Setting up the Python environment for Windows Setting up the Python environment for Linux

REST API MongoDB

Pandas & Scikit-learn Prepare raw data to be "digestible" by our machine learning model

Keras TensorBoard

Optimize params Save model

Predict-worker OandaV20 API Price Pull (streaming_data collection in MongoDB) Use saved model

Welcome guys! The machine learning for stock market trading. My name is Andrew Buleziuk and I hope you will enjoy this course on LiveEdu.tv. A few words about me: I'm a back-end developer in Munich, Germany and primarily, I do work with PHP.js and, as a side project, Python, primarily, general and general rest framework and the front-end javascript library Vue.js. A few years ago I published the book „Algorithmisches Trading“, Algorithmic Trading for the stock market. I published it on German where I basically explained to investors how to develop their own algorithmic systems for a specific trading platform named MetaTrader4 and it was a really good success. My readers were very happy to have this book. So that being said, I think, that we will learn a lot about algorithmic trading for a stock market using Python. So here's what are we gonna do: First of all, we're gonna initialize a GitHub repository which will be public, so you can all clone it and do or test the code on your own machine. We will create a broker account. We will use Oanda as a service provider. We will then see how we can read data from Oanda API and save the data to our Mongo database. And the data will be something like high/low close and open prices for a specific currency or a specific stock. Then, we will prepare the data for the model and define and extract our target variable. To prepare the data, it has to be in a certain shape that the model can read it. After that, we will define the machine learning model and prepare to visualize the model. If you used Keras before, you will know that you can describe a model, which will give you a better sense of how the layers in the neural network or of the machine learning algorithms are actually structured. We will train and save the model. And when we are going to predict an outcome of a stock market, we will use the live prices of our API and then use the saved model in order to predict the price movement. Then, we were going to visualize the current stock data or currency data with the actual prediction movement of our trained and saved machine learning model. Yeah, for that we are going to use only Python libraries. Yeah, that's basically it. We will have about seven to eight sessions and I hope you are thrilled to be a part of it 00:00:00-00:05:00 No machine learning for stock trading. What do we mean with machine learning? Well, basically, where we are a high level speaking. We are going to use Kara's layers to build a neural network that will take input data from the stock market or from the forex market. Predict a price or a direction of movement direction, or the price for the trading instrument and then either output it as Abe... as a number is a prediction. Or maybe it will act on it meaning it will try to buy or sell something. So, what exactly are we going to do? We're going to create a Github repository. I already did this in the last session. We are going to take a look, how to create a broker account in London. It's a forex broker account, which is very neat. The reason why which you sell one is that it has an API, which is very easy to use, especially with Python. Meaning that first, we can get the data from one and second we can play straights. If we decide to take this project to a neck to the next level and make not just a simple prediction model, but a wedding bot that has a machine learning model as a driving engine for the decisions, for a trade. And then, of course, you can create it on email accounts and it's called practice accounts, meaning you don't have to deposit any real funds. So you can play around before you actually put your real funds into your accounts. The second, we're gonna see you're gonna take a look at how to read the data from the API. We're gonna take a look at how to if today the data that we get from the API to MongoDB. And there are two possible ways to do that, you know. We create a cron job, which will pull the data from the API every... I don't know, one second, one minute, five minutes and the data will get saved to moment DB. It's just much easier to use a know simple database because yep it's an emerging project. Today we have... the way we want to save for data in our collection or in our entry. And tomorrow we... we get and new data inputs that we want to save to a sequel database. It would be a little bit painful to do that. In no sequel database, we just added to our collection. We have to prepare the data, we cannot take the raw data from the stock market, or from the Forex and input it to the model. Well, you have to be somehow relative or like, for example, you can take a roll price from, let's say the German DAX index. it's around twelve thousand right now, I guess. You just cannot put it in there. It would be a model that wouldn't have to be able to do anything with that. I think the prediction would be pretty, pretty bad in this case, right? So, we have to prepare the data somehow we have to clean it. We have to... If we are streaming data from the broker, from the stock market, from the forex market, you have to make sure that we do not have double entries in our MongoDB. Because our trailer trader or our predictor will be flawed to do false predictions. Then we will have to define our machine learning model this system, in my opinion, the most fun part. Because you are playing around with errors well, where you just create add or subtract layers. And just play around until you have a model that does what ..what you wanted to do. Now, there are we are going to define this