aws-apigateway-sagemakerendpoint

Language	Package
Python	aws_solutions_constructs.aws_apigateway_sagemakerendpoint
Typescript	@aws-solutions-constructs/aws-apigateway-sagemakerendpoint
Java	software.amazon.awsconstructs.services.apigatewaysagemakerendpoint

Overview

This AWS Solutions Construct implements an Amazon API Gateway connected to an Amazon SageMaker endpoint pattern.

Here is a minimal deployable pattern definition:

Typescript

import { Construct } from 'constructs';
import { Stack, StackProps } from 'aws-cdk-lib';
import { ApiGatewayToSageMakerEndpoint, ApiGatewayToSageMakerEndpointProps } from '@aws-solutions-constructs/aws-apigateway-sagemakerendpoint';

// Below is an example VTL (Velocity Template Language) mapping template for mapping the Api GET request to the Sagemaker POST request
const requestTemplate = `
{
    "instances": [
        # set( $user_id = $input.params("user_id") )
        # set( $items = $input.params("items") )
        # foreach( $item in $items.split(",") )
        # if( $foreach.hasNext ),#end
        {"in0": [$user_id], "in1": [$item]}
            $esc.newline
        # end
    ]
}`

// Replace 'my-endpoint' with your Sagemaker Inference Endpoint
new ApiGatewayToSageMakerEndpoint(this, 'test-apigw-sagemakerendpoint', {
  endpointName: 'my-endpoint',
  resourcePath: '{user_id}',
  requestMappingTemplate: requestTemplate
});

Python

from aws_solutions_constructs.aws_apigateway_sagemakerendpoint import ApiGatewayToSageMakerEndpoint
from aws_cdk import Stack
from constructs import Construct

# Below is an example VTL (Velocity Template Language) mapping template for mapping the Api GET request to the Sagemaker POST request
request_template = """
{
    "instances": [
        # set( $user_id = $input.params("user_id") )
        # set( $items = $input.params("items") )
        # foreach( $item in $items.split(",") )
        # if( $foreach.hasNext ),#end
        {"in0": [$user_id], "in1": [$item]}
            $esc.newline
        # end
    ]
}"""

# Replace 'my-endpoint' with your Sagemaker Inference Endpoint
ApiGatewayToSageMakerEndpoint(self, 'test-apigw-sagemakerendpoint',
                                endpoint_name='my-endpoint',
                                resource_path='{user_id}',
                                request_mapping_template=request_template
                                )

Java

import software.constructs.Construct;

import software.amazon.awscdk.Stack;
import software.amazon.awscdk.StackProps;
import software.amazon.awsconstructs.services.apigatewaysagemakerendpoint.*;

// Create an example VTL (Velocity Template Language) mapping template for mapping the Api GET request to the Sagemaker POST request
final String requestTemplate = "{"
        + "\"instances\": ["
        + "# set( $user_id = $input.params(\"user_id\") )"
        + "# set( $items = $input.params(\"items\") )"
        + "# foreach( $item in $items.split(\",\") )"
        + "# if( $foreach.hasNext ),#end"
        + "{\"in0\": [$user_id], \"in1\": [$item]}"
        + "    $esc.newline"
        + "# end"
        + "]"
        + "}";

// Replace ""my-endpoint"" with your Sagemaker Inference Endpoint
new ApiGatewayToSageMakerEndpoint(this, "ApiGatewayToSageMakerEndpointPattern",
        new ApiGatewayToSageMakerEndpointProps.Builder()
                .endpointName("my-endpoint")
                .resourcePath("{user_id}")
                .requestMappingTemplate(requestTemplate)
                .build());

Pattern Construct Props

Name	Type	Description
apiGatewayProps?	api.RestApiProps	Optional user-provided props to override the default props for the API Gateway.
apiGatewayExecutionRole?	iam.Role	IAM Role used by API Gateway to invoke the SageMaker endpoint. If not specified, a default role is created with access to endpointName.
endpointName	string	Name of the deployed SageMaker inference endpoint.
resourceName?	string	Optional resource name where the GET method will be available.
resourcePath	string	Resource path for the GET method. The variable defined here can be referenced in requestMappingTemplate.
requestMappingTemplate	string	Mapping template to convert GET requests for the default application/json content-type received on the REST API to POST requests expected by the SageMaker endpoint.
additionalRequestTemplates	{ [contentType: string]: string; }	Optional Request Templates for content-types other than application/json. Use the requestMappingTemplate property to set the request template for the application/json content-type.
responseMappingTemplate?	string	Optional mapping template to convert responses received from the SageMaker endpoint.
logGroupProps?	logs.LogGroupProps	User provided props to override the default props for for the CloudWatchLogs LogGroup.

Pattern Properties

Name	Type	Description
apiGateway	api.RestApi	Returns an instance of the API Gateway REST API created by the pattern.
apiGatewayRole	iam.Role	Returns an instance of the iam.Role created by the construct for API Gateway.
apiGatewayCloudWatchRole?	iam.Role	Returns an instance of the iam.Role created by the construct for API Gateway for CloudWatch access.
apiGatewayLogGroup	logs.LogGroup	Returns an instance of the LogGroup created by the construct for API Gateway access logging to CloudWatch.

Sample API Usage

Method	Request Path	Query String	SageMaker Action	Description
GET	/321	items=101,131,162	sagemaker:InvokeEndpoint	Retrieves the predictions for a specific user and items.

Default settings

Out of the box implementation of the Construct without any override will set the following defaults:

Amazon API Gateway

• Deploy an edge-optimized API endpoint

• Enable CloudWatch logging for API Gateway

• Configure least privilege access IAM role for API Gateway

• Set the default authorizationType for all API methods to IAM

• Enable X-Ray Tracing

• Validate request parameters before passing data to SageMaker

Architecture

GitHub

To view the code for this pattern, create/view issues and pull requests, and more:
	@aws-solutions-constructs/aws-apigateway-sagemakerendpoint