[MLOps Course 2] Structuring ML Projects with FastAPI 🚀

All right, now let me quickly walk you through this repo so that you can start familiarizing yourself with how FastAPI works, but most importantly, how you should be structuring all of this in your own ML projects.

Now, you'll have to clone this repo as I already did here.

And the first thing that you will notice is this client directory inside we have call predict.

Now, this is a script that will not live in your actual ML project in your package.

I just added it here for convenience because I want to mimic how it will look like to have a client call one of your endpoints and request a prediction.

So this is all that this is doing.

We also have a models directory.

This is where you will keep your pickle files.

Now, I already called the train model script, which trained the model and generated this pickle file.

So you'll have to do that yourself if you want to see it here.

And then we have our source directory.

Now, inside of source directory, we have my ML project.

That's the name of the project, and that's our package.

And we also have an API directory.

Now, this is standard every time you want to set up endpoints with FastAPI.

This is what the documentation recommends.

This is where we keep basically our endpoints.

All right, so inside of api, we have main.py, this is the entry point.

So if we want to run the server, this is what we are calling, we're calling my main.py, we have routes, which is going to define our specific endpoints.

So one endpoint that I have already defined here is predict, and I left it as a to do for you to define the health checkpoint.

Okay, this is just going to give you some practice on how to define FastAPI endpoints.

All right, so now, something that is also very important, we have a file that is reserved for defining or Pydantic contracts.

So we're using Pydantic so that we can validate the data that comes into our service and the data that we output.

All right?

So we always have a validation on the request and the responses.

Now, you will see in routes here, if we look at the predict endpoint that we are using Prediction response, which is one of the schemas that we have defined here.

Prediction response, right, and that's it.

The readme contains a lot of information on how you should work with this, project.

So, you know, you can find More information in there, but let me quickly show you how to do all of this.

alright, so if I open my terminal, the first thing that I need to do is of course start the server.

So I'm going to do that with uv run FastAPI dev and of course I need to point to the entry point which I mentioned earlier that that was main.py.

So let's do source my ml project API main.py So doing that is just going to start the server.

Now let's open a new terminal because we want to call our call predict, script which is going to mimic how a client will actually call our API.

All right, so all we have to do is uv run client call predict and then so doing this is going to call the predict endpoint and we should receive a response which in this case will be a prediction from our model.

So let's do that.

Great.

So we sent, and by the way, this is in the script, we send this information with the age, income and tenure and then what we got back was a prediction.

All right, Now, just a quick note on why we're using dev here.

This is just mainly for convenience, because as we make changes to our endpoints, as we make changes to our code, we don't want the server to stop running and then having to restart it every time dev comes with auto reload, which makes it very convenient for us, because when we make changes, it just simply restarts the server for us.

Right?

So it's a great way of doing local development.

Now, hopefully you found this useful.

I highly recommend that you do that.

To do it's going to give you great practice on how to define FastAPI endpoints, but also go through the code, try to understand how everything works, how everything is structured, and then revisit the guide because it will help you a lot.

Right.

Good luck.