update pay function

mcp_server.py

@@ -2337,7 +2337,37 @@ async def search_chat_history(user_id: str, query: str, top_k: int = 10):
         raise HTTPException(status_code=500, detail=str(e))
 
 
-# ==================== 投研会议室系统 ====================
+# ==================== 投研会议室系统 (V2 - 流式+工具调用) ====================
+
+import random
+
+# 投研会议室专用模型配置
+MEETING_MODEL_CONFIGS = {
+    "kimi-k2-thinking": {
+        "api_key": "sk-TzB4VYJfCoXGcGrGMiewukVRzjuDsbVCkaZXi2LvkS8s60E5",
+        "base_url": "https://api.moonshot.cn/v1",
+        "model": "kimi-k2-thinking",
+    },
+    "deepseek": {
+        "api_key": "sk-7363bdb28d7d4bf0aa68eb9449f8f063",
+        "base_url": "https://api.deepseek.com",
+        "model": "deepseek-chat",
+    },
+    "deepmoney": {
+        "api_key": "",
+        "base_url": "http://111.62.35.50:8000/v1",
+        "model": "deepmoney",
+    },
+}
+
+# 每个角色可用的工具列表
+ROLE_TOOLS = {
+    "buffett": ["search_china_news", "search_research_reports", "get_stock_basic_info", "get_stock_financial_index"],
+    "big_short": ["search_china_news", "get_stock_financial_index", "get_stock_balance_sheet", "get_stock_cashflow"],
+    "simons": ["get_stock_trade_data", "search_limit_up_stocks", "get_concept_statistics"],
+    "leek": [],  # 韭菜不用工具
+    "fund_manager": ["search_china_news", "search_research_reports", "get_stock_basic_info"],
+}
 
 # 投研会议室角色配置
 MEETING_ROLES = {
@@ -2345,508 +2375,435 @@ MEETING_ROLES = {
         "id": "buffett",
         "name": "巴菲特",
         "nickname": "唱多者",
-        "role_type": "bull",  # 多头
+        "role_type": "bull",
         "avatar": "/avatars/buffett.png",
         "model": "kimi-k2-thinking",
-        "color": "#10B981",  # 绿色（上涨）
+        "color": "#10B981",
         "description": "主观多头，善于分析事件的潜在利好和长期价值",
+        "tools": ROLE_TOOLS["buffett"],
         "system_prompt": """你是"巴菲特"，一位资深的价值投资者和主观多头分析师。
 
 你的特点：
 1. 善于发现事件和公司的潜在利好因素
-2. 关注长期价值，不被短期波动干扰
+2. 关注长期价值，分析护城河、竞争优势
-3. 分析公司的护城河、竞争优势和管理层质量
+3. 对市场保持乐观但理性的态度
-4. 对市场保持乐观但理性的态度
 
-分析风格：
+你可以使用以下工具获取数据：
-- 重点挖掘利好因素和投资机会
+- search_china_news: 搜索新闻
-- 从产业链、市场格局、政策支持等角度分析
+- search_research_reports: 搜索研报
-- 给出清晰的看多逻辑和目标预期
+- get_stock_basic_info: 获取股票基本信息
-- 语言风格：稳重、专业、富有洞察力
+- get_stock_financial_index: 获取财务指标
 
-注意：你的发言要简洁有力，每次发言控制在200字以内。直接表达观点，不要客套。"""
+分析时请先调用工具获取数据，再基于数据发表看多观点。
+注意：参考前面其他人的发言，进行有针对性的回应。发言控制在200字以内。"""
     },
     "big_short": {
         "id": "big_short",
         "name": "大空头",
         "nickname": "大空头",
-        "role_type": "bear",  # 空头
+        "role_type": "bear",
         "avatar": "/avatars/big_short.png",
         "model": "kimi-k2-thinking",
-        "color": "#EF4444",  # 红色（下跌）
+        "color": "#EF4444",
-        "description": "善于分析事件和财报中的风险因素，帮助投资者避雷",
+        "description": "善于分析事件和财报中的风险因素",
-        "system_prompt": """你是"大空头"，一位专业的风险分析师和空头研究员。
+        "tools": ROLE_TOOLS["big_short"],
+        "system_prompt": """你是"大空头"，一位专业的风险分析师。
 
 你的特点：
 1. 善于发现被市场忽视的风险因素
-2. 擅长财报分析，发现财务造假和粉饰的迹象
+2. 擅长财报分析，发现财务造假迹象
-3. 关注行业天花板、竞争加剧、估值泡沫等问题
+3. 关注行业天花板、竞争加剧、估值泡沫
-4. 对市场保持警惕，帮助投资者避雷
 
-分析风格：
+你可以使用以下工具获取数据：
-- 重点挖掘风险因素和潜在隐患
+- search_china_news: 搜索负面新闻
-- 从财务数据、行业周期、估值水平等角度分析
+- get_stock_financial_index: 获取财务指标找问题
-- 给出清晰的风险提示和规避建议
+- get_stock_balance_sheet: 分析资产负债表
-- 语言风格：犀利、直接、善于质疑
+- get_stock_cashflow: 分析现金流
 
-注意：你的发言要简洁有力，每次发言控制在200字以内。直接指出风险，不要绕弯子。"""
+分析时请先调用工具获取数据，再基于数据指出风险。
+注意：参考前面其他人的发言，进行有针对性的反驳。发言控制在200字以内。"""
     },
     "simons": {
         "id": "simons",
         "name": "量化分析员",
         "nickname": "西蒙斯",
-        "role_type": "quant",  # 量化
+        "role_type": "quant",
         "avatar": "/avatars/simons.png",
-        "model": "deepseek-v3",
+        "model": "deepseek",
-        "color": "#3B82F6",  # 蓝色（中性）
+        "color": "#3B82F6",
-        "description": "中性立场，使用量化分析工具分析技术指标",
+        "description": "中性立场，使用量化工具分析技术指标",
+        "tools": ROLE_TOOLS["simons"],
         "system_prompt": """你是"量化分析员"（昵称：西蒙斯），一位专业的量化交易研究员。
 
 你的特点：
 1. 使用数据和技术指标说话，保持中性立场
-2. 擅长均线分析、量价关系、动能指标等技术分析
+2. 擅长均线、量价、动能指标分析
-3. 关注市场情绪、资金流向、筹码分布等量化因素
+3. 用概率思维看待市场
-4. 用概率思维看待市场，不做主观臆断
 
-分析风格：
+你可以使用以下工具获取数据：
-- 基于技术指标给出客观分析
+- get_stock_trade_data: 获取交易数据（价格、成交量）
-- 使用具体数据支撑观点（如：5日均线、MACD、RSI等）
+- search_limit_up_stocks: 搜索涨停股票
-- 给出量化的买卖信号和风险评估
+- get_concept_statistics: 获取概念板块统计
-- 语言风格：理性、客观、数据驱动
 
-注意：你的发言要简洁有力，每次发言控制在200字以内。多用数据说话，少发表主观意见。"""
+分析时请先调用工具获取数据，再基于数据给出技术分析。
+注意：参考前面其他人的发言，用数据说话。发言控制在200字以内。"""
     },
     "leek": {
         "id": "leek",
         "name": "韭菜",
         "nickname": "牢大",
-        "role_type": "retail",  # 散户
+        "role_type": "retail",
         "avatar": "/avatars/leek.png",
         "model": "deepmoney",
-        "color": "#F59E0B",  # 黄色
+        "color": "#F59E0B",
-        "description": "贪婪又讨厌亏损，热爱追涨杀跌的典型散户",
+        "description": "贪婪又讨厌亏损，热爱追涨杀跌",
+        "tools": [],
         "system_prompt": """你是"韭菜"（昵称：牢大），一个典型的散户投资者。
 
 你的特点：
-1. 贪婪但又害怕亏损，典型的追涨杀跌
+1. 贪婪但又害怕亏损，追涨杀跌
-2. 容易被市场情绪影响，看到涨就想追，看到跌就想跑
+2. 容易被市场情绪影响
-3. 喜欢听小道消息，容易被"内幕"吸引
+3. 喜欢听小道消息，期望一夜暴富
-4. 短线思维，缺乏耐心，期望一夜暴富
 
-分析风格：
+你不需要调用工具，直接用散户视角发表看法。
-- 用最朴素的散户思维来分析问题
+注意：参考前面其他人的发言，用最朴素的方式回应。语言口语化、情绪化。发言控制在150字以内。"""
-- 经常关注"这个能赚多少"、"会不会跌"
-- 容易情绪化，看涨时过度乐观，看跌时过度悲观
-- 语言风格：口语化、情绪化、接地气
-
-注意：你的发言要简洁直接，每次发言控制在150字以内。展现真实散户的心态，可以有些搞笑，但不要太出格。"""
     },
     "fund_manager": {
         "id": "fund_manager",
         "name": "基金经理",
         "nickname": "决策者",
-        "role_type": "manager",  # 管理者
+        "role_type": "manager",
         "avatar": "/avatars/fund_manager.png",
         "model": "kimi-k2-thinking",
-        "color": "#8B5CF6",  # 紫色
+        "color": "#8B5CF6",
-        "description": "总结其他人的发言做出最终决策",
+        "description": "综合分析做出最终决策",
-        "system_prompt": """你是"基金经理"，投研会议的主持人和最终决策者。
+        "tools": ROLE_TOOLS["fund_manager"],
+        "system_prompt": """你是"基金经理"，投研会议的最终决策者。
 
 你的角色：
 1. 综合各方观点，做出理性判断
 2. 平衡多空观点，识别有价值的分析
-3. 特别注意：韭菜的观点通常是反向指标
+3. 注意：韭菜的观点通常是反向指标
-4. 给出专业、负责任的投资建议
 
-决策风格：
+如果需要补充信息，可以调用工具：
-- 综合考虑基本面、技术面、情绪面
+- search_china_news: 搜索新闻
-- 权衡风险与收益，给出明确的投资建议
+- search_research_reports: 搜索研报
-- 指出讨论中的关键洞察和需要注意的风险
+- get_stock_basic_info: 获取股票基本信息
-- 语言风格：权威、专业、全面
 
