AWS published a reference workflow for rebooking passengers after a flight cancellation. In it, an agent proposes options, while AWS Step Functions, a service for running task sequences, checks proposals with deterministic code.
Agent tasks do not write reservations or issue compensation.
Deterministic tasks do that after validation succeeds. The validation confirms flight availability, compliance with fare rules, and route validity.
The example sets the parallel-processing limit to 1,000. If the limit is omitted or set to zero, distributed-processing mode runs up to 10,000 child executions in parallel by default.
Claim check:
- AWS published a reference workflow for rebooking passengers after a flight cancellation. (confirmed by the publication itself: evidence; «The pattern is demonstrated here for airline rebooking»)
- In this reference workflow, an agent proposes options and deterministic code validates them. (confirmed by the publication itself: evidence; «The principle is that agents propose, and deterministic code validates.»)
- Agent tasks do not write reservations or issue compensation; deterministic tasks do that after validation. (confirmed by the publication itself: evidence; «no agent Task state writes to the reservation system or issues a payment. Only deterministic Task states do that, and only after a deterministic validation step has passed.»)
- The validation confirms flight availability, compliance with fare rules, and route validity. (confirmed by the publication itself: evidence; «An AWS Lambda Task confirms each proposed flight is bookable by checking live availability, fare rules, and route validity»)
- In the example, the parallel-processing limit is set to 1,000, while distributed-processing mode runs up to 10,000 child executions in parallel by default. (confirmed by the publication itself: evidence; «MaxConcurrency is set to 1000 to bound fan-out and protect downstream booking and inventory systems. If you omit it or set it to 0, you get the default behavior, which runs up to 10,000 parallel child executions.»)
Publications:
Primary sources:
- https://aws.amazon.com/blogs/compute/validating-multi-agent-decisions-with-step-functions-and-bedrock-agentcore
- https://aws.amazon.com/blogs/machine-learning/abnormal-ai-amazon-bedrock-agentcore-for-agentic-email-security-at-scale
score 68.8 out of 100 · kind: guide