model manually. Meaning that we will create the layers. There is a trend right now something called how to auto-machine learning out an ml maybe you have read it at Google... now has an API for that. It could try and use that but all-purpose of doing this project is that we get a feeling of how to create these models and using sound a tool that already does that for you. It would be like, um, not... not one we're going to do. The fifth train and save the model this is where we gonna optimize the hyperparameters and save the model in some format into our hard drive. Then we will be predicting the outcome of new incoming stock data. 00:05:00-00:10:00 But this is where we're gonna create another cron job or a streaming service that will get the data from the broker and apply the saved model on it. And then visualize the stock market data prediction or create trader that will place trades into the stock market based on our predictions. This is the first session. Before we go on to that we're gonna take a look at how our system should look like in the end though. But see can you see my pointer. Yeah. So the green box is the outside world for us. Meaning that this is the API from the broker from onda where the data will be coming into our system. There some magic will happen. We will do whatever we need to do with the data. And then we can or we should or we can do a request back to the outside world like... For example, we get data for the German DAX index price movement... isn't that and then our model says, I predict that in the next one hour the price movements will be up. So, we can decide on that information on our prediction to make a request back to the broker and say all right sir by 10 units of German DAX index and close it of an hour or so. So the outside world now the inside world. We have the MongoDB, which is a no sequel database. It is pretty easy to install it on your PC whether you have Linux or Windows. You just go to Monty be. Install. install MongoDB on Windows. Executive bill files and you install it. Very easy. Then we have something called... Let's just ignore what they do. Just focus on the fact that these are workers. Meaning we will not have one Python file that is it has 50 classes and 100 methods and the said job. We will divide our system into small chunks. We will call them workers or so-called services. And I think this will create a service-oriented architecture. And for doing that we will use a Titan library called Nameko. Which is pretty cool, because it is the only one... only packaged all library that I know that gives you the ability to create a service-oriented architecture and Python. I come from a world of PHP. And no GS and no GS have a pretty developed environment for microservices. And Python does not, but retina nameko. It is a funny name. Na me ko. With nameko, we can do that and we will do that. And I think that is we're not only gonna learn what we actually want to learn in this project. Machine learning for stock market trading. We will also learn how to create these microservices, at least at the basic level. Now, there is a lot of features that we will not cover like security, recurring requests, etc., etc. Though, on am an echo to work, these microservices... like qf1 microservice. For example, the microservice that will get the data from the outside world into your system. And then you have one microservice that will read the data, interpret it and then say I want to buy it the S&P; 500. So, these two services they have to communicate somehow with each other and in order for them to community... community with each other we have to install Rabbitmq. This is, basically, something like... yeah, Well, it's not ready but it's just a platform for the microservices to communicate. Don't we go to rabbit and Q and let's write it in Google. Then you go to download their various versions and if you're on Windows you have to download some visual C++ whatever modules from Microsoft. They offer free. RabbitMQ is for a freeze. It just installs the software. I'm pretty sure you're gonna... you'll be able to do that. Though for our project you need RabbitMQ and MongoDB installed on your machine. So, now, let's take a look at these three workers or microservices that we are going to create. 00:10:00-00:15:00 The first of all we have two data dot pi which is a data collector of a worker. This microservice does nothing else than just it's called the API from the or broker and David back to MongoDB. And whatever... in just a raw form it does not apply any cleaning, no pre-processing, nothing. It just gets the data and saves you more DB. And so, it's there. We have no raw data and every time our model is going to retrieve historical data or the current data, we will not have to make requests to the outside world or to the API every time. The saved resources and traffic and so on. But when the data is our MongoDB, we have our model that PI worker which will get historical data for say the last 100 days. Let's say the model trainer wants to train a training model for S&P; 500. And in order to do that we need price data for the last 300 days. So then it speaks to data that PI dated at PI gets the historical data that our microsomia swans. Later that pi gets the data and then returns it to model that pi and model that fight does the prediction the training... it does the training home. So, when a model is trained, we save it. We save it in some formatting on our hard drive and then we have our predictor worker or trade water that will use the safe data. It will load the... it will load the safe model and then it will retrieve the data from MongoDB or it's a dot PI. So, what are we going to color in this session? First of all, the GitHub repository. I already posted the GitHub repository in the chat, yesterday. I'm not going to do that again. I just scroll up a little