 决策输出格式：
-1. 综合评估：对讨论议题的整体判断
+1. 综合评估
-2. 关键观点：各方有价值的观点总结
+2. 关键观点
-3. 风险提示：需要注意的主要风险
+3. 风险提示
-4. 操作建议：具体的投资建议（买入/持有/观望/卖出）
+4. 操作建议（买入/持有/观望/卖出）
-5. 信心指数：对这个结论的信心程度（1-10分）
+5. 信心指数（1-10分）
 
-注意：如果讨论还不够充分，你可以要求继续讨论。每次发言控制在300字以内。"""
+参考前面所有人的发言，给出综合判断。发言控制在300字以内。"""
     }
 }
 
-# 投研会议室专用模型配置（扩展现有配置）
-MEETING_MODEL_CONFIGS = {
-    **MODEL_CONFIGS,
-    "deepseek-v3": {
-        "api_key": "sk-1cf3dfadf7244a8680cd0a60da6f1efd",
-        "base_url": "https://api.deepseek.com/v1",
-        "model": "deepseek-chat",
-    }
-}
-
-
-class MeetingRoleMessage(BaseModel):
-    """会议角色消息"""
-    role_id: str
-    role_name: str
-    nickname: str
-    avatar: str
-    color: str
-    content: str
-    timestamp: str
-    round_number: int  # 第几轮讨论
-
 
 class MeetingRequest(BaseModel):
     """投研会议请求"""
-    topic: str  # 用户提出的议题
+    topic: str
     user_id: str = "anonymous"
     user_nickname: str = "匿名用户"
     session_id: Optional[str] = None
-    user_message: Optional[str] = None  # 用户在讨论中的插话
+    user_message: Optional[str] = None
-    conversation_history: List[Dict[str, Any]] = []  # 之前的讨论历史
+    conversation_history: List[Dict[str, Any]] = []
 
 
-class MeetingResponse(BaseModel):
+def get_random_speaking_order() -> List[str]:
-    """投研会议响应"""
+    """随机生成发言顺序（不包括基金经理）"""
-    success: bool
+    roles = ["buffett", "big_short", "simons", "leek"]
-    session_id: str
+    random.shuffle(roles)
-    messages: List[Dict[str, Any]]  # 本轮所有角色的发言
+    return roles
-    round_number: int  # 当前轮次
-    is_concluded: bool  # 是否已得出结论
-    conclusion: Optional[Dict[str, Any]] = None  # 基金经理的结论（如果有）
 
 
-async def call_role_llm(role_id: str, prompt: str, context: str = "") -> str:
+async def call_role_tool(role_id: str, tool_name: str, arguments: dict) -> dict:
-    """调用特定角色的LLM生成回复"""
+    """调用角色的工具"""
+    handler = TOOL_HANDLERS.get(tool_name)
+    if not handler:
+        return {"success": False, "error": f"Unknown tool: {tool_name}"}
+
+    try:
+        result = await handler(arguments)
+        return {"success": True, "tool": tool_name, "result": result}
+    except Exception as e:
+        logger.error(f"Tool {tool_name} failed: {e}")
+        return {"success": False, "tool": tool_name, "error": str(e)}
+
+
+async def stream_role_response(
+    role_id: str,
+    topic: str,
+    context: str,
+    tools: List[dict]
+) -> AsyncGenerator[dict, None]:
+    """流式生成角色回复，支持工具调用"""
     role = MEETING_ROLES.get(role_id)
     if not role:
-        raise ValueError(f"Unknown role: {role_id}")
+        yield {"type": "error", "error": f"Unknown role: {role_id}"}
+        return
 
     model_name = role["model"]
-    model_config = MEETING_MODEL_CONFIGS.get(model_name, MODEL_CONFIGS["kimi-k2-thinking"])
+    model_config = MEETING_MODEL_CONFIGS.get(model_name)
+    if not model_config:
+        yield {"type": "error", "error": f"Unknown model: {model_name}"}
+        return
 
     try:
         client = OpenAI(
             api_key=model_config["api_key"],
-            base_url=model_config["base_url"]
+            base_url=model_config["base_url"],
+            timeout=180
         )
 
         messages = [
             {"role": "system", "content": role["system_prompt"]},
+            {"role": "user", "content": f"议题：{topic}\n\n{context}"}
         ]
 
-        if context:
+        # 准备工具定义（如果该角色有工具）
-            messages.append({"role": "user", "content": f"当前讨论背景：\n{context}"})
+        role_tool_names = role.get("tools", [])
+        openai_tools = None
+        if role_tool_names:
+            openai_tools = []
+            for tool in TOOLS:
+                if tool.name in role_tool_names:
+                    openai_tools.append({
+                        "type": "function",
+                        "function": {
+                            "name": tool.name,
+                            "description": tool.description,
+                            "parameters": tool.parameters
+                        }
+                    })
 
-        messages.append({"role": "user", "content": prompt})
+        # 第一次调用：可能触发工具调用
+        tool_calls_made = []
+        if openai_tools:
+            response = client.chat.completions.create(
+                model=model_config["model"],
+                messages=messages,
+                tools=openai_tools,
+                tool_choice="auto",
+                stream=False,  # 工具调用不使用流式
+                temperature=0.7,
+                max_tokens=1000,
+            )
 
-        response = client.chat.completions.create(
+            assistant_message = response.choices[0].message
+
+            # 处理工具调用
+            if assistant_message.tool_calls:
+                messages.append(assistant_message)
+
+                for tool_call in assistant_message.tool_calls:
+                    tool_name = tool_call.function.name
+                    try:
+                        arguments = json.loads(tool_call.function.arguments)
+                    except:
+                        arguments = {}
+
+                    # 发送工具调用开始事件
+                    yield {
+                        "type": "tool_call_start",
+                        "tool": tool_name,
+                        "arguments": arguments
+                    }
+
+                    # 执行工具调用
+                    result = await call_role_tool(role_id, tool_name, arguments)
+                    tool_calls_made.append(result)
+
+                    # 发送工具调用结果事件
+                    yield {
+                        "type": "tool_call_result",
+                        "tool": tool_name,
+                        "result": result
+                    }
+
+                    # 添加工具结果到消息
+                    messages.append({
+                        "role": "tool",
+                        "tool_call_id": tool_call.id,
+                        "content": json.dumps(result, ensure_ascii=False)
+                    })
+
+        # 流式生成最终回复
+        stream = client.chat.completions.create(
             model=model_config["model"],
             messages=messages,
+            stream=True,
             temperature=0.7,
             max_tokens=500,
         )
 
-        return response.choices[0].message.content.strip()
+        full_content = ""
-
+        for chunk in stream:
-    except Exception as e:
+            if chunk.choices and chunk.choices[0].delta.content:
-        logger.error(f"调用角色 {role_id} 的 LLM 失败: {e}")
+                content = chunk.choices[0].delta.content
-        return f"[{role['name']}暂时无法发言，请稍后重试]"
+                full_content += content
-
+                yield {
-
+                    "type": "content_delta",
-async def determine_speaking_order(topic: str) -> List[str]:
+                    "content": content
-    """使用 K2 模型决定发言顺序"""
-    try:
-        client = OpenAI(
-            api_key=MODEL_CONFIGS["kimi-k2-thinking"]["api_key"],
-            base_url=MODEL_CONFIGS["kimi-k2-thinking"]["base_url"]
-        )
-
-        response = client.chat.completions.create(
-            model=MODEL_CONFIGS["kimi-k2-thinking"]["model"],
-            messages=[
-                {
-                    "role": "system",
-                    "content": """你是一个会议主持助手。根据用户提出的议题，决定投研会议中各角色的最佳发言顺序。
-
-可用角色（不包括基金经理，他最后总结）：
-- buffett: 巴菲特（主观多头，分析利好）
-- big_short: 大空头（风险分析师）
-- simons: 量化分析员（技术分析）
-- leek: 韭菜（散户视角）
-
-根据议题性质，安排最合适的发言顺序。比如：
-- 如果是分析某公司/事件，建议先让多头分析利好，再让空头分析风险
-- 如果是技术走势问题，可以先让量化分析
-- 韭菜可以随时插入，提供散户视角
-
-只需要返回角色ID列表，用逗号分隔，例如：buffett,simons,big_short,leek"""
-                },
-                {"role": "user", "content": f"议题：{topic}"}
-            ],
-            temperature=0.3,
-            max_tokens=100,
-        )
-
-        order_str = response.choices[0].message.content.strip()
-        # 解析返回的顺序
-        order = [r.strip() for r in order_str.split(",") if r.strip() in MEETING_ROLES]
-
-        # 确保所有非管理者角色都在列表中
-        for role_id, role in MEETING_ROLES.items():
-            if role["role_type"] != "manager" and role_id not in order:
-                order.append(role_id)
-
-        return order
-
-    except Exception as e:
-        logger.error(f"决定发言顺序失败: {e}")
-        # 返回默认顺序
-        return ["buffett", "big_short", "simons", "leek"]
-
-
-async def check_conclusion_ready(discussion_history: str, topic: str) -> tuple[bool, str]:
-    """基金经理判断是否可以得出结论"""
-    try:
-        client = OpenAI(
-            api_key=MODEL_CONFIGS["kimi-k2-thinking"]["api_key"],
-            base_url=MODEL_CONFIGS["kimi-k2-thinking"]["base_url"]
-        )
-
-        response = client.chat.completions.create(
-            model=MODEL_CONFIGS["kimi-k2-thinking"]["model"],
-            messages=[
-                {
-                    "role": "system",
-                    "content": MEETING_ROLES["fund_manager"]["system_prompt"]
-                },
-                {
-                    "role": "user",
-                    "content": f"""议题：{topic}
-
-目前的讨论内容：
-{discussion_history}
-
-请判断：
-1. 目前的讨论是否足够充分，可以得出最终结论？
-2. 如果可以，请给出你的最终决策。
-3. 如果不可以，请说明还需要讨论什么，并要求继续讨论。
-
-请以JSON格式回复：
-{{
-    "can_conclude": true/false,
-    "reasoning": "判断理由",
-    "conclusion": "如果可以结论，这里是你的完整决策；如果不能，这里是需要继续讨论的方向"
-}}"""
                 }
-            ],
-            temperature=0.5,
-            max_tokens=800,
-        )
 
