Act as a senior AI infrastructure engineer. Build The AI Agent Recovery Policy Test Lab, an interactive simulator for testing how autonomous workflows recover when tools fail or produce unexpected results.
Allow users to define a workflow, available tools, expected outputs, retry limits, recovery actions, fallback paths, and success conditions.
The application must:
Simulate timeouts, malformed responses, rate limits, partial results, and unavailable tools.
Compare retry, fallback, compensation, escalation, and graceful-stop strategies.
Detect recovery loops and repeated actions that cannot resolve the original failure.
Track task completion, recovery attempts, execution time, and estimated cost.
Visualize the workflow before and after recovery.
Allow users to inject failures at specific steps and replay the same scenario.
Generate regression tests from failed recovery paths.
Explain why a strategy succeeded or failed under the configured assumptions.
Export scenarios and results as JSON.
Include examples for research, document processing, and customer-support agents. Use deterministic simulations and keep real tool execution disabled by default. Make recovery rules editable and clearly distinguish simulated success from proven production reliability.
0 Comments