bit in a chat window. It is the first link. As you can see machine learning for stock trading life education. This repository is only for us. Yeah. So, just go ahead clone that. And then you will be able to pull after every session end and yeah... Try this out on your own machine. That was an easy one. Now, create one account an API token. Let's take a look at the broker. Now, I already have a real account meaning that I have some real fun in it. Um. I'm not sure if I... yeah. But really... Let me show you how it looks when you have already in accounts. So, you go to view your funds. You have a few you have some accounts, sub-accounts over here. I have Albert PI - alga PI one empty for. So you just can create I believe an unlimited account, amount of accounts. And we will we are going to use primary Varaha, where I have deposited 50 euros. Just to play around for us. So if we decide to let our predictor or our project to actually trade in a stock market, we will play around 50 euros. It's not much but I think... We will not create a model that will do or they will be profitable from the start. So the money will probably get lost but that's okay. And if you have to you have an ascendant... it just normal. So it takes a while until you have a model. Until we have a bot or a system that is profitable. So, it's not much but it's enough to play around. So for you... You can go to... You don't have an account already, you can go to all calm. Let me open an incognito window. Change the language to English. You just click on open a demo account. Put in your name whatever. I just create accounts. I'm pretty sure you know how to do that. And when you're done you will get a... You will get an account or if you are similar to mine. 00:15:00-00:20:00 And you will get probably a few thousand heroes to play around, which is so-called the demo or fake money. And with that, you don't play around. So then, when you have an account, you go to... You go back to your account over here. Homepage of your account. On the left side, you can go to manage API access. And... And this... In this field you will see your API, which you have to copy and you will need it later in our project. Right now, I already have generated an API key and I have saved it in a separate file, which I will be importing into our notebooks. Or into our passes and by doing that I prevent you guys seeing the API key and doing some funny stuff with my puns. So, that's why if I click on revoke, I will it will generate a new API key. If you don't have created an API key, it will... l you will see the API key over here. All right, so you know how to create an or under the account and what honor is. You know how to get an API token. Really easy. Now, explore on that API documentary and test API with the postman. You're gonna believe that we're gonna test it with just python notebook. If you go on the left side, I mean in the chat. A jet box you can see there is a link to Github and it's called one API v-22. Open that. Now, this is a Python library. An API wrapper. But we are going to use it. And it's pretty easy to install you just click pip install. If you want to install it as a separate package on your machine, or alternatively if you clone the GitHub repository, we have a small... We have small requirements for that x-file, where we have already for any cameras on the API be 20. And nameko for a microservice framework. So, these are the packages that we will need. Right. Let's pray to play around with... with the one API. I have opened the Jupiter server over here, locally. And we have a notebook. One the API exploring notebook. I can zoom in. Like this. You... Oh, even a second. I'm you... you... you... All right. So, I imported just a basic file, where the API key... this... It's located at and then we import Jason, which we... which we need just to encode the data that we get from the API. Then we import the client which is basically from on API we xx. Import API and with that, we will create our client where we can execute our request like trade, get data, adjust stop loss or something like that. Now, in this example, we want to get adjust some pricing. First of all, we define our access to... Okay. You will have to put in here your access token for your account. Then we have our account ID, which is... If you go to your account... few funds and then with this b20 account number just copy and paste it. 00:20:00-00:25:00 And now, we create our client, which is an API access token and this is very important. This has cost me a lot of headaches because this parameter environment is nowhere shown on the official documentation. So, if you do not say that its life you are playing along but your practice account. And if you do not know that it has to be set to live, your play... You're just assuming that your access token is wrong, that API is not working, that the server is down whatever and it took me a while to get into the API class and look what parameters have to be initialized or what parameters are being initialized in the background under the hood. And then I saw the environment and... yeah. It should be in the official documentation can take a look at what... It's even better to do that. So first of all, let's take a look at what methods we can execute with our client. We have an access token, which we provide. We have a client environment, which we now set. We have a request and request parents. A request method is the most important for us because we will be executing request to the API to get some data. Now, our first request will be the following. On this part prices, we have... We call the pricing part of the Elana v20 endpoints pricing and then we call the bus pricing info and our parameters are the account ID. Which we will want to get the prices. Why is that important? So just imagine that you have a basic account and a