AWS wellarchitected documentation change
Summary
Enhanced guidance on stopping conditions for agentic workflows, added implementation details for timeouts and token limits, and updated section titles.
Security assessment
Changes address cost optimization through runtime controls. While timeout mechanisms could indirectly prevent resource exhaustion, the documentation frames them strictly as cost-saving measures without mentioning security vulnerabilities or features. No concrete security evidence exists.
Diff
diff --git a/wellarchitected/latest/generative-ai-lens/gencost05-bp01.md b/wellarchitected/latest/generative-ai-lens/gencost05-bp01.md index 024f739d4..ecccf22f2 100644 --- a//wellarchitected/latest/generative-ai-lens/gencost05-bp01.md +++ b//wellarchitected/latest/generative-ai-lens/gencost05-bp01.md @@ -9 +9 @@ Implementation guidanceResources -Agentic workflows can be long-running, which incurs additional cost to your application. Develop controls to limit agents from running for extended periods of time without stopping. +Agentic workflows can be long-running, which can incur additional cost to your application. Develop controls to limit agents from running for extended periods of time without stopping. @@ -13 +13 @@ Agentic workflows can be long-running, which incurs additional cost to your appl -**Benefits of establishing this best practice:** [Measure overall efficiency](https://docs.aws.amazon.com/wellarchitected/latest/framework/cost-dp.html) \- Agentic workflows can be long-running, which adds additional cost to your workload. By establishing stopping conditions for long-running agentic workflows, you can optimize resources, improve user experience, and optimize workload costs. +**Benefits of establishing this best practice:** [Measure overall efficiency](https://docs.aws.amazon.com/wellarchitected/latest/framework/cost-dp.html) \- Agentic workflows can be long-running, which can add additional cost to your workload. By establishing stopping conditions for long-running agentic workflows, you can optimize resources, improve user experience, and optimize workload costs. @@ -23 +23,3 @@ For generative AI prompt flows where you lack control over the duration of the w - 1. Determine the maximum time needed for an agent to complete its runtime. + 1. Estimate the maximum time needed for an agent to complete its runtime. + + * Include model response times, tool execution times, and network latency in the estimation. @@ -30,0 +33,6 @@ For generative AI prompt flows where you lack control over the duration of the w + 3. Re-architect your workflows to facilitate stopping conditions. + + * Set timeouts on external tools such as Lambda functions or API endpoints, verify that your prompts understand how to handle timeout responses. + + * Set token limits on model responses to simulate timeout functionality by stopping models from printing long-running responses. + @@ -36 +44 @@ For generative AI prompt flows where you lack control over the duration of the w -**Related practices:** +**Related best practices:** @@ -43 +51 @@ For generative AI prompt flows where you lack control over the duration of the w -**Related guides, videos, and documentation:** +**Related documents:**