update

backend/app/agent/smart_agent.py

@@ -3,6 +3,7 @@
 """
 import re
 import json
+import asyncio
 from typing import Dict, Any, Optional, List
 from app.config import get_settings
 from app.agent.context import ContextManager
@@ -37,6 +38,14 @@ class SmartStockAgent:
         else:
             logger.warning("Smart Agent初始化完成（规则模式，建议配置LLM）")
 
+    async def _call_llm_async(self, messages: List[Dict[str, str]], temperature: float = 0.7, max_tokens: int = 2000) -> Optional[str]:
+        """异步调用LLM，避免阻塞事件循环"""
+        loop = asyncio.get_event_loop()
+        return await loop.run_in_executor(
+            None,
+            lambda: llm_service.chat(messages, temperature, max_tokens)
+        )
+
     def _register_skills(self):
         """注册所有技能"""
         skill_manager.register(MarketDataSkill())
@@ -411,7 +420,7 @@ DIF和DEA的位置关系，MACD柱状图变化，判断动能强弱和买卖信
 """
 
         try:
-            analysis = llm_service.chat(
+            analysis = await self._call_llm_async(
                 messages=[{"role": "user", "content": prompt}],
                 temperature=0.7,
                 max_tokens=3000  # 增加到3000，因为分析更详细了
@@ -564,7 +573,7 @@ DIF和DEA的位置关系，MACD柱状图变化，判断动能强弱和买卖信
 """
 
         try:
-            analysis = llm_service.chat(
+            analysis = await self._call_llm_async(
                 messages=[{"role": "user", "content": prompt}],
                 temperature=0.7,
                 max_tokens=1500
@@ -751,7 +760,7 @@ MA60：{f"{ma['ma60']:.2f}" if ma['ma60'] else '计算中'}
 """
 
         try:
-            analysis = llm_service.chat(
+            analysis = await self._call_llm_async(
                 messages=[{"role": "user", "content": prompt}],
                 temperature=0.7,
                 max_tokens=2500
@@ -1032,7 +1041,7 @@ MA60：{f"{ma['ma60']:.2f}" if ma['ma60'] else '计算中'}
 只返回JSON，不要有任何其他内容。"""
 
         try:
-            result = llm_service.chat(
+            result = await self._call_llm_async(
                 messages=[{"role": "user", "content": prompt}],
                 temperature=0.3,
                 max_tokens=300
@@ -1094,50 +1103,18 @@ MA60：{f"{ma['ma60']:.2f}" if ma['ma60'] else '计算中'}
             # 处理美股
             return await self._handle_us_stock(stock_keyword, message)
 
-        # 处理A股和指数
+        # 处理A股和指数 - 使用LLM进行智能匹配
-        # 指数映射表
+        stock_info = await self._match_stock_with_llm(stock_keyword)
-        index_mapping = {
-            "上证指数": "000001.SH",
-            "上证": "000001.SH",
-            "大盘": "000001.SH",
-            "沪指": "000001.SH",
-            "深证成指": "399001.SZ",
-            "深证": "399001.SZ",
-            "深指": "399001.SZ",
-            "创业板指": "399006.SZ",
-            "创业板": "399006.SZ",
-            "科创50": "000688.SH",
-            "沪深300": "000300.SH",
-            "中证500": "000905.SH",
-            "A股": "000001.SH"  # 默认用上证指数代表A股
-        }
 
-        # 检查是否是指数查询
+        if not stock_info:
-        stock_code = None
+            return {
-        stock_name = None
+                "message": f"抱歉，未找到股票或指数\"{stock_keyword}\"。请确认名称或代码是否正确。",
-        is_index = False
+                "metadata": {"type": "error"}
+            }
 
