AWS Certified Generative AI – Professional (AIP-C01) — Question 34

An ecommerce company is using Amazon Bedrock to build a generative AI (GenAI) application. The application uses AWS Step Functions to orchestrate a multi-agent workflow to produce detailed product descriptions. The workflow consists of three sequential states: a description generator, a technical specifications validator, and a brand voice consistency checker. Each state produces intermediate reasoning traces and outputs that are passed to the next state. The application uses an Amazon S3 bucket for process storage and to store outputs.
During testing, the company discovers that outputs between Step Functions states frequently exceed the 256 KB quota and cause workflow failures.
A GenAI Developer needs to revise the application architecture to efficiently handle the Step Functions 256 KB quota and maintain workflow observability. The revised architecture must preserve the existing multi-agent reasoning and acting (ReAct) pattern.
Which solution will meet these requirements with the LEAST operational overhead?

Answer options

• A. Store intermediate outputs in Amazon DynamoDB. Pass only references between states. Create a Map state that retrieves the complete data from DynamoDB when required for each agent's processing step.
• B. Configure an Amazon Bedrock integration to use the S3 bucket URI in the input parameter for large outputs. Use the ResultPath field and the ResultSelector field to route S3 references between the agent steps while maintaining the sequential validation workflow.
• C. Use AWS Lambda functions to compress outputs to less than 256 KB before each agent state. Configure each agent task to decompress the outputs before processing and to compress results before passing them to the next state.
• D. Configure a separate Step Functions state machine to handle each agent's processing. Use Amazon EventBridge to coordinate the execution flow between state machines. Use S3 references for the outputs as event data.

Correct answer: B

Explanation

Option B is the correct answer because it effectively utilizes the S3 bucket to handle large outputs without exceeding the Step Functions 256 KB limit, allowing for seamless data routing while maintaining observability. Options A and C introduce additional complexity and operational overhead by requiring data retrieval and compression, respectively. Option D complicates the architecture by introducing multiple state machines, which is unnecessary for this scenario.