Building a Machine Learning API: Integrating AWS Lambda with API Gateway

Key Takeaways

Welcome back everybody. So what is our next goal? We want to expose deployed ML model features and get a prediction response. Our model is running in AWS Lambda container. So for a quick recap, I have taken the example request which we had passed while doing a test using Lambda function. So this was our request body and this was the example response which we are expecting. So in short we are going to expose our ML model via APIs. I have also mentioned the steps we are going to perform and then I will be taking you all towards performing these steps one by one. So we will be checking whether our lambda is working fine or not. Then we are going to open the API gateway console which is basically one of the very good service in AWS in order to expose the APIs. If you are not aware about API gateway you can get an understanding API gateway act as a front door for all the APIs request coming from the users and then to redirect it to the backend services. In short, it is acting as a middleware everybody or facade layer. You're going to create a new rest API. You will be creating a post method and integrate with lambda. Then you're going to deploy the API. You're going to test from the API gateway console and then you will be testing using curl or a Python script. Okay. So I will be using a Python notebook in order to do a round of testing. So let's go step by step. Our first step is to ensure that our lambda working fine. So I will be navigating to management console and the first landing place would be lambda. Okay, you all would be already aware that how to get into the lambda console. If not search for lambda in the search bar and you will be navigating to lambda dashboard. Okay, you can always click on this last modified in order to check what was the latest function you had created. In my case, I have this iris ML container. Let me open this and I know that I have basically deployed this particular function using container images. I will be going to test and let me click on test button. We are going to ensure that we are getting the expected response which we had taken as a sample response to and yes we have the exactly the same response we were expecting and mentioned as a sample response. Next thing is go to API gateway everybody. So in the search bar type API and it will start giving you suggestions. So I have got a suggestion for API gateway. I will be opening API gateway. So currently I'm in Mumbai region. You can make a choice of a region as per your requirement. Okay, let me show you something. If you are new to API gateway, you can click on this and read about it little bit. Amazon API gateway is a fully managed service that makes it easy for developers to create, publish, maintain, monitor and secure APIs. Either you can click on create an API from here. Okay? Or you can go to this particular API section. From here also you can click on create an API. Let's see the region in which no API is existing. Means I want to show you a layout difference. Currently in Mumbai region I am already having so many APIs created. Let me take any other region in which I don't have any single API created. And let's see if there is any layout difference. Okay. So here I don't have any API created. If I will be going to create an API, I'm going to get a similar kind of layout. Okay. So there is no layout difference even if you are having an API existing in your region or not. I'm going back everybody because I'm going to use Mumbai as my primary region for all my exercises. And even you have to ensure like if you are basically using Mumbai for creating lambda and other resources you had used in the past demo then API you have to create in the Mumbai region only. Okay. So let me click on create API. We have different types of APIs. I'm going to choose REST API because REST API provides more flexibility [laughter] and you can also gain complete control over the request and response along with the API management capabilities. Okay, it works with Lambda HTTP AWS services. I want to integrate this API with Lambda function. That is my need. So, click on build and you have to choose new API. If you're already having an existing API, you can select the second radio button and so okay. But as I am going to create a fresh new API. So let me provide the API name. Okay. So let's say I am giving the API name as Iris ML API. If you want you can give any name of your choice. In descriptions you can mention APIs for Iris ML model. Something like this. Okay, the API endpoint type I will be keeping regional. We have different type of API endpoint. Okay, for the time being I will be talking about regional APIs are deployed in the current AWS region. Just remember that much. Okay, if you have chosen Mumbai, it will be deployed in Mumbai region. Let's click on create API. So our API is created. Now we are going to click on create resource. What is resource? Everybody understand resource as the entry point. Whenever you are going to access the APIs, resources will be entry point or you can also call it as the logical grouping. It is useful logical grouping everybody. So let's say I'm creating a resource. I'm giving a resource name as predict and this would be my resource path. Okay, secret resource you can see a resource got created. You will be seeing a resource with slash predict under the slash. Now what is the next thing everybody? Once the resource got created, we have to create a method. Okay. So we will click on create method. We have different types of method. I am going to create a post method. And then the integration type would be lambda. I want to integrate the API with a lambda function. If you have any other requirement, let's say you want to integrate with existing HTTP endpoint, you want to generate a response based on API gateway mappings and transformations. You can choose other radio button. But in our case, we know that our ML model is deployed on Lambda. So we are going to integrate this API with lambda function. Okay. So method you have chosen post and now we are talking about lambda integration. So in the search type Iris ML container. You have selected this. You can adjust other settings if you want. But for the time being we are not changing any logic. We are not providing or maintaining any request header. I'm keeping it very simple. I just want to show you how can you expose it via APIs. Click on create method. So you have created a post method too.