Title: Vertex AI Gemini streaming silently drops per-request timeout, falling back to litellm.request_timeout (6000s)

当你在 LiteLLM 里对 Vertex AI / Google AI Studio Gemini 模型使用 stream=True (含 timeout= )时,per-request 超时会被静默丢弃,流卡住时不会按你设置的秒数失败,而是回退到 litellm.request_timeout
![[Bug]: Anthropic streaming with response_format and tools reports finish_reason "stop" when a user tool is called](https://www.chat-gpts.plus/wp-content/uploads/2026/10/45082-866d3e0c-768x403.jpg)







