feat(aws-apigateway-sagemakerendpoint): New aws-apigateway-sagemakerendpoint pattern implementation

[dscpinheiro]

Issue #, if available: #75

Description of changes: This PR introduces a new pattern, where inference from a deployed model in SageMaker can be done via API Gateway.

As mentioned on the issue, each ML use case is unique and the construct requires certain properties (such as resource path and request mapping template) to be provided. Also, for the documentation (README.md) and integration tests, I've used this blog post as a reference.

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[aws-solutions-constructs-team]

AWS CodeBuild CI Report

• CodeBuild project: codebuildgithubautobuildPro-rp7jMDIK1wBN
• Commit ID: 235fa5e
• Result: SUCCEEDED
• Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

[hnishar]

Good first cut, few comments to look into

[hnishar]

Can it support both endpoint name and ARN ?

[dscpinheiro]

The path for the integration has to be endpoints/<sagemaker-endpoint-name>/invocations, so if ARN was used we'd have to parse it. Is there an example of this in Constructs right now?

[hnishar]

Check aws-apigateway-iot pattern

[dscpinheiro]

So I looked, but it's a subdomain there (e.g. ab123cdefghij4l-ats.iot.ap-south-1.amazonaws.com), which is different than the ARN. I could do something like this:

// Parse endpoint name
try {
    const components = Arn.parse(props.sageMakerEndpoint);
    
    // This line is required since resourceName is string | undefined
    if (!components.resourceName) {
        throw new Error('Provided ARN does not contain resource name');
    }

    this.sageMakerEndpointName = components.resourceName;
} catch {
    this.sageMakerEndpointName = props.sageMakerEndpoint;
}

Would that be enough? I'm not sure if the complexity added is worth it.

[hnishar]

Probably not worth the complexity, let's make sure the comments clearly say it expects the endpoint Name and not the ARN of the endpoint

[dscpinheiro]

Both the interface comment and README file mention name, not ARN (Name of the deployed SageMaker inference endpoint). Is there anywhere I should change it?

[aws-solutions-constructs-team]

AWS CodeBuild CI Report

• CodeBuild project: codebuildgithubautobuildPro-rp7jMDIK1wBN
• Commit ID: 7c79371
• Result: SUCCEEDED
• Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

[aws-solutions-constructs-team]

AWS CodeBuild CI Report

• CodeBuild project: codebuildgithubautobuildPro-rp7jMDIK1wBN
• Commit ID: bc65b1c
• Result: SUCCEEDED
• Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

[aws-solutions-constructs-team]

AWS CodeBuild CI Report

• CodeBuild project: codebuildgithubautobuildPro-rp7jMDIK1wBN
• Commit ID: 34ee408
• Result: SUCCEEDED
• Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

[hnishar]

Thanks for the contribution, the pattern will be published in the next release.