premium account. And the basic account has had no access to, let's say S&P; 500. And so if you have a basic account and you want to get the data from a premium trading instrument, you won't get it. So, that's why it's important. And then the parameters and whenever we have said that. It will just as the pricing info will just create a probably weekend or a post request and with this line, we will execute the request. Um, right. So and our parameters, in this case, are instruments that we want to get a price for and it is German text index in here oh. And then a currency pair euro versus the British pound. Let’s execute that and see if this works with an instrument. You're right. It's a string. No. No spaces. That's executed again. And here we go. We have the prices do. It is maybe imported P print, I don't know. If I have this, I have P print which is nice. And just wanna you... you... You... I just want to have the price as DJs in prices printed out a bit. 00:25:00-00:30:00 You but... You did leave this line. Did it again. All right. So now, this looks better. Now, we have here... We have a picnic array of prices. We have our asks which means that if you want to buy an instrument you have an asking price. And this is the price that you would get for the German DAX index. We have here of the basic format for a candle, which means you have here the open price fear a high, low and close price. And it might even mate... make sense although the price range is pretty below. But, yeah, we have a weekend, so there is not much going on. And there are the bit prices close out the instrument. There is non-tradable, meaning that's because we are away. Markets are closed and a weakened variable falls by price units available. I don't think it’s that important to us. And then we have the other trading instrument which is the currency euro versus British pounds. And then only you prices. 00:55:00-01:00:00 All right... which is interesting and I think now it's time that we take a look at the developer or at the documentation of Wanda. And then we will take a look at how they actually... you parameters that we set in our notebook in a way that we will have candles in a specified time frame. As you can see we have different... We have been empty for witches to this trading software. We have some clients for Java and then we will use the rest we 20. We have our endpoints over here. We'll just use the endpoint pricing. Let's go over here. You... All right. So, here we have the request parameters that we can set. If you take a look at this one, it's name is instruments. It's a list of instrument names. And take a look over here. These are the parameters that wem and who all request, which is a list of instruments. And the name for the instruments you have to get from this link over here. What can we say else? Since which is a daytime filter to apply to the response only prices and home conversion for a time later in this filter will be provided. And also filter independently. It explains itself what a task. Account ID will already set include units available. You... All right. Take a look at the instrument. Candles. So, this is actually what we probably want to use. We'll use an hour on our project. So this one I will have to explain a little bit deeper. Though if our project... If our project will get a stream of data and getting every second a new data, our data worker then our model cannot be expecting a candle in a timeframe of five minutes or one hour. So our data is to be consistent both an incoming and saving the data and then in the model that is consuming the data. 00:30:00-00:35:00 So in this regard, we will have to use the candles, hopefully, because if you... if you have candles, like this is the first case... If you have candles and you have a time span, let's say 24 hours, and you will have a fixed amount of candles. I don't know what is it. It's C n 12 candles an hour and then twelve, one... I don't know, say 500 candles. I don't know. 1200 candles and every five minutes. Now, we have a fixed amount of input points that we know that will income into our system, into our machine learning model. It's cannot be just not possible but there is one can no more or one can no less. On the other hand, if we use not candles but streaming prices. Then we do not know whether there will be a thousand... We mean input points or a million because you don't have, like a fixed time range, when the price income, when... For example, the euro US dollar and then in the night there is no volatility. There are no new prices like you can wait for five minutes and there's no new price. Because in the night nobody's trading. But on the other hand at 16 o'clock you can have like the Fed Chairman speaks have... has a meeting and you have high volatility. And you have like and new inputs 10 new prices in a matter of a second. It is said it's really unstable and in order to handle that uncertain data with different time frames or incoming frequency. It's gonna be much harder to create a model to handle death. But that's why we will focus on candles. Now and to do that assume maybe so we will not use the pricing. We will use Kendall's API. Because we will... We can adjust the granola red Nova granularity, sorry, of the candles. And they are we going to set, but we want either a new candle every one minute, every five minutes, whatever. Um. Right. So, now we have done that. We have successfully retrieved some data from the API. In the next session, we will retrieve the actual data... data that we want to have. And install Rabbitmq in MongoDB. I showed you where you can get it. You know the names. And you are most probably capable of installing the software and if not then just write me an email. Or... yeah. Eastern tutorial or something. Install and explore in America functionalities. Now, this is an interesting and fun