-        result = response.choices[0].message.content.strip()
+        # 发送完成事件
-        # 尝试解析JSON
+        yield {
-        try:
+            "type": "content_done",
-            # 处理可能的 markdown 代码块
+            "full_content": full_content,
-            if "```json" in result:
+            "tool_calls": tool_calls_made
-                result = result.split("```json")[1].split("```")[0].strip()
+        }
-            elif "```" in result:
-                result = result.split("```")[1].split("```")[0].strip()
-
-            data = json.loads(result)
-            return data.get("can_conclude", False), data.get("conclusion", result)
-        except json.JSONDecodeError:
-            # 如果JSON解析失败，直接返回内容
-            return True, result
 
     except Exception as e:
-        logger.error(f"检查结论状态失败: {e}")
+        logger.error(f"Role {role_id} stream failed: {e}")
-        return True, "基于目前的讨论，建议投资者谨慎对待，继续关注后续发展。"
+        yield {"type": "error", "error": str(e)}
 
 
-@app.post("/agent/meeting/start")
+@app.post("/agent/meeting/stream")
-async def start_investment_meeting(request: MeetingRequest):
+async def stream_investment_meeting(request: MeetingRequest):
     """
-    启动投研会议
+    流式投研会议 V2
 
-    第一轮：所有角色（除基金经理外）依次发言
+    - 随机发言顺序
+    - 每个角色流式输出
+    - 支持工具调用
+    - 支持用户中途发言
     """
-    logger.info(f"启动投研会议: {request.topic} (user: {request.user_id})")
+    logger.info(f"[Meeting V2] 启动: {request.topic}")
 
-    session_id = request.session_id or str(uuid.uuid4())
+    async def generate_meeting_stream() -> AsyncGenerator[str, None]:
-    messages = []
+        session_id = request.session_id or str(uuid.uuid4())
-    round_number = 1
+        round_number = len(request.conversation_history) // 5 + 1
 
-    # 决定发言顺序
+        # 发送会话开始
-    speaking_order = await determine_speaking_order(request.topic)
+        yield f"data: {json.dumps({'type': 'session_start', 'session_id': session_id, 'round': round_number}, ensure_ascii=False)}\n\n"
-    logger.info(f"发言顺序: {speaking_order}")
 
-    # 构建讨论上下文
+        # 构建上下文
-    context = f"议题：{request.topic}\n\n这是第一轮讨论，请针对议题发表你的观点。"
+        context_parts = []
+        if request.conversation_history:
+            context_parts.append("之前的讨论：")
+            for msg in request.conversation_history:
+                context_parts.append(f"【{msg.get('role_name', '未知')}】：{msg.get('content', '')}")
 
-    # 依次让每个角色发言
-    for role_id in speaking_order:
-        role = MEETING_ROLES[role_id]
-        if role["role_type"] == "manager":
-            continue  # 基金经理不在第一轮发言
-
-        # 加入之前角色的发言作为上下文
-        prev_context = context
-        if messages:
-            prev_context += "\n\n其他人的观点：\n"
-            for msg in messages:
-                prev_context += f"- {msg['role_name']}：{msg['content']}\n"
-
-        # 调用LLM生成发言
-        content = await call_role_llm(role_id, request.topic, prev_context)
-
-        message = {
-            "role_id": role_id,
-            "role_name": role["name"],
-            "nickname": role["nickname"],
-            "avatar": role["avatar"],
-            "color": role["color"],
-            "content": content,
-            "timestamp": datetime.now().isoformat(),
-            "round_number": round_number
-        }
-        messages.append(message)
-
-    # 第一轮结束后，基金经理判断是否可以得出结论
-    discussion_summary = "\n".join([
-        f"【{msg['role_name']}】：{msg['content']}"
-        for msg in messages
-    ])
-
-    can_conclude, conclusion_content = await check_conclusion_ready(discussion_summary, request.topic)
-
-    # 添加基金经理的发言
-    fund_manager = MEETING_ROLES["fund_manager"]
-    fund_manager_message = {
-        "role_id": "fund_manager",
-        "role_name": fund_manager["name"],
-        "nickname": fund_manager["nickname"],
-        "avatar": fund_manager["avatar"],
-        "color": fund_manager["color"],
-        "content": conclusion_content,
-        "timestamp": datetime.now().isoformat(),
-        "round_number": round_number,
-        "is_conclusion": can_conclude
-    }
-    messages.append(fund_manager_message)
-
-    return {
-        "success": True,
-        "session_id": session_id,
-        "messages": messages,
-        "round_number": round_number,
-        "is_concluded": can_conclude,
-        "conclusion": fund_manager_message if can_conclude else None
-    }
-
-
-@app.post("/agent/meeting/continue")
-async def continue_investment_meeting(request: MeetingRequest):
-    """
-    继续投研会议讨论
-
-    根据之前的讨论历史，继续新一轮讨论
-    支持用户在讨论中插话
-    """
-    logger.info(f"继续投研会议: {request.topic} (round: {len(request.conversation_history) // 5 + 1})")
-
-    session_id = request.session_id or str(uuid.uuid4())
-    messages = []
-    round_number = len(request.conversation_history) // 5 + 2  # 估算轮次
-
-    # 构建历史讨论上下文
-    history_context = "历史讨论：\n"
-    for msg in request.conversation_history:
-        history_context += f"【{msg.get('role_name', '未知')}】：{msg.get('content', '')}\n"
-
-    # 如果用户有插话，先处理用户消息
-    if request.user_message:
-        history_context += f"\n【用户】：{request.user_message}\n"
-        messages.append({
-            "role_id": "user",
-            "role_name": "用户",
-            "nickname": request.user_nickname,
-            "avatar": "",
-            "color": "#6366F1",
-            "content": request.user_message,
-            "timestamp": datetime.now().isoformat(),
-            "round_number": round_number
-        })
-
-    # 新一轮讨论的发言顺序
-    speaking_order = await determine_speaking_order(request.topic)
-
-    # 依次让每个角色发言
-    for role_id in speaking_order:
-        role = MEETING_ROLES[role_id]
-        if role["role_type"] == "manager":
-            continue
-
-        # 构建本次发言的上下文
-        current_context = f"议题：{request.topic}\n\n{history_context}"
-        if messages:
-            current_context += "\n本轮讨论：\n"
-            for msg in messages:
-                if msg["role_id"] != "user":
-                    current_context += f"- {msg['role_name']}：{msg['content']}\n"
-
-        # 调用LLM
-        prompt = f"这是第{round_number}轮讨论，请根据之前的讨论内容，进一步阐述或补充你的观点。"
         if request.user_message:
-            prompt += f"\n\n用户刚才说：{request.user_message}\n请也回应用户的观点。"
+            context_parts.append(f"\n用户刚才说：{request.user_message}")
 
-        content = await call_role_llm(role_id, prompt, current_context)
+        context = "\n".join(context_parts) if context_parts else "这是第一轮讨论，请针对议题发表你的观点。"
 
-        message = {
+        # 随机发言顺序
-            "role_id": role_id,
+        speaking_order = get_random_speaking_order()
-            "role_name": role["name"],
+        yield f"data: {json.dumps({'type': 'order_decided', 'order': speaking_order}, ensure_ascii=False)}\n\n"
-            "nickname": role["nickname"],
+
-            "avatar": role["avatar"],
+        all_messages = []
-            "color": role["color"],
+        accumulated_context = context
-            "content": content,
+
+        # 依次让每个角色发言
+        for role_id in speaking_order:
+            role = MEETING_ROLES[role_id]
+
+            # 发送开始发言事件
+            yield f"data: {json.dumps({'type': 'speaking_start', 'role_id': role_id, 'role_name': role['name'], 'color': role['color']}, ensure_ascii=False)}\n\n"
+
+            # 准备工具列表
+            role_tools = [t for t in TOOLS if t.name in role.get("tools", [])]
+
+            # 流式生成回复
+            full_content = ""
+            tool_calls = []
+
+            async for event in stream_role_response(role_id, request.topic, accumulated_context, role_tools):
+                if event["type"] == "tool_call_start":
+                    yield f"data: {json.dumps({'type': 'tool_call_start', 'role_id': role_id, 'tool': event['tool'], 'arguments': event['arguments']}, ensure_ascii=False)}\n\n"
+
+                elif event["type"] == "tool_call_result":
+                    yield f"data: {json.dumps({'type': 'tool_call_result', 'role_id': role_id, 'tool': event['tool'], 'result': event['result']}, ensure_ascii=False)}\n\n"
+                    tool_calls.append(event["result"])
+
+                elif event["type"] == "content_delta":
+                    yield f"data: {json.dumps({'type': 'content_delta', 'role_id': role_id, 'content': event['content']}, ensure_ascii=False)}\n\n"
+                    full_content += event["content"]
+
+                elif event["type"] == "content_done":
+                    full_content = event["full_content"]
+                    tool_calls = event.get("tool_calls", [])
+
+                elif event["type"] == "error":
+                    yield f"data: {json.dumps({'type': 'error', 'role_id': role_id, 'error': event['error']}, ensure_ascii=False)}\n\n"
+                    full_content = f"[{role['name']}暂时无法发言]"
+
+            # 构建完整消息
+            message = {
+                "role_id": role_id,
+                "role_name": role["name"],
+                "nickname": role["nickname"],
+                "avatar": role["avatar"],
+                "color": role["color"],
+                "content": full_content,
+                "tool_calls": tool_calls,
+                "timestamp": datetime.now().isoformat(),
+                "round_number": round_number
+            }
+            all_messages.append(message)
+
+            # 发送消息完成事件
+            yield f"data: {json.dumps({'type': 'message_complete', 'message': message}, ensure_ascii=False)}\n\n"
+
+            # 更新上下文
+            accumulated_context += f"\n\n【{role['name']}】：{full_content}"
+
+            await asyncio.sleep(0.3)
+
+        # 基金经理总结
+        fund_manager = MEETING_ROLES["fund_manager"]
+        yield f"data: {json.dumps({'type': 'speaking_start', 'role_id': 'fund_manager', 'role_name': fund_manager['name'], 'color': fund_manager['color']}, ensure_ascii=False)}\n\n"
+
+        fm_full_content = ""
+        fm_tool_calls = []
+        fm_tools = [t for t in TOOLS if t.name in fund_manager.get("tools", [])]
+
+        async for event in stream_role_response("fund_manager", request.topic, accumulated_context, fm_tools):
+            if event["type"] == "tool_call_start":
+                yield f"data: {json.dumps({'type': 'tool_call_start', 'role_id': 'fund_manager', 'tool': event['tool'], 'arguments': event['arguments']}, ensure_ascii=False)}\n\n"
+            elif event["type"] == "tool_call_result":
+                yield f"data: {json.dumps({'type': 'tool_call_result', 'role_id': 'fund_manager', 'tool': event['tool'], 'result': event['result']}, ensure_ascii=False)}\n\n"
+                fm_tool_calls.append(event["result"])
+            elif event["type"] == "content_delta":
+                yield f"data: {json.dumps({'type': 'content_delta', 'role_id': 'fund_manager', 'content': event['content']}, ensure_ascii=False)}\n\n"
+                fm_full_content += event["content"]
+            elif event["type"] == "content_done":
+                fm_full_content = event["full_content"]
+            elif event["type"] == "error":
+                fm_full_content = "[基金经理暂时无法发言]"
+
+        fm_message = {
+            "role_id": "fund_manager",
+            "role_name": fund_manager["name"],
+            "nickname": fund_manager["nickname"],
+            "avatar": fund_manager["avatar"],
+            "color": fund_manager["color"],
+            "content": fm_full_content,
+            "tool_calls": fm_tool_calls,
             "timestamp": datetime.now().isoformat(),
-            "round_number": round_number
+            "round_number": round_number,
+            "is_conclusion": True
         }
-        messages.append(message)
 
