📄 中文摘要
在生产环境中，LLM 应用的成本回归问题往往是无声的，尽管一切正常，但费用却显著增加。大多数服务提供商的仪表板只能显示总支出，而生产团队更需要了解费用激增的具体原因，包括是哪个端点、哪个提示部署或哪个客户造成的。这篇文章提供了一个实用的操作手册，帮助团队及早发现提示部署导致的成本回归问题。
提示部署可能悄然增加你的 OpenAI 账单——如何捕捉这一现象
📄 English Summary
Prompt deploys can silently spike your OpenAI bill — here’s how to catch it
In production environments, cost regressions in LLM applications can occur silently, where everything appears to function normally, yet expenses rise significantly. Most provider dashboards only display total expenditures, while production teams require insights into the specific causes of cost spikes, such as which endpoint, which prompt deployment, or which customer is responsible. This article offers a practical playbook to help teams catch prompt deployment cost regressions early.