A travel multi-agent solution calls an airline inventory tool. Intermittent 503s and schema validation failures occur. Which tool error-handling and fallback design should you implement?
Select an answer to reveal the explanation.
Short Explanation
Treat tools like distributed dependencies. Use bounded retries with exponential backoff for transient 503s, validate results against schema, and fall back to alternate tools or graceful uncertainty when failures persist. Don't hang the orchestrator, invent flights, or shut everything down forever.
Full Explanation
Building tool ecosystems includes error-handling, fallback mechanisms, and result validation. The bounded-retry design uses exponential backoff for transient 503s, validates tool results against schema, and falls back to alternate tools or graceful uncertainty. Infinite blocking retries risk cascading latency and resource exhaustion. Fabricating flight availability produces incorrect bookings and destroys trust. Disabling all tools after one error confuses a single fault with total system retirement. Patterns: classify errors as transient versus permanent, circuit-break unhealthy tools, and notify the orchestrator so it can replan. Exam tip: Never “hallucinate tool success” when the tool failed.