-    # 本轮结束后，基金经理再次判断
+        yield f"data: {json.dumps({'type': 'message_complete', 'message': fm_message}, ensure_ascii=False)}\n\n"
-    all_discussion = history_context + "\n本轮讨论：\n" + "\n".join([
-        f"【{msg['role_name']}】：{msg['content']}"
-        for msg in messages if msg["role_id"] != "user"
-    ])
 
-    can_conclude, conclusion_content = await check_conclusion_ready(all_discussion, request.topic)
+        # 发送会议状态（不强制结束，用户可以继续）
+        yield f"data: {json.dumps({'type': 'round_end', 'round_number': round_number, 'can_continue': True}, ensure_ascii=False)}\n\n"
 
-    # 添加基金经理的发言
+    return StreamingResponse(
-    fund_manager = MEETING_ROLES["fund_manager"]
+        generate_meeting_stream(),
-    fund_manager_message = {
+        media_type="text/event-stream",
-        "role_id": "fund_manager",
+        headers={
-        "role_name": fund_manager["name"],
+            "Cache-Control": "no-cache",
-        "nickname": fund_manager["nickname"],
+            "Connection": "keep-alive",
-        "avatar": fund_manager["avatar"],
+            "X-Accel-Buffering": "no",
-        "color": fund_manager["color"],
+        },
-        "content": conclusion_content,
+    )
-        "timestamp": datetime.now().isoformat(),
-        "round_number": round_number,
-        "is_conclusion": can_conclude
-    }
-    messages.append(fund_manager_message)
-
-    return {
-        "success": True,
-        "session_id": session_id,
-        "messages": messages,
-        "round_number": round_number,
-        "is_concluded": can_conclude,
-        "conclusion": fund_manager_message if can_conclude else None
-    }
 
 
 @app.get("/agent/meeting/roles")
@@ -2863,111 +2820,13 @@ async def get_meeting_roles():
                 "avatar": role["avatar"],
                 "color": role["color"],
                 "description": role["description"],
+                "tools": role.get("tools", []),
             }
             for role in MEETING_ROLES.values()
         ]
     }
 
 
-@app.post("/agent/meeting/stream")
-async def stream_investment_meeting(request: MeetingRequest):
-    """
-    流式投研会议
-
-    以 SSE 方式逐个角色流式返回发言
-    """
-    logger.info(f"流式投研会议: {request.topic} (user: {request.user_id})")
-
-    async def generate_meeting_stream() -> AsyncGenerator[str, None]:
-        session_id = request.session_id or str(uuid.uuid4())
-        round_number = 1
-        all_messages = []
-
-        # 发送会话开始事件
-        yield f"data: {json.dumps({'type': 'session_start', 'session_id': session_id}, ensure_ascii=False)}\n\n"
-
-        # 决定发言顺序
-        speaking_order = await determine_speaking_order(request.topic)
-
-        yield f"data: {json.dumps({'type': 'order_decided', 'order': speaking_order}, ensure_ascii=False)}\n\n"
-
-        context = f"议题：{request.topic}\n\n这是第一轮讨论，请针对议题发表你的观点。"
-
-        # 依次让每个角色发言
-        for role_id in speaking_order:
-            role = MEETING_ROLES[role_id]
-            if role["role_type"] == "manager":
-                continue
-
-            # 发送"正在发言"状态
-            yield f"data: {json.dumps({'type': 'speaking_start', 'role_id': role_id, 'role_name': role['name']}, ensure_ascii=False)}\n\n"
-
-            # 构建上下文
-            prev_context = context
-            if all_messages:
-                prev_context += "\n\n其他人的观点：\n"
-                for msg in all_messages:
-                    prev_context += f"- {msg['role_name']}：{msg['content']}\n"
-
-            # 调用LLM生成发言
-            content = await call_role_llm(role_id, request.topic, prev_context)
-
-            message = {
-                "role_id": role_id,
-                "role_name": role["name"],
-                "nickname": role["nickname"],
-                "avatar": role["avatar"],
-                "color": role["color"],
-                "content": content,
-                "timestamp": datetime.now().isoformat(),
-                "round_number": round_number
-            }
-            all_messages.append(message)
-
-            # 发送完整发言
-            yield f"data: {json.dumps({'type': 'message', 'message': message}, ensure_ascii=False)}\n\n"
-
-            # 短暂延迟，让前端有时间处理
-            await asyncio.sleep(0.5)
-
-        # 基金经理总结
-        fund_manager = MEETING_ROLES["fund_manager"]
-        yield f"data: {json.dumps({'type': 'speaking_start', 'role_id': 'fund_manager', 'role_name': fund_manager['name']}, ensure_ascii=False)}\n\n"
-
-        discussion_summary = "\n".join([
-            f"【{msg['role_name']}】：{msg['content']}"
-            for msg in all_messages
-        ])
-        can_conclude, conclusion_content = await check_conclusion_ready(discussion_summary, request.topic)
-
-        fund_manager_message = {
-            "role_id": "fund_manager",
-            "role_name": fund_manager["name"],
-            "nickname": fund_manager["nickname"],
-            "avatar": fund_manager["avatar"],
-            "color": fund_manager["color"],
-            "content": conclusion_content,
-            "timestamp": datetime.now().isoformat(),
-            "round_number": round_number,
-            "is_conclusion": can_conclude
-        }
-
-        yield f"data: {json.dumps({'type': 'message', 'message': fund_manager_message}, ensure_ascii=False)}\n\n"
-
-        # 发送会议结束事件
-        yield f"data: {json.dumps({'type': 'meeting_end', 'is_concluded': can_conclude, 'round_number': round_number}, ensure_ascii=False)}\n\n"
-
-    return StreamingResponse(
-        generate_meeting_stream(),
-        media_type="text/event-stream",
-        headers={
-            "Cache-Control": "no-cache",
-            "Connection": "keep-alive",
-            "X-Accel-Buffering": "no",
-        },
-    )
-
-
 # ==================== 健康检查 ====================
 
 @app.get("/health")

