diff --git a/_conf_schema.json b/_conf_schema.json
index 7e9fe9d..3584d9d 100644
--- a/_conf_schema.json
+++ b/_conf_schema.json
@@ -81,6 +81,24 @@
         "default": 35,
         "hint": "⚠️ 警告：增加此值会导致大量Token消耗！默认值35轮（有一定消耗）。设置过高可能导致LLM费用激增"
     },
+    "llm_timeout": {
+        "type": "int",
+        "description": "LLM 请求超时时间（秒）",
+        "default": 30,
+        "hint": "LLM 单次请求的超时时间，单位秒。可根据模型响应速度适当调整，增大可减少超时失败，但会占用更多等待时间。"
+    },
+    "llm_retries": {
+        "type": "int",
+        "description": "LLM 请求重试次数",
+        "default": 2,
+        "hint": "当请求超时或失败时自动重试的次数，建议设置为1~3之间。"
+    },
+    "llm_backoff": {
+        "type": "int",
+        "description": "LLM 请求重试退避基值（秒）",
+        "default": 2,
+        "hint": "重试之间的基准等待时间（秒），实际等待时间为基值乘以尝试次数。"
+    },
     "pdf_output_dir": {
         "type": "string",
         "description": "PDF输出目录",
diff --git a/src/analysis/llm_analyzer.py b/src/analysis/llm_analyzer.py
index bb91437..f7c2689 100644
--- a/src/analysis/llm_analyzer.py
+++ b/src/analysis/llm_analyzer.py
@@ -6,6 +6,7 @@ LLM分析器模块
 import json
 import re
 from datetime import datetime
+import asyncio
 from typing import List, Dict, Tuple
 from astrbot.api import logger
 from ...src.models.data_models import SummaryTopic, UserTitle, GoldenQuote, TokenUsage
@@ -18,6 +19,33 @@ class LLMAnalyzer:
         self.context = context
         self.config_manager = config_manager
 
+    async def _call_provider_with_retry(self, provider, prompt: str, max_tokens: int, temperature: float):
+        """调用LLM提供者，带超时、重试与退避。"""
+        
+        timeout = self.config_manager.get_llm_timeout()
+        retries = self.config_manager.get_llm_retries()
+        backoff = self.config_manager.get_llm_backoff()
+
+        last_exc = None
+        for attempt in range(1, retries + 1):
+            try:
+                coro = provider.text_chat(prompt=prompt, max_tokens=max_tokens, temperature=temperature)
+                return await asyncio.wait_for(coro, timeout=timeout)
+            except asyncio.TimeoutError as e:
+                last_exc = e
+                logger.warning(f"LLM请求超时: 第{attempt}次, timeout={timeout}s")
+            except Exception as e:
+                last_exc = e
+                logger.warning(f"LLM请求失败: 第{attempt}次, 错误: {e}")
+
+            # 若非最后一次，等待退避后重试
+            if attempt < retries:
+                await asyncio.sleep(backoff * attempt)
+
+        # 最终仍失败，记录错误并返回 None 由调用方处理降级，避免抛出异常
+        logger.error(f"LLM请求全部重试失败: {last_exc}")
+        return None
+
     async def analyze_topics(self, messages: List[Dict]) -> Tuple[List[SummaryTopic], TokenUsage]:
         """使用LLM分析话题"""
         try:
@@ -108,11 +136,10 @@ class LLMAnalyzer:
                 logger.warning("未配置LLM提供商，跳过话题分析")
                 return [], TokenUsage()
 
-            response = await provider.text_chat(
-                prompt=prompt,
-                max_tokens=10000,
-                temperature=0.6
-            )
+            response = await self._call_provider_with_retry(provider, prompt, max_tokens=10000, temperature=0.6)
+            if response is None:
+                logger.error("话题分析调用LLM失败: provider返回None（重试失败）")
+                return [], TokenUsage()
 
             # 提取token使用统计
             token_usage = TokenUsage()
@@ -336,11 +363,10 @@ class LLMAnalyzer:
                 logger.warning("未配置LLM提供商，跳过用户称号分析")
                 return [], TokenUsage()
 
-            response = await provider.text_chat(
-                prompt=prompt,
-                max_tokens=1500,
-                temperature=0.5
-            )
+            response = await self._call_provider_with_retry(provider, prompt, max_tokens=1500, temperature=0.5)
+            if response is None:
+                logger.error("用户称号分析调用LLM失败: provider返回None（重试失败）")
+                return [], TokenUsage()
 
             # 提取token使用统计
             token_usage = TokenUsage()
@@ -440,11 +466,10 @@ class LLMAnalyzer:
                 logger.warning("未配置LLM提供商，跳过金句分析")
                 return [], TokenUsage()
 
-            response = await provider.text_chat(
-                prompt=prompt,
-                max_tokens=1500,
-                temperature=0.7
-            )
+            response = await self._call_provider_with_retry(provider, prompt, max_tokens=1500, temperature=0.7)
+            if response is None:
+                logger.error("金句分析调用LLM失败: provider返回None（重试失败）")
+                return [], TokenUsage()
 
             # 提取token使用统计
             token_usage = TokenUsage()
diff --git a/src/core/config.py b/src/core/config.py
index 7bb908b..0243ac3 100644
--- a/src/core/config.py
+++ b/src/core/config.py
@@ -71,6 +71,18 @@ class ConfigManager:
         """获取最大查询轮数"""
         return self.config.get("max_query_rounds", 35)
 
+    def get_llm_timeout(self) -> int:
+        """获取LLM请求超时时间（秒）"""
+        return self.config.get("llm_timeout", 30)
+
+    def get_llm_retries(self) -> int:
+        """获取LLM请求重试次数"""
+        return self.config.get("llm_retries", 2)
+
+    def get_llm_backoff(self) -> int:
+        """获取LLM请求重试退避基值（秒），实际退避会乘以尝试次数"""
+        return self.config.get("llm_backoff", 2)
+
     def get_pdf_output_dir(self) -> str:
         """获取PDF输出目录"""
         return self.config.get("pdf_output_dir", "data/plugins/astrbot-qq-group-daily-analysis/reports")