Integrate Adaptive Reasoning Budgets

Modify your LangGraph agents' use of Claude Opus 4.1 to incorporate adaptive reasoning budgets. For low-severity alerts, agents should use a smaller token budget and faster inference. For high-severity, multi-region incidents, they should dynamically switch to a larger budget and potentially more 'extended thinking' steps (simulated by multi-turn interactions or specific prompt structures) to ensure thorough analysis, balancing cost and performance.