-        for key, code in index_mapping.items():
+        stock_code = stock_info['code']
-            if key in stock_keyword or stock_keyword in key:
+        stock_name = stock_info['name']
-                stock_code = code
+        is_index = stock_info['is_index']
-                stock_name = key if key in stock_keyword else stock_keyword
-                is_index = True
-                logger.info(f"识别为指数查询: {stock_name} -> {stock_code}")
-                break
-
-        # 如果不是指数，使用Tushare搜索股票
-        if not is_index:
-            search_results = tushare_service.search_stock(stock_keyword)
-
-            if not search_results:
-                return {
-                    "message": f"抱歉，未找到股票\"{stock_keyword}\"。请确认股票名称或代码是否正确。",
-                    "metadata": {"type": "error"}
-                }
-
-            stock = search_results[0]
-            stock_code = stock['symbol']
-            stock_name = stock['name']
 
         logger.info(f"处理{'指数' if is_index else '股票'}问题: {stock_name}({stock_code})")
 
@@ -1152,6 +1129,126 @@ MA60：{f"{ma['ma60']:.2f}" if ma['ma60'] else '计算中'}
         else:
             return await self._single_query(stock_code, stock_name, message)
 
+    async def _match_stock_with_llm(self, keyword: str) -> Optional[Dict[str, Any]]:
+        """
+        使用LLM智能匹配股票或指数
+
+        Args:
+            keyword: 用户输入的关键词
+
+        Returns:
+            匹配结果: {'code': '股票代码', 'name': '股票名称', 'is_index': bool}
+        """
+        if not self.use_llm:
+            # 降级方案：使用Tushare搜索
+            search_results = tushare_service.search_stock(keyword)
+            if search_results:
+                return {
+                    'code': search_results[0]['symbol'],
+                    'name': search_results[0]['name'],
+                    'is_index': False
+                }
+            return None
+
+        prompt = f"""你是一个专业的A股市场专家。请根据用户输入的关键词，识别对应的股票代码或指数代码。
+
+用户输入：{keyword}
+
+常见指数代码：
+- 上证指数/大盘/沪指/A股 → 000001.SH
+- 深证成指/深证/深指 → 399001.SZ
+- 创业板指/创业板 → 399006.SZ
+- 科创50 → 000688.SH
+- 沪深300 → 000300.SH
+- 中证500 → 000905.SH
+
+如果是指数，请直接返回对应的指数代码。
+如果是股票名称或代码，请使用Tushare数据库进行搜索匹配。
+
+请以JSON格式返回：
+{{
+    "is_index": true/false,
+    "code": "股票或指数代码（如000001.SH）",
+    "name": "股票或指数名称",
+    "confidence": 0.0-1.0
+}}
+
+如果无法匹配，返回：
+{{
+    "is_index": false,
+    "code": null,
+    "name": null,
+    "confidence": 0.0
+}}
+
+只返回JSON，不要有任何其他内容。"""
+
+        try:
+            result = await self._call_llm_async(
+                messages=[{"role": "user", "content": prompt}],
+                temperature=0.3,
+                max_tokens=200
+            )
+
+            if not result:
+                logger.warning("LLM匹配返回空结果")
+                return None
+
+            # 清理结果
+            result = result.strip()
+            if result.startswith("```json"):
+                result = result[7:]
+            if result.startswith("```"):
+                result = result[3:]
+            if result.endswith("```"):
+                result = result[:-3]
+            result = result.strip()
+
+            # 解析JSON
+            match_result = json.loads(result)
+
+            # 如果LLM无法匹配或置信度太低，使用Tushare搜索
+            if not match_result.get('code') or match_result.get('confidence', 0) < 0.5:
+                logger.info(f"LLM匹配置信度低，使用Tushare搜索: {keyword}")
+                search_results = tushare_service.search_stock(keyword)
+                if search_results:
+                    return {
+                        'code': search_results[0]['symbol'],
+                        'name': search_results[0]['name'],
+                        'is_index': False
+                    }
+                return None
+
+            logger.info(f"LLM匹配成功: {keyword} -> {match_result['name']}({match_result['code']})")
+            return {
+                'code': match_result['code'],
+                'name': match_result['name'],
+                'is_index': match_result['is_index']
+            }
+
+        except json.JSONDecodeError as e:
+            logger.error(f"LLM匹配JSON解析失败: {e}, 原始响应: {result[:200] if result else 'None'}")
+            # 降级方案
+            search_results = tushare_service.search_stock(keyword)
+            if search_results:
+                return {
+                    'code': search_results[0]['symbol'],
+                    'name': search_results[0]['name'],
+                    'is_index': False
+                }
+            return None
+        except Exception as e:
+            logger.error(f"LLM匹配失败: {e}")
+            # 降级方案
+            search_results = tushare_service.search_stock(keyword)
+            if search_results:
+                return {
+                    'code': search_results[0]['symbol'],
+                    'name': search_results[0]['name'],
+                    'is_index': False
+                }
+            return None
+
     async def _handle_macro_question(
         self,
         intent_analysis: Dict[str, Any],
@@ -1193,7 +1290,7 @@ MA60：{f"{ma['ma60']:.2f}" if ma['ma60'] else '计算中'}
 5. 最后声明："以上分析仅供参考，不构成投资建议。股市有风险，投资需谨慎。"
 """
 