src/mocks/handlers/agent.js

@@ -434,4 +434,157 @@ export const agentHandlers = [
       conclusion: messages[messages.length - 1],
     });
   }),
+
+  // POST /mcp/agent/meeting/stream - 流式会议接口（V2）
+  http.post('/mcp/agent/meeting/stream', async ({ request }) => {
+    const body = await request.json();
+    const { topic, user_id } = body;
+
+    const sessionId = `meeting-${Date.now()}`;
+
+    // 定义会议角色和他们的消息
+    const roleMessages = [
+      {
+        role_id: 'buffett',
+        role_name: '巴菲特',
+        content: `关于「${topic}」，我认为这里存在显著的投资机会。从价值投资的角度看，我们应该关注以下几点：\n\n1. **长期价值**：该标的具有较强的护城河\n2. **盈利能力**：ROE持续保持在较高水平\n3. **管理层质量**：管理团队稳定且执行力强\n\n我的观点是**看多**，建议逢低布局。`,
+        tools: [
+          { name: 'search_china_news', result: { articles: [{ title: '相关新闻1' }, { title: '相关新闻2' }] } },
+          { name: 'get_stock_basic_info', result: { pe: 25.6, pb: 3.2, roe: 18.5 } },
+        ],
+      },
+      {
+        role_id: 'big_short',
+        role_name: '大空头',
+        content: `等等，让我泼点冷水。关于「${topic}」，市场似乎过于乐观了：\n\n⚠️ **风险提示**：\n1. 当前估值处于历史高位，安全边际不足\n2. 行业竞争加剧，利润率面临压力\n3. 宏观环境不确定性增加\n\n建议投资者**保持谨慎**，不要追高。`,
+        tools: [
+          { name: 'get_stock_financial_index', result: { debt_ratio: 45.2, current_ratio: 1.8 } },
+        ],
+      },
+      {
+        role_id: 'simons',
+        role_name: '量化分析员',
+        content: `从量化角度分析「${topic}」：\n\n📊 **技术指标**：\n- MACD：金叉形态，动能向上\n- RSI：58，处于中性区域\n- 均线：5日>10日>20日，多头排列\n\n📈 **资金面**：\n- 主力资金：近5日净流入2.3亿\n- 北向资金：持续加仓\n\n**结论**：短期技术面偏多，但需关注60日均线支撑。`,
+        tools: [
+          { name: 'get_stock_trade_data', result: { volume: 1234567, turnover: 5.2 } },
+          { name: 'get_concept_statistics', result: { concepts: ['AI概念', '半导体'], avg_change: 2.3 } },
+        ],
+      },
+      {
+        role_id: 'leek',
+        role_name: '韭菜',
+        content: `哇！「${topic}」看起来要涨啊！\n\n🚀 我觉得必须满仓干！隔壁老王都赚翻了！\n\n不过话说回来...万一跌了怎么办？会不会套住？\n\n算了不管了，先冲一把再说！错过这村就没这店了！\n\n（内心OS：希望别当接盘侠...）`,
+        tools: [], // 韭菜不用工具
+      },
+      {
+        role_id: 'fund_manager',
+        role_name: '基金经理',
+        content: `## 投资建议总结\n\n综合各方观点，对于「${topic}」，我的判断如下：\n\n### 综合评估\n多空双方都提出了有价值的观点。技术面短期偏多，但估值确实需要关注。\n\n### 关键观点\n- ✅ 基本面优质，长期价值明确\n- ⚠️ 短期估值偏高，需要耐心等待\n- 📊 技术面处于上升趋势\n\n### 风险提示\n注意仓位控制，避免追高\n\n### 操作建议\n**观望为主**，等待回调至支撑位再考虑建仓\n\n### 信心指数：7/10`,
+        tools: [
+          { name: 'search_research_reports', result: { reports: [{ title: '深度研报1' }] } },
+        ],
+        is_conclusion: true,
+      },
+    ];
+
+    // 创建 SSE 流
+    const encoder = new TextEncoder();
+    const stream = new ReadableStream({
+      async start(controller) {
+        // 发送 session_start
+        controller.enqueue(encoder.encode(`data: ${JSON.stringify({
+          type: 'session_start',
+          session_id: sessionId,
+        })}\n\n`));
+
+        await delay(300);
+
+        // 发送 order_decided
+        controller.enqueue(encoder.encode(`data: ${JSON.stringify({
+          type: 'order_decided',
+          order: roleMessages.map(r => r.role_id),
+        })}\n\n`));
+
+        await delay(300);
+
+        // 依次发送每个角色的消息
+        for (const role of roleMessages) {
+          // speaking_start
+          controller.enqueue(encoder.encode(`data: ${JSON.stringify({
+            type: 'speaking_start',
+            role_id: role.role_id,
+            role_name: role.role_name,
+          })}\n\n`));
+
+          await delay(200);
+
+          // 发送工具调用
+          for (const tool of role.tools) {
+            const toolCallId = `tc-${Date.now()}-${Math.random().toString(36).slice(2, 8)}`;
+
+            // tool_call_start
+            controller.enqueue(encoder.encode(`data: ${JSON.stringify({
+              type: 'tool_call_start',
+              role_id: role.role_id,
+              tool_call_id: toolCallId,
+              tool_name: tool.name,
+              arguments: {},
+            })}\n\n`));
+
+            await delay(500);
+
+            // tool_call_result
+            controller.enqueue(encoder.encode(`data: ${JSON.stringify({
+              type: 'tool_call_result',
+              role_id: role.role_id,
+              tool_call_id: toolCallId,
+              result: tool.result,
+              status: 'success',
+              execution_time: 0.5 + Math.random() * 0.5,
+            })}\n\n`));
+
+            await delay(200);
+          }
+
+          // 流式发送内容
+          const chunks = role.content.match(/.{1,20}/g) || [];
+          for (const chunk of chunks) {
+            controller.enqueue(encoder.encode(`data: ${JSON.stringify({
+              type: 'content_delta',
+              role_id: role.role_id,
+              content: chunk,
+            })}\n\n`));
+            await delay(30);
+          }
+
+          // message_complete
+          controller.enqueue(encoder.encode(`data: ${JSON.stringify({
+            type: 'message_complete',
+            role_id: role.role_id,
+            content: role.content,
+            is_conclusion: role.is_conclusion || false,
+          })}\n\n`));
+
+          await delay(500);
+        }
+
+        // round_end
+        controller.enqueue(encoder.encode(`data: ${JSON.stringify({
+          type: 'round_end',
+          round_number: 1,
+          is_concluded: false,
+        })}\n\n`));
+
+        controller.close();
+      },
+    });
+
+    return new Response(stream, {
+      headers: {
+        'Content-Type': 'text/event-stream',
+        'Cache-Control': 'no-cache',
+        'Connection': 'keep-alive',
+      },
+    });
+  }),
 ];

src/views/AgentChat/components/MeetingRoom/MeetingMessageBubble.js

@@ -1,8 +1,8 @@
 // src/views/AgentChat/components/MeetingRoom/MeetingMessageBubble.js
-// 会议消息气泡组件
+// 会议消息气泡组件 - V2: 支持工具调用展示和流式输出
 
-import React from 'react';
+import React, { useState } from 'react';
-import { motion } from 'framer-motion';
+import { motion, AnimatePresence } from 'framer-motion';
 import {
   Box,
   Flex,
@@ -15,6 +15,9 @@ import {
   Tooltip,
   Card,
   CardBody,
+  Spinner,
+  Code,
+  Collapse,
 } from '@chakra-ui/react';
 import {
   TrendingUp,
@@ -24,6 +27,11 @@ import {
   Crown,
   Copy,
   ThumbsUp,
+  ChevronRight,
+  Database,
+  Check,
+  Wrench,
+  AlertCircle,
 } from 'lucide-react';
 import { getRoleConfig, MEETING_ROLES } from '../../constants/meetingRoles';
 import { MarkdownWithCharts } from '@components/ChatBot/MarkdownWithCharts';
@@ -48,6 +56,231 @@ const getRoleIcon = (roleType) => {
   }
 };
 
+/**
+ * 工具名称映射
+ */
+const TOOL_NAME_MAP = {
+  search_china_news: '搜索新闻',
+  search_research_reports: '搜索研报',
+  get_stock_basic_info: '获取股票信息',
+  get_stock_financial_index: '获取财务指标',
+  get_stock_balance_sheet: '获取资产负债表',
+  get_stock_cashflow: '获取现金流量表',
+  get_stock_trade_data: '获取交易数据',
+  search_limit_up_stocks: '搜索涨停股',
+  get_concept_statistics: '获取概念统计',
+};
+
+/**
+ * 格式化结果数据用于显示
+ */
+const formatResultData = (data) => {
+  if (data === null || data === undefined) return null;
+  if (typeof data === 'string') return data;
+  try {
+    return JSON.stringify(data, null, 2);
+  } catch {
+    return String(data);
+  }
+};
+
+/**
+ * 获取结果数据的预览文本
+ */
+const getResultPreview = (result) => {
+  if (!result) return '无数据';
+
+  if (result.data) {
+    const data = result.data;
+    if (data.chart_data) {
+      return `图表数据: ${data.chart_data.labels?.length || 0} 项`;
+    }
+    if (data.sector_data) {
+      const sectorCount = Object.keys(data.sector_data).length;
+      return `${sectorCount} 个板块分析`;
+    }
+    if (data.stocks) {
+      return `${data.stocks.length} 只股票`;
+    }
+    if (Array.isArray(data)) {
+      return `${data.length} 条记录`;
+    }
+  }
+
+  if (Array.isArray(result)) {
+    return `${result.length} 条记录`;
+  }
+
+  if (typeof result === 'object') {
+    const keys = Object.keys(result);
+    return `${keys.length} 个字段`;
+  }
+
+  return '查看详情';
+};
+
+/**
+ * 单个工具调用卡片
+ */
+const ToolCallCard = ({ toolCall, idx, roleColor }) => {
+  const [isExpanded, setIsExpanded] = useState(false);
+  const [copied, setCopied] = useState(false);
+
+  const hasResult = toolCall.result && (
+    typeof toolCall.result === 'object'
+      ? Object.keys(toolCall.result).length > 0
+      : toolCall.result
+  );
+
+  const handleCopy = async (e) => {
+    e.stopPropagation();
+    try {
+      await navigator.clipboard.writeText(formatResultData(toolCall.result));
+      setCopied(true);
+      setTimeout(() => setCopied(false), 2000);
+    } catch (err) {
+      console.error('复制失败:', err);
+    }
+  };
+
+  const toolDisplayName = TOOL_NAME_MAP[toolCall.tool_name] || toolCall.tool_name;
+
+  return (
+    <motion.div
+      initial={{ opacity: 0, x: -10 }}
+      animate={{ opacity: 1, x: 0 }}
+      transition={{ delay: idx * 0.05 }}
+    >
+      <Card
+        bg="rgba(255, 255, 255, 0.03)"
+        border="1px solid"
+        borderColor={isExpanded ? `${roleColor}40` : 'rgba(255, 255, 255, 0.1)'}
+        borderRadius="md"
+        transition="all 0.2s"
+        _hover={{
+          borderColor: `${roleColor}30`,
+        }}
+        size="sm"
+      >
+        <CardBody p={2}>
+          {/* 工具调用头部 */}
+          <Flex
+            align="center"
+            justify="space-between"
+            gap={2}
+            cursor={hasResult ? 'pointer' : 'default'}
+            onClick={() => hasResult && setIsExpanded(!isExpanded)}
+          >
+            <HStack flex={1} spacing={2}>
+              {toolCall.status === 'calling' ? (
+                <Spinner size="xs" color={roleColor} />
+              ) : toolCall.status === 'success' ? (
+                <Box color="green.400">
+                  <Check className="w-3 h-3" />
+                </Box>
+              ) : (
+                <Box color="red.400">
+                  <AlertCircle className="w-3 h-3" />
+                </Box>
+              )}
+              <Wrench className="w-3 h-3" style={{ color: roleColor }} />
+              <Text fontSize="xs" fontWeight="medium" color="gray.300">
+                {toolDisplayName}
+              </Text>
+              {hasResult && (
+                <Box
+                  color="gray.500"
+                  transition="transform 0.2s"
+                  transform={isExpanded ? 'rotate(90deg)' : 'rotate(0deg)'}
+                >
+                  <ChevronRight className="w-3 h-3" />
+                </Box>
+              )}
+            </HStack>
+
+            <HStack spacing={2}>
+              {hasResult && (
+                <Tooltip label={copied ? '已复制' : '复制数据'} placement="top">
+                  <IconButton
+                    size="xs"
+                    variant="ghost"
+                    icon={copied ? <Check className="w-2 h-2" /> : <Copy className="w-2 h-2" />}
+                    onClick={handleCopy}
+                    color={copied ? 'green.400' : 'gray.500'}
+                    _hover={{ bg: 'rgba(255, 255, 255, 0.1)' }}
+                    aria-label="复制"
+                    minW="20px"
+                    h="20px"
+                  />
+                </Tooltip>
+              )}
+              {toolCall.execution_time && (
+                <Text fontSize="10px" color="gray.500">
+                  {toolCall.execution_time.toFixed(2)}s
+                </Text>
+              )}
+            </HStack>
+          </Flex>
+
+          {/* 展开的详细数据 */}
+          <Collapse in={isExpanded} animateOpacity>
+            {isExpanded && hasResult && (
+              <Box mt={2}>
+                <Code
+                  display="block"
+                  p={2}
+                  borderRadius="sm"
+                  fontSize="10px"
+                  whiteSpace="pre-wrap"
+                  bg="rgba(0, 0, 0, 0.3)"
+                  color="gray.300"
+                  maxH="200px"
+                  overflowY="auto"
+                  sx={{
+                    '&::-webkit-scrollbar': { width: '4px' },
+                    '&::-webkit-scrollbar-track': { bg: 'transparent' },
+                    '&::-webkit-scrollbar-thumb': { bg: 'gray.600', borderRadius: 'full' },
+                  }}
+                >
+                  {formatResultData(toolCall.result)}
+                </Code>
+              </Box>
+            )}
+          </Collapse>
+        </CardBody>
+      </Card>
+    </motion.div>
+  );
+};
+
+/**
+ * 工具调用列表组件
+ */
+const ToolCallsList = ({ toolCalls, roleColor }) => {
+  if (!toolCalls || toolCalls.length === 0) return null;
+
+  return (
+    <Box mt={3} mb={2}>
+      <HStack spacing={2} mb={2}>
+        <Wrench className="w-3 h-3" style={{ color: roleColor }} />
+        <Text fontSize="xs" color="gray.400">
+          工具调用 ({toolCalls.length})
+        </Text>
+      </HStack>
+      <VStack spacing={1} align="stretch">
+        {toolCalls.map((toolCall, idx) => (
+          <ToolCallCard
+            key={toolCall.tool_call_id || idx}
+            toolCall={toolCall}
+            idx={idx}
+            roleColor={roleColor}
+          />
+        ))}
+      </VStack>
+    </Box>
+  );
+};
+
 /**
  * MeetingMessageBubble - 会议消息气泡组件
  *
@@ -67,6 +300,8 @@ const MeetingMessageBubble = ({ message, isLatest }) => {
   const isUser = message.role_id === 'user';
   const isManager = roleConfig.roleType === 'manager';
   const isConclusion = message.is_conclusion;
+  const isStreaming = message.isStreaming;
+  const hasToolCalls = message.tool_calls && message.tool_calls.length > 0;
 
