AWS wellarchitected documentation change
Summary
Expanded error handling guidance with detailed failure classification, recovery mechanisms, monitoring, and continuous improvement processes
Security assessment
Changes focus on reliability and fault tolerance rather than security. Added error recovery patterns (retries, circuit breakers) improve resilience but don't address security vulnerabilities or document security features.
Diff
diff --git a/wellarchitected/latest/generative-ai-lens/genrel03-bp01.md b/wellarchitected/latest/generative-ai-lens/genrel03-bp01.md index 200d6d256..565e636ed 100644 --- a//wellarchitected/latest/generative-ai-lens/genrel03-bp01.md +++ b//wellarchitected/latest/generative-ai-lens/genrel03-bp01.md @@ -25 +25 @@ Agentic workflows commonly make calls to external systems. Develop agents with e -When developing multi-step prompt flows, consider using Amazon Bedrock Flows to orchestrate multi-step prompts. Bedrock Flows enables graceful failure and recovery for long prompt chains, which allows your applications to take appropriate action on failure. Bedrock Flows has nodes for controlling flow logic, which include iterator nodes and condition nodes. Customers may consider using these nodes to implement graceful recovery instead of developing a custom abstraction layer. +When developing multistep prompt flows or prompt chains, consider using Amazon Bedrock Flows to orchestrate multistep prompts. Bedrock Flows enables graceful failure and recovery for long prompt chains, which allows your applications to take appropriate action on failure. Bedrock Flows has nodes for controlling flow logic, which include iterator nodes and condition nodes. Customers may consider using these nodes to implement graceful recovery instead of developing a custom abstraction layer. @@ -29 +29 @@ When developing multi-step prompt flows, consider using Amazon Bedrock Flows to - 1. Leverage Amazon Bedrock Flows to capture model output and dynamically determine the next step in the flow. + 1. Establish error classification system: @@ -31 +31 @@ When developing multi-step prompt flows, consider using Amazon Bedrock Flows to - 2. Implement response capture and conditional logic at the application layer to account for and recover from unexpected model behavior. + * Categorize common failure types @@ -33 +33 @@ When developing multi-step prompt flows, consider using Amazon Bedrock Flows to - * Classify systems responses based on actionable or inactionable outputs. + * Define severity levels @@ -35 +35,33 @@ When developing multi-step prompt flows, consider using Amazon Bedrock Flows to - * Leverage contextual grounding guardrails to capture error and recover from error scenarios. + * Create response templates for each error category + + * Set up automated detection mechanisms + + 2. Implement recovery mechanisms: + + * Design retries strategies with exponential backoff + + * Create fallback prompt templates + + * Develop circuit breaker implementations + + * Set up automated recovery workflows + + 3. Configure monitoring and alerting: + + * Track recovery success rates + + * Monitor remediation effectiveness + + * Set up alerts for repeated failures + + * Implement performance tracking + + 4. Create continuous improvement process: + + * Analyze failure patterns + + * Update remediation strategies + + * Refine prompt templates + + * Optimize recovery workflows @@ -42 +74 @@ When developing multi-step prompt flows, consider using Amazon Bedrock Flows to -**Related practices:** +**Related best practices:** @@ -49 +81 @@ When developing multi-step prompt flows, consider using Amazon Bedrock Flows to -**Related guides, videos, and documentation:** +**Related documents:**