Transcription
Hello everyone. For those who don't know yet, I am Misha. Natural Intelligence. The DVR project. DVR is an ecosystem in which we develop neural networks, various solutions. Very many people are already using it. And I am one of those who are rooting for this project, who would like it to develop, to grow. Today I offer to your attention what many of you have long asked for. We will talk about API. API or API in the vernacular. Such a protocol by which we can connect to neural networks, to models, to assistants through various automation systems, such as NM, simply with code, there to add some solutions to your applications and so on. That is, here already, uh, more material for those who have already tried to work with this in one way or another, know what API is, what it is used for, and who needs it. Most often, such people have already tried themselves in various with various services, for example, with such a service as Open Router. Uh, there are also analogues that live on this. The principle is that we, of course, can get, get API access along with keys, with accounts, creating payment pages, adding cards. Most often, this is a big problem, especially in countries where there are sanctions on neural networks, Russia, Belarus. So, the popularity of universal APIs, universal APIs is due to the fact that there is no need to make multiple registrations, control accounts, try to pay for something abroad, because the main providers of artificial intelligence are still located abroad on another continent, in the USA or in China. The largest providers of artificial intelligence of large language models. And all this is great, but there are no such convenient, good solutions, apparently, because you continued and continue to write until now, when will the API appear on Dovar. And today this moment has come. Now I will show where to look for what. First, I will remind you that we have an assistant constructor. This assistant constructor allows you to quickly create any assistants for any taste and color, embed them anywhere in Telegram, on a website as a widget, send them by link to someone, and so on. Well, and use them simply for yourself. It also has pre-configured assistants already pre-set. Well, for example, I have created something like DVR GRKS, a parody of Grok from Elon. So, one click and it's already created. I can use it. So, in this solution, the API was already present. If we click opposite its cell, we will see the assistant's API. We click on it and see clear detailed instructions. Well, here I will disable it, delete it, change it at any time, so it's not a big deal that the key and the database are displayed here. The principle of operation is very simple. Each assistant gets its own key. And it also gets its own database identifier, into which it stores its knowledge, because our assistants are functional, they are not just language models, they store information, they access databases, vectors, RAG, documents, and so on. Especially here we have pre-configured, so these functions are closed here. Well, for example, even I created it for myself. So, here we have the ability to connect documents, spreadsheets, databases, enable, add links to websites from where it will get information, specifically on web pages, select models, configure context, well, all that sort of thing. Many already know, I had a video about this. Therefore, I just wanted to mention that we already have such a solution. Each assistant has its own API API. And the advantage of DVR assistants again is that they work like MCP, yes, this fashionable protocol now, model context protocol, which, uh, everyone, that is, the MCP server takes on all the problems, all the complexities, all the nuances, and we simply access it, solve all tasks. The same happens with this API. Assistants work, uh, logically similar. Their principle is similar. Well, today I present to you a broader, more powerful solution. Again, for the purpose of bypassing blocking here, sanctions blocking on models. I wrote my proxy, wrote it myself, hosted it on an intermediate server. It was quite expensive, relatively expensive. $160 per year to maintain just this platform, which routes traffic through itself so that the IP address is replaced, so that there are no blocks from providers of large language models. In the new solution, which I will now show, where it is located immediately, all sections of the site. And at the bottom, API to all LLMs. If we click, again, all these are keys I will demonstrate. After this video, I will change all of them. It changes instantly. Just refresh the key. We click, and our key is reissued. There are no problems with this. Everything, I have described it here. How to access. Described. Here is a single point of access, API is a single endpoint, very simple. Here it is. Each user will find their key, if you are registered in DVAR, log in, you will see your key, this red one. You can copy it and use it immediately. How to use it? Here I have placed examples of the request structure, examples of how everything works, what is sent where, what data we send, what we receive. Everything, everything is very simple. Currently, I have added support, support for the main models, the most advanced, the most top ones. Most of them are multimodal, with digital vision. Only Psikh, unfortunately, still has models that it returns are not multimodal, which is quite inconvenient. In principle, that's all. There is not much else to explain here. If you have worked with API, if you know what it is, then you can start using it right away. Well, once again, I will explain, once again, I will explain how everything works. DVR is a platform, organized on the principle of digital communism. This is what we jokingly call the principle of operation, when all payments are made collectively. That is, you have an internal account that you can top up. You can also earn it. There is such an opportunity here. By filling out questionnaires, recommending the service to other users. The system tracks their transitions, registrations, and in case they top up their accounts, a percentage of 10% goes to your account. So, such a principle is laid down. Therefore, everyone has their own account. From this account, funds are spent when working with our services. DVR has many services already. I have to remove some because there are too many of them. Each of them performs its task using artificial intelligence in a specific direction. So, you don't have to worry, you topped up your account. We do everything else on our side. We top up real accounts. All this with tax payments, of course, with all the necessary nuances. That is, we solve all these issues, plus we pay for proxying, and you use it, and you don't need to look for anything anywhere, set up anything anywhere, insert keys, or anything else. Although, on the other hand, if you want to use your own keys for free, so that this settlement account, these internal points are not spent, you can add your own here. By clicking here, you can add your own keys to the cells, real ones, from the providers, and requests will be free for you, because they will go past the service. Only proxying will be used, nothing else. That's how it's set up. Also, in AI, at the moment, we are giving DVAR models for free. We already have our own servers. We now have two servers on which neural networks are hosted, various neural networks, language, diffusion. Transcription is already, in fact, gradually, little by little, we are deploying more and more capacities. So, what we have now, our own, we try to give for free. Currently, all this is free. That is, through the same AI or these assistants, you can use DVAR models. They are usually lower, local, free. Yes, they are not the most powerful. We don't have the ability to deploy large models ourselves yet. On our capacities, we now have, here is hidden a Chinese one, 32 billion parameters with digital vision. And we have another second server. On it, there is now DEMA 27 billion parameters as the main one. There are also smaller ones. They perform more specialized tasks. Well, these are our main models that we give for free. So that you can connect, test, experiment. In principle, that's all. Uh, probably a bit dry. I've explained a little, it won't be interesting to everyone, because it's exactly for those who are really creating something. And you can create something if you haven't encountered this, but you want to create something of your own, you can use existing platforms. This will be discussed in my next video. So, I will finish this video now and proceed to the next one. And the next video will be more interesting. It will be related to automation. I will show what it will be about. It will be about the NN or N8N platform, in which DVAR APIs are already used in the form of ready-made nodes that I have configured, that connect to the process, that work, and on which you can independently, without knowing code, create multi-agent systems, create various automations. What I will try to show in my next video. Well, for those who are not used to this style, this is a fork that I wrote. Here is the classic view, classic style design. This is how it looks more standard for those who are already working on this platform, automating, trying. Therefore, I invite all enthusiasts to experiment. The one that I showed you today, told you about, is ready for use. You can safely start, try, get positive results, rejoice. Well, and at the end, I wanted to ask, so that more people see this video, it is useful, I think it will be useful for many. Therefore, I ask you to like it, it's not difficult. Write comments, if you don't know what to write, write. I like this video. That's enough for the YouTube algorithms to push the video higher in the search results and show it to other thinking, enthusiastic, interested people. Therefore, I will be glad. for your reactions of any kind. I read your comments. I don't always manage to respond promptly, but I read all your comments. There is not a single comment that I have not read. That's all. All the best, all the best to you. See you soon. Bye. เฮ