From 42dd5fabc629e0cef9066cddb85aa9cf18f92579 Mon Sep 17 00:00:00 2001 From: xiaoxia Date: Sun, 4 Oct 2026 18:34:11 +0800 Subject: [PATCH] =?UTF-8?q?fix(#2180):=20=E5=85=A8=E5=B1=80=20max=5Fretrie?= =?UTF-8?q?s=202=E2=86=921=EF=BC=8C=E9=81=BF=E5=85=8D=E9=87=8D=E8=AF=95?= =?UTF-8?q?=E5=8F=A0=E5=8A=A0=E5=88=B0351s?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit 原配置max_retries=2（即3次尝试），lite失败→3次超时→pro降级→3次超时， VLM最坏46.6+76.7=123s，文案三级重试最坏351s。 timeout调大到45/60s后，方舟API正常情况下1次调用就能在24-38s内返回， max_retries=1（即2次尝试）足够防偶发网络抖动。 最坏情况VLM总耗时：45+45+60+60=210s，比351s改善40%。 --- packages/config/base.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/packages/config/base.py b/packages/config/base.py index 8a081e0d3..8b4d3ecc4 100755 --- a/packages/config/base.py +++ b/packages/config/base.py @@ -96,7 +96,7 @@ class SharedSettings(BaseSettings): ) doubao_base_url: str = "https://ark.cn-beijing.volces.com/api/v3" doubao_timeout: int = 45 # #2180: 方舟LLM高峰期响应6-8s，原30s太紧提到45s - doubao_max_retries: int = 2 + doubao_max_retries: int = 1 # #2180: timeout调大后一次调用就够，1次重试防偶发抖动；避免6次重试叠加到351s doubao_vision_model: str = ( "doubao-seed-2-1-pro-260915" # 高精度视觉（Seed 2.1 Pro 原生多模态；原 vision-pro-250328 已下线） )