Orchestrate and Validate the Full Workflow

Step Functions State Machine

Create a state machine that coordinates the processing stages. A simple bootcamp version can use these states:

1. MarkAnalyzing
2. AnalyzeTranscript
3. SaveResult
4. MarkCompleted
5. HandleFailure

If you implement audio input, add:

1. StartTranscription
2. WaitForTranscription
3. CheckTranscriptionStatus

Minimal Transcript Path

For a reliable final demo, the transcript path is enough:

Uploaded transcript -> Step Functions -> Lambda -> Bedrock -> S3 report -> DynamoDB status

Test Case 1: Successful Transcript Analysis

Input:

{
  "jobId": "job-test-001",
  "bucket": "<bucket-name>",
  "transcriptKey": "uploads/job-test-001/sample_conversation.txt"
}

Expected result:

• Step Functions execution succeeds.
• DynamoDB status becomes COMPLETED.
• Report exists under reports/job-test-001/report.json.
• CloudWatch logs show no unhandled exception.

Test Case 2: Missing Transcript

Input references a transcript key that does not exist.

Expected result:

• Workflow moves to failure handling.
• DynamoDB status becomes FAILED.
• errorMessage explains that the S3 object is missing.
• CloudWatch logs contain the failure details.

Test Case 3: Bedrock Access Error

Temporarily use a model ID that is not enabled, or remove Bedrock permission from the test role.

Expected result:

• Lambda catches the access error.
• Workflow fails in a controlled way.
• Error is visible in CloudWatch and DynamoDB.

Final Validation Checklist

• Upload path works.
• Job status can be retrieved.
• Bedrock output follows the report structure.
• Logs are available.
• Failure cases are visible.
• Cleanup steps are documented.