part. Finally, we get to code something. As I already explained in the beginning. The last session did not happen sadly because of some streaming issues. And we have, um, we have heard and some code and which I didn't revert so far. So that's why I'm going to close everything and we're gonna take a look at what we... What code we do have. First of all, we're gonna focus on the data that PI. It's a data collector worker which will do nothing else than just get the data from the on the API, in a similar way that we did over here. We've read a request we get the data and we save the data to MongoDB. So data API worker. Data dot pie. So, let's open it and take a look at what this actually is. Though, we have... reporting from nameko see and nameko timers and a basically cron job. Later on, we will import the only API. Of course, because we want to make the request. Um. I think it would be wiser to do the documentation example simply because we have done a little bit more work than we should. First of all, let's begin at the very basic level. What is actually in ma-Kun? How does it work? I will create a file called my codify and I input the basic... the basic example. So what does it? You and we important the maker or PC we create a class which is basically a class service and other service have a name. In this case, it's a greeting service and it has a method that an artist and another service came called. 00:35:00-00:40:00 And what does this method do? It just as an input parameter name and then it prints out a string with a hello and the name that it does know. Now we need to Python consoles... you... you... Now, we will start our microservice. Think about nameko is that you started doubting your microservice and then it will just run indefinitely until either your process or your server dies. I saw the restarts. So in order to run our microservice, we call the name echo method and we say run a microservice. And in order to run a microservice, we have to input the name of the file. In our case, it's nameko a pile with just nameko. You... All right. let me work out. Maybe I refactor or rename the file. I'll say, hello. You. You mom... you... Let's rename it one more time. Maybe they're not making some issues with now a seat. Yes. The... the inputs package name from the maker - hello and there's, of course, no such package. Though again... Here we go. You know what nameko has done it started the service, which is named reading servers, like over here. And it is connected to a PQ. Bla Bla Bla. It is our rebellion q server and under the hoods, if we now make a request this microservice to greeting service... This request will travel through rubble MQ, through this Python process. It will be handled the hello function. Do whatever to us and then the output the return will travel to revering to you back to us. But let's do that. We will create manually a request, instead of a request with another microservice. And for that, we go to Namak Rochelle. You... And now we say and for nameko, RPC is the way of speaking to our microservice. Tears or visitors as cron jobs. There is HTTP I like for AP I, saw something and pops up. Are we... are speaking to our microservice to our PC? Then the name of our microservice and the function that we want to call. And let's say life Edu and then we get the results back... hello, life education and if you take a look at our code. This is the function, hello and it returns hello life education. Just to get a better sense out works. Let's make it our own little function, which we will call some and then we have first, second... second and we are going to return the sum of these two parameters. First plus second. David and now we do the same nameko RPC greeting service. 00:40:00-00:45:00 But this time we say it's a ton and first is life... and is ten. And we get an error, which is great. [Music]. But see. Yeah. Right. This is... This... This mistake happens to me every time. So now, we edit the function, but we did not restart the service. We have to stop it or they start to service again. And now we can do this or someone as you can see we get some bag which is 15. In our case and yeah... Now you have a greeting service you can have an unlimited amount of functions that do whatever you need them to do. And yeah... This is where this is really, really cool. This way you can deconstruct your monolith in a service-oriented way. And then just do your architecture more maintainable. All right. So, this is the data API. Data dot pi part of our project, meaning that it is a data collector and a data collector should do what? First of all, we should have a method that will get... Open low close data from the API. And what you wanna... I would say we don't need anything. Return and we want to return under data. Now, what is it on the data? The data is for now just an empty array. And our to do is... how to do this get data from Wanda if your eye. And the second part is you do Elsa is safe on the API to Malmo dB. You... So every time this method gets cold, it automatically makes a request to under that pie. Gets the data to save it to MongoDB and this is basically all it does. Now it's an RPC. We do not really need that. We are going to do a cron job out of that. Because we just want to work this microservice in the background and independently, meaning that it does not have to receive a request in order to do something. Now and to do that we go to nameko documentation. We have a user guide over here and you have these built-in extension topics over here. Which I mentioned before you have RPC events, which pops up. HTTP trigger and timer. We will use a timer. You... Here we will use a timer because we want to have a five-minute candle every five minutes. And for this purpose, a timer is more than enough to do this. The timer is a simple entry pointed fires one pair configurable number of seconds. So for that, we need to import a timer. And we have to change our decorator from our PC to the timer. Now we can delete our PC because we don't need that anymore. We can change the greeting service to eita service and it is not a greeting service. It is a data service. All right. Here we get all HT a timer that will run not every second but every five minutes. And we have to put it in in seconds which is 60... 