-            analysis = llm_service.chat(
+            analysis = await self._call_llm_async(
                 messages=[{"role": "user", "content": prompt}],
                 temperature=0.7,
                 max_tokens=1500
@@ -1262,7 +1359,7 @@ MA60：{f"{ma['ma60']:.2f}" if ma['ma60'] else '计算中'}
 """
 
         try:
-            answer = llm_service.chat(
+            answer = await self._call_llm_async(
                 messages=[{"role": "user", "content": prompt}],
                 temperature=0.7,
                 max_tokens=1200
@@ -1318,7 +1415,7 @@ MA60：{f"{ma['ma60']:.2f}" if ma['ma60'] else '计算中'}
 直接返回回复内容，不要有其他格式。"""
 
         try:
-            reply = llm_service.chat(
+            reply = await self._call_llm_async(
                 messages=[{"role": "user", "content": prompt}],
                 temperature=0.8,
                 max_tokens=400
@@ -1584,7 +1681,7 @@ MACD：{f"{technical.get('macd'):.4f}" if technical.get('macd') else '计算中'
 """
 
         try:
-            analysis = llm_service.chat(
+            analysis = await self._call_llm_async(
                 messages=[{"role": "user", "content": prompt}],
                 temperature=0.7,
                 max_tokens=2000
@@ -1746,34 +1843,16 @@ RSI：{technical.get('rsi', 0):.2f if technical.get('rsi') else '计算中'}
 
     async def _handle_a_stock_stream(self, stock_keyword: str, message: str):
         """流式处理A股分析"""
-        # 指数映射
+        # 使用LLM进行智能匹配
-        index_mapping = {
+        stock_info = await self._match_stock_with_llm(stock_keyword)
-            "上证指数": "000001.SH", "上证": "000001.SH", "大盘": "000001.SH", "沪指": "000001.SH",
-            "深证成指": "399001.SZ", "深证": "399001.SZ", "深指": "399001.SZ",
-            "创业板指": "399006.SZ", "创业板": "399006.SZ",
-            "科创50": "000688.SH", "沪深300": "000300.SH", "中证500": "000905.SH",
-            "A股": "000001.SH"
-        }
 
-        stock_code = None
+        if not stock_info:
-        stock_name = None
+            yield f"抱歉，未找到股票或指数\"{stock_keyword}\"。请确认名称或代码是否正确。"
-        is_index = False
+            return
 
-        for key, code in index_mapping.items():
+        stock_code = stock_info['code']
-            if key in stock_keyword or stock_keyword in key:
+        stock_name = stock_info['name']
-                stock_code = code
+        is_index = stock_info['is_index']
-                stock_name = key if key in stock_keyword else stock_keyword
-                is_index = True
-                break
-
-        if not is_index:
-            search_results = tushare_service.search_stock(stock_keyword)
-            if not search_results:
-                yield f"抱歉，未找到股票\"{stock_keyword}\"。请确认股票名称或代码是否正确。"
-                return
-            stock = search_results[0]
-            stock_code = stock['symbol']
-            stock_name = stock['name']
 
