A travel multi-agent app calls a flaky third-party fare API. Intermittent 429 and 503 responses currently cause the planner to invent prices. You must harden tool use. What should you implement?
Select an answer to reveal the explanation.
Short Explanation
You need to harden tool use by designing error-handling and fallbacks: retry transient 429 and 503 errors with backoff, use alternate tools or cached fares when appropriate, return explicit failure states to the orchestrator, and validate results so models can't silently invent prices.
Full Explanation
Production multi-agent systems must assume tools fail. Design tool error-handling and fallbacks: implement retries with backoff for 429 and 503 responses, circuit breakers for prolonged outages, secondary data sources or degraded modes such as alternate tools or cached fares when appropriate, and clear error objects the orchestrator can branch on (retry, ask user, cancel). Result validation and quality checks prevent the model from treating missing data as license to hallucinate. Treating any tool failure as success and letting the model fabricate numeric results creates business and trust risk. Crashing the entire multi-agent process on the first tool timeout with no retry or user messaging worsens availability. Removing error status codes from the tool response schema removes signal required for control flow. Log tool failures with correlation identifiers and include them in reliability monitors.