60 seconds times five and there we have. It's basically, it's a simple plan job but with a... yeah, what if what am echo... but in a makeup package which means that you can have here one method with a decorator for a cron job and then another method for HTTP, another matter method for RCP. 00:45:00-00:50:00 And this is actually what we need. So our next method will get historical. Get historical data. And before we continue I will actually switch to our 8 at PI because this is something we already started wit you. And let's see. We have a timer. We have a PC. We have our gear on HD data. The module is getting data from honor API is something we will do in the next chapter. And then our next method that has to work is getting historical. Open high/low close candles. Um. Why is this important? Let's hope for our structure. So this worker or this training... training file we'll need like hundreds of days of data. And in order to get them, you have to either first get the data to get historical data by making a request and getting a big chunk of data back. Or you get the data by letting this beta collective run for one hundred, two hundred, three hundred days and after it's... the data collector has worked for three hundred days with no interruptions. You will have a data collection of three hundred days in your MongoDB database. So but you see where the problem lies, we would have to let this worker-run for three annotations order to get the data. What we do not want to do that we want to have the data here and now. So and for that, we have now two possibilities you know we get the data once for 300 days we get at once and save it into our MongoDB. And then our... and then our model gets the data every time from MongoDB and not from the API directly. And [Music] this would be a much wiser thing to do because you get the data for 300 days once. And then every time you train your model, you will not have to make these API calls. Now and this is something it requires a little bit more code and a little bit... a little bit... a little bit more understanding. Because you have to go through the data. You have to clean the data and to avoid doubles and so forth. And it's really... it is more important when you are living when your trader bot is on a server, where you have to pay for your resource usage. Where you have to pay for your network traffic and so on. Because we're doing this locally, we will do it the easier way, which is just getting the data every time from the API. And later on, we could if we continue to work on this project beyond this... this tutorial on life education, it would be wise to change that direction. Now, in order for our machine learning model to get the data, we need this RPC model. This means that the model that PI is a microservices and nameko microservice will make an RPC call to the data collector. The data collector will make an API rest call to oh under the PI will get the data and then rebel me. It will pass on the data to our model training worker and that water will then do its stuff with Kara's. Now, it's a little bit rainy but long logical chain over here. Don't... Let's do this everything step-by-step. That historical data... What do we need to do? What do we need to know? It is the instrument we want to have the data from life. Instrument. Is it is that the S&P; 500, is that the euro versus US dollar and from eight and to date. This is all we want to half and then historical well open high low close is... It's just an empty... It is just an empty list... out of course. 00:50:00-00:55:00 Afterward, you say to get Oracle price data from one API. Now, let me input a basic doc string over here. Women from today, blah blah blah. Cool. Very honor from API like something we may want to do is just leave a to do here. Save historical data. Since we are doing this anyway maybe we will change the direction of the stream still. Right. I would suggest that we simply run our microservice in order to take to see what I wear it works at all. Maybe we have some typos. Namiko runs data. The module name, hello. You... There was a typo of... All right. So now, we have to start so versus data service at blah blah blah over. Now, we can um it's a... First of all, let's take a look at how our cron job works. Now we have set it to 5 minutes. We do not want to wait 5 minutes in order for this method to get one iteration. Let's set it to one second. Save it. Restart our microservice. Run it again. And as you can see every... Every second the microservice as it executes the function gets open high, low, close data. And for now, it's... It's out the same data which... which is in our case just a simple... simple list. A simple variable and as you can see to do we will replace this data that gets spit out with the... API data that we get over here. And then when the data arrives in our variable we will... our next to do is save it to mon bébé. Oh, yeah. You know what? Let's leave it that way. Let's see our numerical handles to... you... two methods of one microservice at once. So we have a crown job, which is being executed every second. And now let's change too and let's make one request to the RPC method. Get historical or open high, low, close and in order for us to see that it was actually executed. We'll print. Are we gonna Prince? Must be... How about you just copy/paste the data we got over here? But we got over here print... Cool. You... and historical