         # 获取数据（非流式）
         try:
@@ -1926,13 +2005,19 @@ RSI：{technical.get('rsi', 0):.2f if technical.get('rsi') else '计算中'}
 5. 最后声明："以上分析仅供参考，不构成投资建议。股市有风险，投资需谨慎。"
 """
 
-        # 流式调用LLM（同步生成器）
+        # 流式调用LLM（同步生成器，使用线程避免阻塞）
+        import asyncio
         stream = llm_service.chat_stream(
             messages=[{"role": "user", "content": prompt}],
             temperature=0.7,
             max_tokens=2000
         )
+
+        # 在线程中迭代同步生成器，避免阻塞事件循环
+        loop = asyncio.get_event_loop()
         for chunk in stream:
+            # 每次yield后让出控制权
+            await asyncio.sleep(0)
             yield chunk
 
     async def _llm_us_stock_analysis_stream(self, data: Dict[str, Any], user_message: str):
@@ -2022,13 +2107,18 @@ MACD：{f"{technical.get('macd'):.4f}" if technical.get('macd') else '计算中'
 6. 最后声明："以上分析仅供参考，不构成投资建议。美股投资有风险，请谨慎决策。"
 """
 
-        # 流式调用LLM（同步生成器）
+        # 流式调用LLM（同步生成器，使用线程避免阻塞）
+        import asyncio
         stream = llm_service.chat_stream(
             messages=[{"role": "user", "content": prompt}],
             temperature=0.7,
             max_tokens=2000
         )
+
+        # 在线程中迭代同步生成器，避免阻塞事件循环
         for chunk in stream:
+            # 每次yield后让出控制权
+            await asyncio.sleep(0)
             yield chunk
 
 

backend/app/services/multi_llm_service.py

@@ -146,10 +146,12 @@ class MultiLLMService:
                 )
             elif provider == 'deepseek':
                 # DeepSeek调用（OpenAI兼容）
+                # DeepSeek对参数更严格，确保temperature在有效范围内
+                safe_temperature = max(0.0, min(2.0, temperature))
                 response = client.chat.completions.create(
                     model=model_id,
                     messages=messages,
-                    temperature=temperature,
+                    temperature=safe_temperature,
                     max_tokens=max_tokens
                 )
             else:
@@ -216,10 +218,12 @@ class MultiLLMService:
 
             elif provider == 'deepseek':
                 # DeepSeek流式调用（OpenAI兼容）
+                # DeepSeek对参数更严格，确保temperature在有效范围内
+                safe_temperature = max(0.0, min(2.0, temperature))
                 response = client.chat.completions.create(
                     model=model_id,
                     messages=messages,
-                    temperature=temperature,
+                    temperature=safe_temperature,
                     max_tokens=max_tokens,
                     stream=True
                 )

frontend/index.html

@@ -64,10 +64,9 @@
                         <div class="example-queries">
                             <button class="example-btn" @click="sendExample('分析贵州茅台')">分析贵州茅台</button>
                             <button class="example-btn" @click="sendExample('比亚迪怎么样')">比亚迪怎么样</button>
-                            <button class="example-btn" @click="sendExample('上证指数走势')">上证指数走势</button>
                             <button class="example-btn" @click="sendExample('分析特斯拉')">分析特斯拉</button>
                             <button class="example-btn" @click="sendExample('苹果股票')">苹果股票</button>
-                            <button class="example-btn" @click="sendExample('NVDA基本面')">NVDA基本面</button>
+                            <button class="example-btn" @click="sendExample('英伟达股票怎么样')">英伟达股票怎么样</button>
                         </div>
                     </div>