AWS CodeBuild CI Report

• CodeBuild project: AutoBuildv2Project1C6BFA3F-wQm2hXv2jqQv
• Commit ID: 52a175c
• Result: SUCCEEDED
• Build Logs (available for 30 days)

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

I was thinking if it's possible we can have a InferenceProfile class with a fromFoundationModel() method that would return a class with inference profile arn?

For example,

claud 3.5 sonnet v2

• model ID - us.anthropic.claude-3-5-sonnet-20241022-v2:0
• inference profile arn - arn:aws:bedrock::<ACCOUNT_ID>:inference-profile/us.anthropic.claude-3-5-sonnet-20241022-v2:0

I was considering a class like this

export interface IInferenceProfile {
  profileArn: string;
}

export class InferenceProfile implements IInferenceProfile {
  readonly profileArn: string;

  private static getCountryCode(scope: Construct): string {
    const stack = Stack.of(scope);
    if (!Token.isUnresolved(stack.region)) {
      return stack.region.startsWith('us') ? 'us' : 'eu';
    }
    return 'us';
  }

  static fromModel(scope: Construct, model: bedrock.FoundationModelIdentifier): IInferenceProfile {
    // Construct the inference profile ARN dynamically
    const profileArn = Stack.of(scope).formatArn({
      service: 'bedrock',
      resource: 'inference-profile',
      resourceName: `${this.getCountryCode(scope)}.${model.modelId}`,
    });
    return new InferenceProfile(profileArn);
  }

  static fromInferenceProfileArn(arn: string): IInferenceProfile {
    return new InferenceProfile(arn);
  }
  
  constructor(profileArn: string) {
    this.profileArn = profileArn;
  }
}

So we can bulid up the inference profile for Claude 3.5 Sonnet v2 like this

// Create an inference profile for the model
const inferenceProfile = InferenceProfile.fromModel(this, bedrock.FoundationModelIdentifier.ANTHROPIC_CLAUDE_3_5_SONNET_20241022_V2_0);

Things we need to check:

• Not all foundationModels have their inference profiles, we probably need to maintain a support list in the InferenceProfile class?
• other considerations?

Thanks for the suggestions @pahud, I think that'd be a good way to proceed and I'll take a closer look in the coming weeks

We are adding support for it also in awslabs/generative-ai-cdk-constructs#800 for the L2 Bedrock construct, which we expect to merge back to the core cdk

This PR has been in the MERGE CONFLICTS state for 3 weeks, and looks abandoned. To keep this PR from being closed, please continue work on it. If not, it will automatically be closed in a week.