or hoc equals 2 FET. You... It's an object. All right. We have to restart our run. Now, create a manual request or here we go again nameko RPC and why is RPC. It is because our method gets historical of night look laws works on the RPC channel. Now, what is the name of our microservice? It is as you can see the class data collector and the name data service. So we have to put in your data underscore service and some... We do not have that method anymore. Our method is... method is called yet Oracle open high, low, close. And then we have three parameters that we have to submit an instrument. Which is would say all row versus u.s. dollars the currency pair. And then we have from the date and put some random eight. Later on, we will have, of course, to submit a precise date. Date equals 20 in the 18 and funds and yeah... Let's execute that's one... get Oracle method not found. All right. But this curl open high/low close but if I copy and paste that. Got an unexpected keyword from the data. Horses from date. Now, here we go. So it spits it out nothing. Why? Because before we return to the historical data of arrival we set it to zero. But we printed a historical ohlc which means that this one Danika like the microservice. It's fitted out data in its own, of course, can simply delete this historic of this deletion of the historical data of arrival. And if we save that... Let us start a microservice... and I'll imagine that when I call this RPC request on the left-bottom, imagine that this request is not manually being handled by me. Imagine it's a request that our... second... that our model trainer is doing. So the model comes in and says oh alright it's already been two weeks and I have to retrain my model for the stock market trading. And then it says but now I need historical data for the last three years. And then we have this data collector microservice over here. So the model microservice passed the request and says, hey data service. How are you doing? And I have the historical data for a year US dollar from this date to this date. And then the model trainer gets this big fat chunk of data. Of course, this is just an example of data. When we will actually fill the data with... with a well it's API request. You will have like it where we have very, very... where we pick JSON object with thousands of thousands of data points. And yeah... In this way, the model will get the data and do whatever is needed to do it. All right, guys. Um. I am very thankful that you joined me. I'm also very happy that this time the stream worked out! I hope. Let me take it. I can check it real quick. Yeah. Not loading right now. We reload the whole page. Um, yeah. It's what... It works. Great. So, thank you for watching guys! For those of you who didn't... Who weren't from the beginning with me the last session did not work out because of some streaming issue. And this session was yesterday. So and in order to fix that I'm going to do two sessions today. This was the first one we have accomplished very much actually so far. We have you know how to create an own account. We have our GitHub repository, which is in the chat window at the top. We have created a... You know how to create an undercount. Where to get an API key. You know how to use the API key in order to create a client or here to get your account ID. And now we know how to get... how to create a request to get the prices from the Oregon foreign exchange or stock market program. I assume that you will have installed RabbitMQ and MongoDB by the next session. Or maybe the next week. 01:00:00-01:04:15 And we know how to use name echoes with all the functionalities that we need. And in our case, these are actually only our PC or a microservices to speak with themselves. And a timer and just a cron job that will pull the data from the API. In the next session, let's take a quick look at what are we gonna do today. On explore v20 Python wrapper. This is what we're gonna focus mainly on an instrument to get the right data. Then create nameko services for data updated. We are we actually already did that. Create a Cron job for fetching. We already did that. Save data to memory be collection. This is the important part and creates a new maker or ECM of our PC root for getting fresh OC data. This is something that we did accept is a fresher Oracle's or HD beta. As you can see we already did some basic stuff from the second session. And the second session we will actually fill it not just with some basic examples that do not do any good. We will fill it with actual beta with good actual data getting the important stuff. All right. So, thank you for watching! I hope you enjoyed it. I would, as always, I would really really be happy if you would be more interactive in the chat. If you have any questions. And I will see you... I don't know. Let's make a break of half an hour, one hour. What... What do you want guys? Well, it's really up to you. What I will do now is I'm going to push the changes that we did over here. You... you... you... So, I commit the changes as a first session. It push. You... you... No. One more time. You... you... All right. Here we go. You... Now, we have a third commit with the example of hello at PI. If you go to another PI, you will see my API key. And this is something you may want to know. If your API key gets exposed somehow or you lose it, I don't know. Or if like 40 people do you see your women key or your API key in a stream. Then you have to revoke it and for that, you just go to your API access page and revoke it. So this is what I'm gonna do because it hopes now has my API key. Alright, guys. Thank you for watching! And I will see you in an hour. Take a break. It would be cool if you would install MongoDB and RabbitMQ the stuff that we need for... for our next session. Thank you for watching! And I will see you in one hour.