A multi-agent HR FAQ system repeatedly answers near-identical policy questions, driving high token cost. You must design caching without serving stale policy after updates. Which caching strategy is most appropriate?
Select an answer to reveal the explanation.
Short Explanation
The combined strategy is right because it uses prompt caching, semantic caching, and response caching with TTLs and invalidation when policy sources change. That's how you cut repeated FAQ cost without serving stale HR policy. You should key caches by policy version or content hash and watch hit rate against freshness incidents.
Full Explanation
The combined strategy is correct because multi-agent solutions should use prompt caching, semantic caching, and response caching, balanced with invalidation for changing knowledge. Semantic thresholds catch paraphrases, while TTLs and invalidation when policy sources change prevent stale answers. Caching nothing ignores cost and latency optimization and wastes money. Caching model weights on devices misses the point and does not address repeated FAQ completions. Forever keys from the first three characters are collision-prone toy keys causing wrong answers. Use policy version or content hash keys, enforce TTLs for regulated HR content, and monitor hit rate versus freshness incidents.