   // 复制到剪贴板
   const handleCopy = () => {
@@ -120,6 +355,19 @@ const MeetingMessageBubble = ({ message, isLatest }) => {
                 主持人
               </Badge>
             )}
+            {isStreaming && (
+              <Badge
+                colorScheme="blue"
+                size="sm"
+                variant="subtle"
+                display="flex"
+                alignItems="center"
+                gap={1}
+              >
+                <Spinner size="xs" />
+                发言中
+              </Badge>
+            )}
             {isConclusion && (
               <Badge
                 colorScheme="green"
@@ -188,6 +436,15 @@ const MeetingMessageBubble = ({ message, isLatest }) => {
           )}
 
           <CardBody p={4}>
+            {/* 工具调用列表 */}
+            {hasToolCalls && (
+              <ToolCallsList
+                toolCalls={message.tool_calls}
+                roleColor={roleConfig.color}
+              />
+            )}
+
+            {/* 消息内容 */}
             <Box
               fontSize="sm"
               color="gray.100"
@@ -212,7 +469,25 @@ const MeetingMessageBubble = ({ message, isLatest }) => {
                 '& strong': { color: roleConfig.color },
               }}
             >
-              <MarkdownWithCharts content={message.content} variant="dark" />
+              {message.content ? (
+                <MarkdownWithCharts content={message.content} variant="dark" />
+              ) : isStreaming ? (
+                <HStack spacing={2} color="gray.500">
+                  <Spinner size="sm" />
+                  <Text>正在思考...</Text>
+                </HStack>
+              ) : null}
+
+              {/* 流式输出时的光标 */}
+              {isStreaming && message.content && (
+                <motion.span
+                  animate={{ opacity: [1, 0, 1] }}
+                  transition={{ duration: 0.8, repeat: Infinity }}
+                  style={{ color: roleConfig.color }}
+                >
+                  ▌
+                </motion.span>
+              )}
             </Box>
 
             {/* 操作按钮 */}

src/views/AgentChat/components/MeetingRoom/index.js

@@ -91,18 +91,18 @@ const MeetingRoom = ({ user, onToast }) => {
       // 启动新会议
       startMeeting(inputValue.trim());
       setInputValue('');
+    } else if (status === MeetingStatus.CONCLUDED) {
+      // 如果已结论，开始新会议
+      resetMeeting();
+      startMeeting(inputValue.trim());
+      setInputValue('');
     } else if (
       status === MeetingStatus.WAITING_INPUT ||
-      status === MeetingStatus.CONCLUDED
+      status === MeetingStatus.DISCUSSING ||
+      status === MeetingStatus.SPEAKING
     ) {
-      // 用户插话或开始新话题
+      // 用户可以在任何时候插话（包括讨论中和发言中）
-      if (isConcluded) {
+      sendUserMessage(inputValue.trim());
-        // 如果已结论，开始新会议
-        resetMeeting();
-        startMeeting(inputValue.trim());
-      } else {
-        sendUserMessage(inputValue.trim());
-      }
       setInputValue('');
     }
   };
@@ -135,11 +135,15 @@ const MeetingRoom = ({ user, onToast }) => {
     if (status === MeetingStatus.IDLE) {
       return '输入投研议题，如：分析茅台最新财报...';
     } else if (status === MeetingStatus.WAITING_INPUT) {
-      return '输入您的观点参与讨论，或等待继续...';
+      return '输入您的观点参与讨论，或点击继续按钮...';
     } else if (status === MeetingStatus.CONCLUDED) {
       return '会议已结束，输入新议题开始新会议...';
+    } else if (status === MeetingStatus.STARTING) {
+      return '正在召集会议成员...';
+    } else if (status === MeetingStatus.DISCUSSING || status === MeetingStatus.SPEAKING) {
+      return '随时输入您的观点参与讨论...';
     }
-    return '会议进行中...';
+    return '输入您的观点...';
   };
 
   return (
@@ -352,11 +356,7 @@ const MeetingRoom = ({ user, onToast }) => {
               onChange={(e) => setInputValue(e.target.value)}
               onKeyDown={handleKeyPress}
               placeholder={getPlaceholder()}
-              isDisabled={
+              isDisabled={status === MeetingStatus.STARTING}
-                isLoading ||
-                status === MeetingStatus.DISCUSSING ||
-                status === MeetingStatus.SPEAKING
-              }
               size="lg"
               bg="rgba(255, 255, 255, 0.05)"
               border="1px solid"
@@ -378,13 +378,11 @@ const MeetingRoom = ({ user, onToast }) => {
             >
               <IconButton
                 size="lg"
-                icon={isLoading ? <Spinner size="sm" /> : <Send className="w-5 h-5" />}
+                icon={isLoading && status === MeetingStatus.STARTING ? <Spinner size="sm" /> : <Send className="w-5 h-5" />}
                 onClick={handleSend}
                 isDisabled={
                   !inputValue.trim() ||
-                  isLoading ||
+                  status === MeetingStatus.STARTING
-                  status === MeetingStatus.DISCUSSING ||
-                  status === MeetingStatus.SPEAKING
                 }
                 bgGradient="linear(to-r, orange.400, red.500)"
                 color="white"
@@ -427,9 +425,11 @@ const MeetingRoom = ({ user, onToast }) => {
                 {status === MeetingStatus.IDLE ? '开始会议' : '发送消息'}
               </Text>
             </HStack>
-            {status === MeetingStatus.WAITING_INPUT && (
+            {(status === MeetingStatus.WAITING_INPUT ||
+              status === MeetingStatus.DISCUSSING ||
+              status === MeetingStatus.SPEAKING) && (
               <Text color="orange.400">
-                💡 您可以插话参与讨论，或点击继续按钮进行下一轮
+                💡 随时输入观点参与讨论，您的发言会影响分析师的判断
               </Text>
             )}
           </HStack>

src/views/AgentChat/constants/meetingRoles.ts

@@ -39,6 +39,26 @@ export interface MeetingRoleConfig {
   icon: React.ReactNode;
 }
 
+/**
+ * 工具调用结果接口
+ */
+export interface ToolCallResult {
+  /** 工具调用 ID */
+  tool_call_id: string;
+  /** 工具名称 */
+  tool_name: string;
+  /** 工具参数 */
+  arguments?: Record<string, any>;
+  /** 调用状态 */
+  status: 'calling' | 'success' | 'error';
+  /** 调用结果 */
+  result?: any;
+  /** 错误信息 */
+  error?: string;
+  /** 执行时间（秒） */
+  execution_time?: number;
+}
+
 /**
  * 会议消息接口
  */
@@ -63,6 +83,10 @@ export interface MeetingMessage {
   round_number: number;
   /** 是否为结论 */
   is_conclusion?: boolean;
+  /** 工具调用列表 */
+  tool_calls?: ToolCallResult[];
+  /** 是否正在流式输出 */
+  isStreaming?: boolean;
 }
 
