Developed and delivered a First Byte Timeout Control feature for LLM streaming requests in the alibaba/higress repository, focusing on enhancing the reliability and responsiveness of AI proxy plugin interactions with large language models. The work involved introducing a new configuration option and implementing logic to set upstream request headers, allowing for more precise timeout management during streaming. Leveraging Go for backend development, the solution utilized API Gateway concepts and Wasm to optimize streaming performance. This targeted improvement addressed latency concerns in AI-driven proxy workflows, supporting better SLA adherence and user experience without introducing new bugs during the development period.