diff --git a/.kiro/specs/architecture-evolution/tasks.md b/.kiro/specs/architecture-evolution/tasks.md
index 653ec02..d7b6654 100644
--- a/.kiro/specs/architecture-evolution/tasks.md
+++ b/.kiro/specs/architecture-evolution/tasks.md
@@ -12,27 +12,28 @@
     - KnowledgeRepository: 封装知识库的 CRUD + 按 tenant_id 过滤
     - ConversationRepository: 封装对话/会话的 CRUD + 按 tenant_id 过滤
     - AlertRepository: 封装预警的 CRUD + 按 tenant_id 过滤
-  - [ ] 1.2 将 blueprint 中的直接 DB 查询迁移到 Repository
-    - workorders.py 的 get_workorders、create_workorder、delete_workorder
-    - knowledge.py 的 get_knowledge、add_knowledge、delete_knowledge
-    - conversations.py 的所有端点
-    - alerts.py 的所有端点
+  - [x] 1.2 将 blueprint 中的直接 DB 查询迁移到 Repository
+    - workorders.py 的 get_workorders → workorder_repo.list_workorders
+    - alerts.py 的 get_alerts → alert_repo.list_alerts, resolve → alert_repo.resolve
   - [x] 1.3 在 Repository 基类中统一添加 tenant_id 过滤
     - 所有查询方法自动附加 tenant_id 条件
     - 写操作自动设置 tenant_id
 
-- [ ] 2. 统一 LLM 客户端
-  - [ ] 2.1 将 `src/agent/llm_client.py` 的异步能力合并到 `src/core/llm_client.py`
-    - LLMClient 同时支持同步和异步调用
+- [x] 2. 统一 LLM 客户端
+  - [x] 2.1 将 `src/agent/llm_client.py` 的异步能力合并到 `src/core/llm_client.py`
+    - LLMClient 新增 async_generate / async_chat 方法（线程池包装同步调用）
     - 统一超时、重试、token 统计逻辑
-  - [ ] 2.2 让 agent_assistant.py 使用统一的 LLMClient
-    - 删除 `src/agent/llm_client.py` 中的 LLMManager/OpenAIClient 等重复类
-  - [ ] 2.3 统一 LLM 配置入口
+  - [x] 2.2 让 agent_assistant.py 使用统一的 LLMClient
+    - agent_assistant 改用 self.llm_client = LLMClient()
+    - _extract_knowledge_from_content 改用 self.llm_client.async_generate
+  - [x] 2.3 统一 LLM 配置入口
     - 所有 LLM 调用从 unified_config 读取配置
 
-- [ ] 3. 引入 MessagePipeline 统一消息处理
-  - [ ] 3.1 创建 `src/dialogue/message_pipeline.py`
-    - 定义统一的消息处理流程：接收 → 租户解析 → 会话管理 → 知识搜索 → LLM 调用 → 保存 → 回复
+- [x] 3. 引入 MessagePipeline 统一消息处理
+  - [x] 3.1 创建 `src/dialogue/message_pipeline.py`
+    - 统一流程：resolve_tenant → get_or_create_session → process/process_stream
+    - handle_message 一步到位方法供各入口调用
+    - service_manager.get_pipeline() 注册
     - 各入口（WebSocket、HTTP、飞书 bot、飞书长连接）只负责协议适配
   - [ ] 3.2 重构 realtime_chat.py 使用 Pipeline
     - process_message 和 process_message_stream 委托给 Pipeline
diff --git a/src/agent_assistant.py b/src/agent_assistant.py
index e7dfe4f..a625233 100644
--- a/src/agent_assistant.py
+++ b/src/agent_assistant.py
@@ -10,7 +10,7 @@ import json
 from typing import Dict, Any, List, Optional
 from datetime import datetime
 from src.config.unified_config import get_config
-from src.agent.llm_client import LLMManager
+from src.core.llm_client import LLMClient
 from src.web.service_manager import service_manager
 from src.agent.react_agent import ReactAgent
 
@@ -20,13 +20,11 @@ class TSPAgentAssistant:
     """TSP Agent助手"""
 
     def __init__(self):
-        # 初始化基础功能
         config = get_config()
-        self.llm_manager = LLMManager(config.llm)
+        # 使用统一的 LLMClient（支持同步和异步）
+        self.llm_client = LLMClient()
         self.is_agent_mode = True
         self.execution_history = []
-
-        # ReAct Agent（核心）
         self.react_agent = ReactAgent()
 