 /**
@@ -209,8 +233,12 @@ export type MeetingEventType =
   | 'session_start'
   | 'order_decided'
   | 'speaking_start'
-  | 'message'
+  | 'tool_call_start'
-  | 'meeting_end';
+  | 'tool_call_result'
+  | 'content_delta'
+  | 'message_complete'
+  | 'round_end'
+  | 'error';
 
 /**
  * SSE 事件接口
@@ -224,4 +252,15 @@ export interface MeetingEvent {
   message?: MeetingMessage;
   is_concluded?: boolean;
   round_number?: number;
+  /** 工具调用相关 */
+  tool_call_id?: string;
+  tool_name?: string;
+  arguments?: Record<string, any>;
+  result?: any;
+  status?: string;
+  execution_time?: number;
+  /** 流式内容 */
+  content?: string;
+  /** 错误信息 */
+  error?: string;
 }

src/views/AgentChat/hooks/useInvestmentMeeting.ts

@@ -1,5 +1,6 @@
 // src/views/AgentChat/hooks/useInvestmentMeeting.ts
 // 投研会议室 Hook - 管理会议状态、发送消息、处理 SSE 流
+// V2: 支持流式输出、工具调用展示、用户中途发言
 
 import { useState, useCallback, useRef } from 'react';
 import axios from 'axios';
@@ -8,6 +9,7 @@ import {
   MeetingStatus,
   MeetingEvent,
   MeetingResponse,
+  ToolCallResult,
   getRoleConfig,
 } from '../constants/meetingRoles';
 
@@ -129,7 +131,129 @@ export const useInvestmentMeeting = ({
   }, []);
 
   /**
-   * 启动会议（使用流式 SSE）
+   * 更新消息内容（用于流式输出）
+   */
+  const updateMessageContent = useCallback((roleId: string, content: string) => {
+    setMessages((prev) => {
+      const lastIndex = prev.findIndex(
+        (m) => m.role_id === roleId && m.isStreaming
+      );
+      if (lastIndex >= 0) {
+        const newMessages = [...prev];
+        newMessages[lastIndex] = {
+          ...newMessages[lastIndex],
+          content: newMessages[lastIndex].content + content,
+        };
+        return newMessages;
+      }
+      return prev;
+    });
+  }, []);
+
+  /**
+   * 添加工具调用到消息
+   */
+  const addToolCallToMessage = useCallback(
+    (roleId: string, toolCall: ToolCallResult) => {
+      setMessages((prev) => {
+        const lastIndex = prev.findIndex(
+          (m) => m.role_id === roleId && m.isStreaming
+        );
+        if (lastIndex >= 0) {
+          const newMessages = [...prev];
+          const existingToolCalls = newMessages[lastIndex].tool_calls || [];
+          newMessages[lastIndex] = {
+            ...newMessages[lastIndex],
+            tool_calls: [...existingToolCalls, toolCall],
+          };
+          return newMessages;
+        }
+        return prev;
+      });
+    },
+    []
+  );
+
+  /**
+   * 更新工具调用结果
+   */
+  const updateToolCallResult = useCallback(
+    (roleId: string, toolCallId: string, result: any, status: string, executionTime?: number) => {
+      setMessages((prev) => {
+        const lastIndex = prev.findIndex(
+          (m) => m.role_id === roleId && m.isStreaming
+        );
+        if (lastIndex >= 0) {
+          const newMessages = [...prev];
+          const toolCalls = newMessages[lastIndex].tool_calls || [];
+          const toolIndex = toolCalls.findIndex((t) => t.tool_call_id === toolCallId);
+          if (toolIndex >= 0) {
+            const newToolCalls = [...toolCalls];
+            newToolCalls[toolIndex] = {
+              ...newToolCalls[toolIndex],
+              result,
+              status: status as 'success' | 'error',
+              execution_time: executionTime,
+            };
+            newMessages[lastIndex] = {
+              ...newMessages[lastIndex],
+              tool_calls: newToolCalls,
+            };
+          }
+          return newMessages;
+        }
+        return prev;
+      });
+    },
+    []
+  );
+
+  /**
+   * 完成消息流式输出
+   */
+  const finishStreamingMessage = useCallback((roleId: string, finalContent?: string) => {
+    setMessages((prev) => {
+      const lastIndex = prev.findIndex(
+        (m) => m.role_id === roleId && m.isStreaming
+      );
+      if (lastIndex >= 0) {
+        const newMessages = [...prev];
+        newMessages[lastIndex] = {
+          ...newMessages[lastIndex],
+          content: finalContent || newMessages[lastIndex].content,
+          isStreaming: false,
+        };
+        return newMessages;
+      }
+      return prev;
+    });
+  }, []);
+
+  /**
+   * 创建流式消息占位
+   */
+  const createStreamingMessage = useCallback(
+    (roleId: string, roleName: string, roundNumber: number): MeetingMessage => {
+      const roleConfig = getRoleConfig(roleId);
+      return {
+        id: `${roleId}-${Date.now()}`,
+        role_id: roleId,
+        role_name: roleName,
+        nickname: roleConfig?.nickname || roleName,
+        avatar: roleConfig?.avatar || '',
+        color: roleConfig?.color || '#6366F1',
+        content: '',
+        timestamp: new Date().toISOString(),
+        round_number: roundNumber,
+        tool_calls: [],
+        isStreaming: true,
+      };
+    },
+    []
+  );
+
+  /**
+   * 启动会议（使用 POST + fetch 流式 SSE）
    */
   const startMeetingStream = useCallback(
     async (topic: string) => {
@@ -137,156 +261,189 @@ export const useInvestmentMeeting = ({
       setStatus(MeetingStatus.STARTING);
       setIsLoading(true);
       setMessages([]);
+      setCurrentRound(1);
 
       try {
-        // 使用 EventSource 进行 SSE 连接
+        // 使用 fetch 进行 POST 请求的 SSE
-        const params = new URLSearchParams({
+        const response = await fetch('/mcp/agent/meeting/stream', {
-          topic,
+          method: 'POST',
-          user_id: userId,
+          headers: {
-          user_nickname: userNickname,
+            'Content-Type': 'application/json',
-        });
+          },
-
+          body: JSON.stringify({
-        const eventSource = new EventSource(
-          `/mcp/agent/meeting/stream?${params.toString()}`
-        );
-        eventSourceRef.current = eventSource;
-
-        eventSource.onmessage = (event) => {
-          try {
-            const data: MeetingEvent = JSON.parse(event.data);
-
-            switch (data.type) {
-              case 'session_start':
-                setSessionId(data.session_id || null);
-                setStatus(MeetingStatus.DISCUSSING);
-                break;
-
-              case 'order_decided':
-                // 发言顺序已决定
-                break;
-
-              case 'speaking_start':
-                setSpeakingRoleId(data.role_id || null);
-                setStatus(MeetingStatus.SPEAKING);
-                break;
-
-              case 'message':
-                if (data.message) {
-                  addMessage(data.message);
-                  setSpeakingRoleId(null);
-
-                  // 检查是否是结论
-                  if (data.message.is_conclusion) {
-                    setConclusion(data.message);
-                    setIsConcluded(true);
-                  }
-                }
-                break;
-
-              case 'meeting_end':
-                setCurrentRound(data.round_number || 1);
-                setIsConcluded(data.is_concluded || false);
-                setStatus(
-                  data.is_concluded
-                    ? MeetingStatus.CONCLUDED
-                    : MeetingStatus.WAITING_INPUT
-                );
-                setIsLoading(false);
-                eventSource.close();
-                break;
-            }
-          } catch (e) {
-            console.error('解析 SSE 事件失败:', e);
-          }
-        };
-
-        eventSource.onerror = (error) => {
-          console.error('SSE 连接错误:', error);
-          eventSource.close();
-          setStatus(MeetingStatus.ERROR);
-          setIsLoading(false);
-          onToast?.({
-            title: '连接失败',
-            description: '会议连接中断，请重试',
-            status: 'error',
-          });
-        };
-      } catch (error) {
-        console.error('启动会议失败:', error);
-        setStatus(MeetingStatus.ERROR);
-        setIsLoading(false);
-        onToast?.({
-          title: '启动会议失败',
-          description: '请稍后重试',
-          status: 'error',
-        });
-      }
-    },
-    [userId, userNickname, addMessage, onToast]
-  );
-
-  /**
-   * 启动会议（非流式，获取完整响应）
-   */
-  const startMeeting = useCallback(
-    async (topic: string) => {
-      setCurrentTopic(topic);
-      setStatus(MeetingStatus.STARTING);
-      setIsLoading(true);
-      setMessages([]);
-
-      try {
-        const response = await axios.post<MeetingResponse>(
-          '/mcp/agent/meeting/start',
-          {
             topic,
             user_id: userId,
             user_nickname: userNickname,
+            conversation_history: [],
+          }),
+        });
+
+        if (!response.ok) {
+          throw new Error(`HTTP error! status: ${response.status}`);
+        }
+
+        const reader = response.body?.getReader();
+        if (!reader) {
+          throw new Error('无法获取响应流');
+        }
+
+        const decoder = new TextDecoder();
+        let buffer = '';
+
+        const processLine = (line: string) => {
+          if (line.startsWith('data: ')) {
+            try {
+              const data: MeetingEvent = JSON.parse(line.slice(6));
+              handleSSEEvent(data, 1);
+            } catch (e) {
+              console.error('解析 SSE 数据失败:', e, line);
+            }
           }
-        );
+        };
 
-        if (response.data.success) {
+        const handleSSEEvent = (data: MeetingEvent, roundNum: number) => {
-          const data = response.data;
+          switch (data.type) {
+            case 'session_start':
+              setSessionId(data.session_id || null);
+              setStatus(MeetingStatus.DISCUSSING);
+              break;
 
