📄 中文摘要
速率限制是导致 AI API 调用在生产环境中失败的主要原因。这并不是一个错误，而是服务提供商为了保护其基础设施而采取的措施。应用程序在运行一段时间后，用户可能会在某些情况下遇到错误，尤其是在高峰时段。错误信息通常为 HTTP 429 — 请求过多，表明应用程序正在受到速率限制。如果处理不当，问题可能会加剧。了解速率限制的机制、如何解读相关信号以及采取适当措施来避免影响应用程序的正常运行至关重要。
API 速率限制与节流：实际发生了什么以及如何修复
📄 English Summary
API Rate Limits & Throttling: What's Actually Happening and How to Fix It
Rate limiting is the primary reason AI API calls fail in production environments. It is not a bug; rather, it is a measure taken by providers to protect their infrastructure. After running smoothly for weeks, an application may start to show errors for some users, particularly during peak times. The error message typically indicates HTTP 429 — Too Many Requests, signaling that the application is being rate limited. If not handled correctly, the situation can worsen. Understanding the mechanisms of rate limiting, how to interpret related signals, and taking appropriate actions to prevent disruptions in application functionality is crucial.