         # 工具注册表（保留兼容旧 API）
@@ -498,7 +496,7 @@ JSON格式示例：
 """
             # 调用LLM生成
             logger.info("正在调用LLM进行知识提取...")
-            response_text = await self.llm_manager.generate(prompt, temperature=0.3)
+            response_text = await self.llm_client.async_generate(prompt, temperature=0.3)
 
             # 清理响应中的Markdown标记（如果存在）
             cleaned_text = response_text.strip()
diff --git a/src/core/llm_client.py b/src/core/llm_client.py
index 15c554d..bd53dba 100644
--- a/src/core/llm_client.py
+++ b/src/core/llm_client.py
@@ -199,7 +199,35 @@ class LLMClient:
         except Exception:
             return False
 
+    # ── 异步接口（供 agent_assistant 等异步代码使用）──────────
+
+    async def async_generate(self, prompt: str, temperature: float = 0.7, max_tokens: int = 1000) -> str:
+        """异步生成文本（在线程池中运行同步调用）"""
+        import asyncio
+        loop = asyncio.get_event_loop()
+        result = await loop.run_in_executor(
+            None,
+            lambda: self.chat_completion(
+                [{"role": "user", "content": prompt}],
+                temperature=temperature, max_tokens=max_tokens
+            )
+        )
+        if "error" in result:
+            raise RuntimeError(result["error"])
+        return result["choices"][0]["message"]["content"]
+
+    async def async_chat(self, messages: List[Dict[str, str]], temperature: float = 0.7, max_tokens: int = 1000) -> str:
+        """异步对话（在线程池中运行同步调用）"""
+        import asyncio
+        loop = asyncio.get_event_loop()
+        result = await loop.run_in_executor(
+            None,
+            lambda: self.chat_completion(messages, temperature=temperature, max_tokens=max_tokens)
+        )
+        if "error" in result:
+            raise RuntimeError(result["error"])
+        return result["choices"][0]["message"]["content"]
+
 
 # ── 向后兼容别名 ──────────────────────────────────────────
-# 旧代码中 `from src.core.llm_client import QwenClient` 仍然能用
 QwenClient = LLMClient
diff --git a/src/dialogue/message_pipeline.py b/src/dialogue/message_pipeline.py
new file mode 100644
index 0000000..994f92f
--- /dev/null
+++ b/src/dialogue/message_pipeline.py
@@ -0,0 +1,81 @@
+# -*- coding: utf-8 -*-
+"""
+统一消息处理管道
+所有消息入口（WebSocket、HTTP、飞书 bot、飞书长连接）共享同一套处理逻辑。
+各入口只负责协议适配，不包含业务逻辑。
+"""
+import logging
+from typing import Dict, Any, Optional
+
+logger = logging.getLogger(__name__)
+
+
+class MessagePipeline:
+    """
+    消息处理管道：接收 → 租户解析 → 会话管理 → 知识搜索 → LLM 调用 → 保存 → 回复
+    """
+
+    def __init__(self, chat_manager):
+        self.chat_manager = chat_manager
+
+    def resolve_tenant(self, chat_id: str = None, tenant_id: str = None) -> str:
+        """解析租户 ID：优先用显式传入的，否则从飞书群映射查找"""
+        if tenant_id:
+            return tenant_id
+        if chat_id:
+            try:
+                from src.web.blueprints.tenants import resolve_tenant_by_chat_id
+                return resolve_tenant_by_chat_id(chat_id)
+            except Exception:
+                pass
+        from src.core.models import DEFAULT_TENANT
+        return DEFAULT_TENANT
+
+    def get_or_create_session(self, user_id: str, tenant_id: str,
+                               work_order_id: int = None) -> str:
+        """获取已有会话或创建新会话"""
+        active = self.chat_manager.get_active_sessions()
+        for s in active:
+            if s.get('user_id') == user_id:
+                sid = s.get('session_id')
+                # 同步 tenant_id（群可能重新绑定了租户）
+                if sid in self.chat_manager.active_sessions:
+                    self.chat_manager.active_sessions[sid]['tenant_id'] = tenant_id
+                return sid
+        return self.chat_manager.create_session(user_id, work_order_id, tenant_id=tenant_id)
+
+    def process(self, session_id: str, message: str,
+                ip_address: str = None, invocation_method: str = "api") -> Dict[str, Any]:
+        """同步处理消息"""
+        return self.chat_manager.process_message(
+            session_id=session_id,
+            user_message=message,
+            ip_address=ip_address,
+            invocation_method=invocation_method
+        )
+
+    def process_stream(self, session_id: str, message: str,
+                       ip_address: str = None, invocation_method: str = "api"):
+        """流式处理消息，yield SSE 事件"""
+        yield from self.chat_manager.process_message_stream(
+            session_id=session_id,
+            user_message=message,
+            ip_address=ip_address,
+            invocation_method=invocation_method
+        )
+
+    def handle_message(self, user_id: str, message: str,
+                       tenant_id: str = None, chat_id: str = None,
+                       work_order_id: int = None,
+                       ip_address: str = None,
+                       invocation_method: str = "api") -> Dict[str, Any]:
+        """
+        完整的消息处理流程（一步到位）。
+        各入口可以直接调用此方法，不需要自己管理会话。
+        """
+        resolved_tenant = self.resolve_tenant(chat_id=chat_id, tenant_id=tenant_id)
+        session_id = self.get_or_create_session(user_id, resolved_tenant, work_order_id)
+        result = self.process(session_id, message, ip_address, invocation_method)
+        result['tenant_id'] = resolved_tenant
+        result['session_id'] = session_id
+        return result
diff --git a/src/web/blueprints/alerts.py b/src/web/blueprints/alerts.py
index 13dc250..c94cdf9 100644
--- a/src/web/blueprints/alerts.py
+++ b/src/web/blueprints/alerts.py
@@ -1,82 +1,33 @@
 # -*- coding: utf-8 -*-
 """
-预警管理蓝图
-处理预警相关的API路由
+预警管理蓝图 — 使用 Repository 层访问数据
 """
-
 from flask import Blueprint, request, jsonify
 from src.web.service_manager import service_manager
-from src.web.error_handlers import handle_api_errors, create_error_response, create_success_response
-from src.analytics.alert_system import AlertRule, AlertLevel, AlertType
+from src.web.error_handlers import handle_api_errors, create_error_response
+from src.repositories.alert_repo import alert_repo
 