-          setSessionId(data.session_id);
+            case 'order_decided':
-          setCurrentRound(data.round_number);
+              // 发言顺序已决定，可以显示提示
-          setIsConcluded(data.is_concluded);
+              break;
 
-          // 添加所有消息
+            case 'speaking_start':
-          data.messages.forEach((msg) => {
+              setSpeakingRoleId(data.role_id || null);
-            addMessage(msg);
+              setStatus(MeetingStatus.SPEAKING);
-          });
+              // 创建流式消息占位
+              if (data.role_id && data.role_name) {
+                const streamingMsg = createStreamingMessage(
+                  data.role_id,
+                  data.role_name,
+                  roundNum
+                );
+                addMessage(streamingMsg);
+              }
+              break;
 
-          // 设置结论
+            case 'tool_call_start':
-          if (data.conclusion) {
+              if (data.role_id && data.tool_call_id && data.tool_name) {
-            setConclusion(data.conclusion);
+                const toolCall: ToolCallResult = {
+                  tool_call_id: data.tool_call_id,
+                  tool_name: data.tool_name,
+                  arguments: data.arguments,
+                  status: 'calling',
+                };
+                addToolCallToMessage(data.role_id, toolCall);
+              }
+              break;
+
+            case 'tool_call_result':
+              if (data.role_id && data.tool_call_id) {
+                updateToolCallResult(
+                  data.role_id,
+                  data.tool_call_id,
+                  data.result,
+                  data.status || 'success',
+                  data.execution_time
+                );
+              }
+              break;
+
+            case 'content_delta':
+              if (data.role_id && data.content) {
+                updateMessageContent(data.role_id, data.content);
+              }
+              break;
+
+            case 'message_complete':
+              if (data.role_id) {
+                finishStreamingMessage(data.role_id, data.content);
+                setSpeakingRoleId(null);
+              }
+              break;
+
+            case 'round_end':
+              setCurrentRound(data.round_number || 1);
+              setIsConcluded(data.is_concluded || false);
+              setStatus(
+                data.is_concluded
+                  ? MeetingStatus.CONCLUDED
+                  : MeetingStatus.WAITING_INPUT
+              );
+              setIsLoading(false);
+              break;
+
+            case 'error':
+              console.error('会议错误:', data.error);
+              setStatus(MeetingStatus.ERROR);
+              setIsLoading(false);
+              onToast?.({
+                title: '会议出错',
+                description: data.error || '未知错误',
+                status: 'error',
+              });
+              break;
           }
+        };
 
-          setStatus(
+        // 读取流
-            data.is_concluded
+        while (true) {
-              ? MeetingStatus.CONCLUDED
+          const { done, value } = await reader.read();
-              : MeetingStatus.WAITING_INPUT
+          if (done) break;
-          );
+
-        } else {
+          buffer += decoder.decode(value, { stream: true });
-          throw new Error('会议启动失败');
+          const lines = buffer.split('\n');
+          buffer = lines.pop() || '';
+
+          for (const line of lines) {
+            if (line.trim()) {
+              processLine(line);
+            }
+          }
+        }
+
+        // 处理剩余 buffer
+        if (buffer.trim()) {
+          processLine(buffer);
         }
       } catch (error: any) {
         console.error('启动会议失败:', error);
         setStatus(MeetingStatus.ERROR);
+        setIsLoading(false);
         onToast?.({
           title: '启动会议失败',
-          description: error.response?.data?.detail || error.message,
+          description: error.message || '请稍后重试',
           status: 'error',
         });
-      } finally {
-        setIsLoading(false);
       }
     },
-    [userId, userNickname, addMessage, onToast]
+    [
+      userId,
+      userNickname,
+      addMessage,
+      createStreamingMessage,
+      addToolCallToMessage,
+      updateToolCallResult,
+      updateMessageContent,
+      finishStreamingMessage,
+      onToast,
+    ]
   );
 
   /**
-   * 继续会议讨论
+   * 启动会议（默认使用流式）
+   */
+  const startMeeting = useCallback(
+    async (topic: string) => {
+      // 使用流式版本
+      await startMeetingStream(topic);
+    },
+    [startMeetingStream]
+  );
+
+  /**
+   * 继续会议讨论（使用流式）
    */
   const continueMeeting = useCallback(
     async (userMessage?: string) => {
@@ -301,55 +458,184 @@ export const useInvestmentMeeting = ({
 
       setStatus(MeetingStatus.DISCUSSING);
       setIsLoading(true);
+      const nextRound = currentRound + 1;
+      setCurrentRound(nextRound);
 
       try {
-        const response = await axios.post<MeetingResponse>(
+        // 构建会话历史（排除正在流式传输的消息）
-          '/mcp/agent/meeting/continue',
+        const historyMessages = messages
-          {
+          .filter((m) => !m.isStreaming)
+          .map((m) => ({
+            role_id: m.role_id,
+            role_name: m.role_name,
+            content: m.content,
+          }));
+
+        // 使用 fetch 进行 POST 请求的 SSE
+        const response = await fetch('/mcp/agent/meeting/stream', {
+          method: 'POST',
+          headers: {
+            'Content-Type': 'application/json',
+          },
+          body: JSON.stringify({
             topic: currentTopic,
             user_id: userId,
             user_nickname: userNickname,
             session_id: sessionId,
             user_message: userMessage,
-            conversation_history: messages,
+            conversation_history: historyMessages,
+          }),
+        });
+
+        if (!response.ok) {
+          throw new Error(`HTTP error! status: ${response.status}`);
+        }
+
+        const reader = response.body?.getReader();
+        if (!reader) {
+          throw new Error('无法获取响应流');
+        }
+
+        const decoder = new TextDecoder();
+        let buffer = '';
+
+        const processLine = (line: string) => {
+          if (line.startsWith('data: ')) {
+            try {
+              const data: MeetingEvent = JSON.parse(line.slice(6));
+              handleSSEEvent(data);
+            } catch (e) {
+              console.error('解析 SSE 数据失败:', e, line);
+            }
           }
-        );
+        };
 
-        if (response.data.success) {
+        const handleSSEEvent = (data: MeetingEvent) => {
-          const data = response.data;
+          switch (data.type) {
+            case 'session_start':
+              setSessionId(data.session_id || null);
+              break;
 
-          setCurrentRound(data.round_number);
+            case 'speaking_start':
-          setIsConcluded(data.is_concluded);
+              setSpeakingRoleId(data.role_id || null);
+              setStatus(MeetingStatus.SPEAKING);
+              if (data.role_id && data.role_name) {
+                const streamingMsg = createStreamingMessage(
+                  data.role_id,
+                  data.role_name,
+                  nextRound
+                );
+                addMessage(streamingMsg);
+              }
+              break;
 
-          // 添加新的消息
+            case 'tool_call_start':
-          data.messages.forEach((msg) => {
+              if (data.role_id && data.tool_call_id && data.tool_name) {
-            addMessage(msg);
+                const toolCall: ToolCallResult = {
-          });
+                  tool_call_id: data.tool_call_id,
+                  tool_name: data.tool_name,
+                  arguments: data.arguments,
+                  status: 'calling',
+                };
+                addToolCallToMessage(data.role_id, toolCall);
+              }
+              break;
 
-          // 设置结论
+            case 'tool_call_result':
-          if (data.conclusion) {
+              if (data.role_id && data.tool_call_id) {
-            setConclusion(data.conclusion);
+                updateToolCallResult(
+                  data.role_id,
+                  data.tool_call_id,
+                  data.result,
+                  data.status || 'success',
+                  data.execution_time
+                );
+              }
+              break;
+
+            case 'content_delta':
+              if (data.role_id && data.content) {
+                updateMessageContent(data.role_id, data.content);
+              }
+              break;
+
+            case 'message_complete':
+              if (data.role_id) {
+                finishStreamingMessage(data.role_id, data.content);
+                setSpeakingRoleId(null);
+              }
+              break;
+
+            case 'round_end':
+              setCurrentRound(data.round_number || nextRound);
+              setIsConcluded(data.is_concluded || false);
+              setStatus(
+                data.is_concluded
+                  ? MeetingStatus.CONCLUDED
+                  : MeetingStatus.WAITING_INPUT
+              );
+              setIsLoading(false);
+              break;
+
+            case 'error':
+              console.error('会议错误:', data.error);
+              setStatus(MeetingStatus.ERROR);
+              setIsLoading(false);
+              onToast?.({
+                title: '会议出错',
+                description: data.error || '未知错误',
+                status: 'error',
+              });
+              break;
           }
+        };
 
-          setStatus(
+        // 读取流
-            data.is_concluded
+        while (true) {
-              ? MeetingStatus.CONCLUDED
+          const { done, value } = await reader.read();
-              : MeetingStatus.WAITING_INPUT
+          if (done) break;
-          );
+
+          buffer += decoder.decode(value, { stream: true });
+          const lines = buffer.split('\n');
+          buffer = lines.pop() || '';
+
+          for (const line of lines) {
+            if (line.trim()) {
+              processLine(line);
+            }
+          }
+        }
+
+        // 处理剩余 buffer
+        if (buffer.trim()) {
+          processLine(buffer);
         }
       } catch (error: any) {
         console.error('继续会议失败:', error);
         setStatus(MeetingStatus.ERROR);
+        setIsLoading(false);
         onToast?.({
           title: '继续会议失败',
-          description: error.response?.data?.detail || error.message,
+          description: error.message || '请稍后重试',
           status: 'error',
         });
-      } finally {
-        setIsLoading(false);
       }
     },
-    [currentTopic, userId, userNickname, sessionId, messages, addMessage, onToast]
+    [
+      currentTopic,
+      userId,
+      userNickname,
+      sessionId,
+      messages,
+      currentRound,
+      addMessage,
+      createStreamingMessage,
+      addToolCallToMessage,
+      updateToolCallResult,
+      updateMessageContent,
+      finishStreamingMessage,
+      onToast,
+    ]
   );
 
   /**