in NN. In NN, I looked at the vector, everything else. Everything, of course, is quite primitive. For the data search and extraction system to work well and qualitatively, it must be not only semantic, it must be combined. The combination consists of three basic pillars: full-text search, graph search, and semantic search. Three things that must be implemented. In addition, it is necessary to provide for the weight of pages, special formulas, page ranks. And it is very necessary to explain to the neural network what needs to be done if we have extracted chunks and pieces of information with contradictory information, that different documents with different dates can have different values. If two documents extracted from the knowledge base contradict each other, but one of them is newer, and the second is older, more outdated, it is implied logically, purely in a human manner, that the newer document will be more relevant. Therefore, if they have a conflict, Russia should not understand that a choice must be made in favor of the newer document. These are the principles, the foundations for the fact that it does not hallucinate, but answers correctly with a large number of documents. So, here I am collecting this thing. Next, we will have, since we already have our own credentials and our own nodes. Here we now have these blocks, yes, DVR, these round systems, our brains. This is what we connected to our agent. But we see that the same agent allows connecting databases. And here, the next stage, when I finish working with the vector, with that powerful one, with the database, we will make a node that will connect to the agent, and you will be able to create very powerful assistants, agents, that will manipulate and work with large data, with complex data, with complexly structured data. Therefore, as you can see, there is a lot of work going on. There is a lot of work, it is difficult. In these 3 weeks, I have gone a huge way, I have gone a very difficult way, I have read documentation, I have written my own nodes, my own connections, nodes, deployed, published all this. Very difficult, little sleep, a lot of work. Therefore, I ask for such simple things from you. Many, I know, want to help the project in some way, but cannot, because they don't know how. So, I ask you to press like. It is not difficult at all to write in the comments. If you don't know what to write in the comments, write it. I like this video. This is enough for our video to take off, for other people to watch us. And we will bring benefit to more people, because, well, what is the meaning of life if we do not do good to each other? We must bring good into the world, the light that we carry. And this is our task. Therefore, I will be glad if you press like, write a comment. Also, many of you write, I remind you that we have a Telegram group besides the YouTube channel, we also have a Telegram group where we discuss technologies and many ideas that I implement here. This is your direct merit. This is what you asked for, and this is what I did for you with my small, not large, but loyal team. So, most people now, who write, are only upset that the system is not very reliable. Servers are standing, we assembled them from equipment. Everything is assembled on the fly. Our capacities are not very large yet, and we try to give everything we have for free. But even power outages often deprive the system of stability, because of heavy rain, wind, power outages, and our servers turn off. Naturally, we have no automation, no anything anymore. Without electricity, all this does not work. Therefore, under this video, besides all the links that I will try to provide for everything I mentioned here, for the API, for the DVR service, for the cabinet so that you can log into NN, into N8N and work in it, I will provide all this. There is also a link to donate. This is not for me, I do not earn on the project. If you put an amount there that is enough to buy a cup of coffee, that is also very good. But I will not drink this coffee. I am trying now with my small team to raise money for a generator. Since we already have deployed equipment, everything is working, loads are holding up better and better. We are constantly working on it. We need a reliable power supply system. We need a diesel generator that can operate around the clock in case of power outages. It should also start automatically without manual startup. Therefore, in the first few days when I opened TN, at least 500 people registered who started actively using the system. Now there are many more. So, even if everyone chipped in $5, we would immediately buy not just a good generator, but a very good, very reliable one, and it would increase the system's efficiency. If you want your automations to work stably, I, you see, I give you all this for free, the cabinet is free, open, n is free, no subscriptions, nothing is needed. Create all your automations and let it all work and bring you joy. But things like power outages, I cannot solve in any way other than by purchasing the necessary equipment. Therefore, we are now collecting for a diesel generator. And if you want to participate, if you consider this matter useful and necessary, I will be glad if you drop a donation via the link under this video. All this money will go towards purchasing the generator. I also wanted to thank the first people who have already participated. This is Halim, this is Ruslan 1C, Mikhail 67, Natalia. Thank you very much, you are great. All your money does not go into anyone's pocket, it is not used for any personal needs. Everything goes towards purchasing equipment. Look, I am now recording a video, and then I am going to add RAM to our servers. We now have the opportunity to add more memory, and this means that all Dyvar solutions will work more stably and reliably. That's all. All the best, all the good. The video turned out to be very long, but I hope you will not scold me too much. I hope you have learned some useful things from this video. Write, join our community on Telegram, subscribe, I will be glad. All the best, all the good. Bye everyone.