Transcription
Hi everyone! Alex is here, let's check MCP. So what is MCP? Is something like USB for all of your AI needs, and let's go to the code to check out. It's better, and I think that it'll be much easier for you to understand in this way.
So we have our favorite folder with CLI, with everything that we should have. So firstly, we're downloading our venv, so our virtual environment where we can store all of our information. Let's go further, information, I mean libraries that we'll be using. Then we're creating that. Let's go and create it together, creating venv, and voila, we have. Let's activate that to work on this. For activation, we are using, for Linux and Mac, different commands, this one, and for Windows, this one. It's just because of the architecture, it's quite different in this way. We're running that in our venv, in our bin folder, activate file. That's how it runs. And then you can see that in our terminal, we have before our folder where we are working, we see activation of virtual environment.
Let's go further, how to install libraries. Here we have something interesting. First, let's upgrade our pip because I have been doing that for a while, and then install in quotes, MCP-CLI. That's important. That's what we'll be working on with, and let's go. Yeah. Here we should wait that a little, and will be able to go next.
Then, important thing we should discover, Python PEs. This one, we can store it, for example, here for a further return. Then let's run our MCP. Okay, let's do that a little bit later. After one of the main things, after we'll discover what is that and how does work. So you know, what about, what is that? We'll be talking a little bit later. Let's discover how we can work with that, so we have something similar to what we have been doing before. It's our dictionary with our size, bedrooms, prices, data frame, our standard scaler. Why we should do that? Because we should just say that these numbers should be from zero to one, so it'll be easier for our machine learning to understand them. And also, machine learning mostly works with them, so that's why. And also, we should feed in our transformer instantly, our size and bedrooms. And what is the most interesting? We're instantly creating our model with linear regression, and we're feeding our, each scale, our data that we have scaled and our prices due, what we are expecting to have. And here we are not testing them, we're not taking all the things because here we have interesting thing about our MCP. So MCP is the one tool that we can use to run all of our projects. And we see the same presets information. So let's just look at that, and I will give you a try to play with that and hope that you'll figure out how you can use that in your life, in your project.
So MCP, firstly, we are starting the server, MCP, fast MCP, and here we have two things like optional resources exposed to data as CSV. So we have our current data set end. What we're doing, we are putting everything from that to CSV because it understands that in CSV format. But thanks to Pandas, we can instantly and very easily create CSV, and the next thing, the next tool, we'll have to predict price, what we'll be able, we'll be able to put. Size, square feet, float, and bedrooms like in, and then transform our data and not instantly putting that to our model. And then we are taking prediction and writing our prediction, and then just running our MCP. That's all, that's quite interesting that we don't have our usual form of our sequentials, etc. We can have that, but here is the way to see how does MCP work, and the interesting thing that you can have ever in one place and understand how does it work is pretty important here.
So we can go to our CSV, run MCP, I mean, and to run MCP, copying that, open terminal and just running. And file wasn't found, of course, because we have been doing that inside of our, inside of our SRC folder. So we're going to our SRC, and we're running it, and now we are having Python in cache, our proxy session talking, and voila, here what we have, and let's do the team the next week. Firstly, to the comment, we should put where our project is, like which Python we will be using, and we'll be running just mainly that by work hitting, connect. Waiting a little, maybe a little more. We have received post messages through the session. And hope that it'll connect us soon. And here we are. I don't know if it worked because I just hit that two times, like this run. But in any case, that's what we have. We have history of our nationalization, what we can do. It's, let me show one of the simplest things. And, thing that not everyone knows. For example, we have this one tool or one of the things that is one tool for all of your AI things because Claude, ChatGPT, all them are supporting MCP, and you can totally work with that, and it'll be easier for you. For example, let's go to the tools, list tools, predict price, take number, input, scale them with the same scaler, predicts with the aligner model, and returns JSON-friendly dictionary. What we have here, important parameters, let's go with thousand and four. Run to, it's much easier, faster, and more understandable. How does it work? Do you see? It's all in one place. Like you can instantly understand, for example, the success tool in the comfort way to write your data that you wanna, you can also write it with a code, but here it just makes that easier and faster. We can go to our resources, list resources, current dataset, and discover, which in general dataset we have. And discover all of the data that we have there, especially if we have a quite big one of them. We don't have any other things here, so this is our resources, resources, templates we don't have. So you can go here. Oh, here. We can close this one. You can see like we are using resources and tools. That's the only two things that we have written. Resources, tools. If we could have more, we'll have more things. We have, we don't have list of the problems. We don't have ping like, oh no, we have ping. We just can ping the server and discover if we have something, for example, history, we have ping method, ping and, and response. Nothing. And we can write for our. Receiving, for example, what we can have to discover that everything works correct, samplings, destination's, roots, authorization that you can use to work with your specific, specific models for different companies. In any case, it's, it's not a thing. The thing is that it is one tool for AI engineering to understand one thing, and you can use it whenever you want. So with that, you can faster and easier work with your tooling like during all of the daytime. So that's just the beginning. What you can have, like the basis, easiest way what, what we can show that just, just to run. But it's definitely the thing to check it out because not everyone, sadly, try that for, I haven't been trying that for a long time, and I really like that. I know it's that, so that's the thing. You know what? Just try to discover new things, and especially in our case, MCP.
Thanks everyone for watching this video. I was happy to see here, and wishing you all the best, and see you the next time, and write in the comments what do you want to see more. I wish you all the best, and see you. Bye.