From 21aec70e929f32a781bcf3817a5d545336a42812 Mon Sep 17 00:00:00 2001
From: tt-P607 <68868379+tt-P607@users.noreply.github.com>
Date: Mon, 25 Aug 2025 01:55:18 +0800
Subject: [PATCH] =?UTF-8?q?=E6=88=91=E6=88=96=E8=AE=B8=E4=BF=AE=E5=A5=BD?=
 =?UTF-8?q?=E4=BA=86=E5=8F=8D=E6=88=AA=E6=96=AD=EF=BC=9F=E4=BD=86=E6=88=91?=
 =?UTF-8?q?=E6=97=A0=E6=B3=95=E9=AA=8C=E8=AF=81=EF=BC=8C=E7=94=A8gemini=20?=
 =?UTF-8?q?2.5=20pro=E7=94=9F=E6=88=90=E6=97=A5=E7=A8=8B=E6=88=AA=E6=96=AD?=
 =?UTF-8?q?=E5=A4=AA=E4=B8=A5=E9=87=8D=E4=BA=86=EF=BC=8C=E8=BF=9E=E5=AE=8C?=
 =?UTF-8?q?=E6=95=B4=E7=9A=84=E4=BA=8C=E5=8D=81=E5=9B=9B=E5=B0=8F=E6=97=B6?=
 =?UTF-8?q?=E9=83=BD=E6=97=A0=E6=B3=95=E8=BE=93=E5=87=BA=EF=BC=8Cgemini=20?=
 =?UTF-8?q?2.5=20Flash=E6=88=AA=E6=96=AD=E5=A5=BD=E5=BE=88=E5=A4=9A?=
 =?UTF-8?q?=EF=BC=8C=E6=88=91=E8=AF=95=E4=BA=865=E6=AC=A1=E7=94=9F?=
 =?UTF-8?q?=E6=88=90=E7=9A=84=E9=83=BD=E6=98=AF=E5=AE=8C=E6=95=B4=E7=9A=84?=
 =?UTF-8?q?=EF=BC=8C=E5=8F=AF=E8=83=BD=E4=BF=AE=E5=A5=BD=E4=BA=86=EF=BC=9F?=
 =?UTF-8?q?=E4=BD=86=E6=9C=89=E5=BE=85=E9=AA=8C=E8=AF=81=EF=BC=8C=E5=BB=BA?=
 =?UTF-8?q?=E8=AE=AE=E5=9C=A8=E5=BE=85=E5=8A=9E=E4=B8=8A=E6=89=93=E4=B8=AA?=
 =?UTF-8?q?=E9=97=AE=E5=8F=B7?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 src/llm_models/utils_model.py | 22 ++++++++++++++--------
 1 file changed, 14 insertions(+), 8 deletions(-)

diff --git a/src/llm_models/utils_model.py b/src/llm_models/utils_model.py
index 8432302c5..d4ed32fb8 100644
--- a/src/llm_models/utils_model.py
+++ b/src/llm_models/utils_model.py
@@ -340,16 +340,22 @@ class LLMRequest:
                         is_truncated = True
                         logger.warning("未检测到 [done] 标记，判定为截断")
                 
-                if (is_empty_reply or is_truncated) and empty_retry_count < max_empty_retry:
-                    empty_retry_count += 1
-                    reason = "空回复" if is_empty_reply else "截断"
-                    logger.warning(f"检测到{reason}，正在进行第 {empty_retry_count}/{max_empty_retry} 次重新生成")
+                if is_empty_reply or is_truncated:
+                    if empty_retry_count < max_empty_retry:
+                        empty_retry_count += 1
+                        reason = "空回复" if is_empty_reply else "截断"
+                        logger.warning(f"检测到{reason}，正在进行第 {empty_retry_count}/{max_empty_retry} 次重新生成")
 
-                    if empty_retry_interval > 0:
-                        await asyncio.sleep(empty_retry_interval)
+                        if empty_retry_interval > 0:
+                            await asyncio.sleep(empty_retry_interval)
 
-                    model_info, api_provider, client = self._select_model()
-                    continue
+                        model_info, api_provider, client = self._select_model()
+                        continue
+                    else:
+                        # 已达到最大重试次数，但仍然是空回复或截断
+                        reason = "空回复" if is_empty_reply else "截断"
+                        # 抛出异常，由外层重试逻辑或最终的异常处理器捕获
+                        raise RuntimeError(f"经过 {max_empty_retry + 1} 次尝试后仍然是{reason}的回复")
 
                 # 记录使用情况
                 if usage := response.usage: