自主智能体陷阱：我的 AI 如何每天消耗 300 多次 LLM 调用来检查自身状态 💸🤖

Building autonomous AI agents with frameworks like OpenClaw, LangChain, or AutoGPT often leads to the fear of unexpected API bills. Recently, my AI assistant, running on Google's Gemini models, began throwing 429 RESOURCE_EXHAUSTED errors. Initially, I thought it was merely checking system health, but it turned out to be running a full-blown medical drama every 55 minutes, racking up significant token usage without my knowledge. This situation serves as a reminder for developers to carefully manage resource usage when granting LLMs the ability to autonomously call tools, in order to avoid unexpected cost surges.