aws-apigateway-sagemakerendpoint module

All classes are under active development and subject to non-backward compatible changes or removal in any future version. These are not subject to the Semantic Versioning model. This means that while you may use them, you may need to update your source code when upgrading to a newer version of this package.

| Reference Documentation:| https://docs.aws.amazon.com/solutions/latest/constructs/| |:-------------|:-------------|

| 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

Note: Each SageMaker endpoint is unique, and the response from the API will depend on the deployed model. The example given below assumes the sample from this blog post. For a reference on how that'd be implemented, please refer to integ.apigateway-sagemakerendpoint-overwrite.ts.

| 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

© Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.