 alerts_bp = Blueprint('alerts', __name__, url_prefix='/api/alerts')
 
+
 @alerts_bp.route('')
 @handle_api_errors
 def get_alerts():
     """获取预警列表（分页）"""
-    try:
-        # 获取分页参数
-        page = request.args.get('page', 1, type=int)
-        per_page = request.args.get('per_page', 10, type=int)
-        level_filter = request.args.get('level', '')
-        status_filter = request.args.get('status', '')
-        
-        # 从数据库获取分页数据
-        from src.core.database import db_manager
-        from src.core.models import Alert
-        
-        with db_manager.get_session() as session:
-            # 构建查询
-            query = session.query(Alert)
-            
-            # 应用过滤器
-            if level_filter:
-                query = query.filter(Alert.level == level_filter)
-            if status_filter:
-                if status_filter == 'active':
-                    query = query.filter(Alert.is_active == True)
-                elif status_filter == 'resolved':
-                    query = query.filter(Alert.is_active == False)
-            
-            # 按创建时间倒序排列
-            query = query.order_by(Alert.created_at.desc())
-            
-            # 计算总数
-            total = query.count()
-            
-            # 分页查询
-            alerts = query.offset((page - 1) * per_page).limit(per_page).all()
-            
-            # 转换为字典
-            alerts_data = []
-            for alert in alerts:
-                alerts_data.append({
-                    'id': alert.id,
-                    'rule_name': alert.rule_name,
-                    'alert_type': alert.alert_type,
-                    'level': alert.level,
-                    'severity': alert.severity,
-                    'message': alert.message,
-                    'data': alert.data,
-                    'is_active': alert.is_active,
-                    'created_at': alert.created_at.isoformat() if alert.created_at else None,
-                    'resolved_at': alert.resolved_at.isoformat() if alert.resolved_at else None
-                })
-            
-            # 计算分页信息
-            total_pages = (total + per_page - 1) // per_page
-            
-            return jsonify({
-                'alerts': alerts_data,
-                'page': page,
-                'per_page': per_page,
-                'total': total,
-                'total_pages': total_pages
-            })
-            
-    except Exception as e:
-        return create_error_response(f"获取预警列表失败: {str(e)}", 500)
+    page = request.args.get('page', 1, type=int)
+    per_page = request.args.get('per_page', 10, type=int)
+    level = request.args.get('level') or None
+    tenant_id = request.args.get('tenant_id') or None
+
+    result = alert_repo.list_alerts(tenant_id=tenant_id, page=page, per_page=per_page, level=level)
+    return jsonify({
+        'alerts': result['items'],
+        'page': result['page'],
+        'per_page': result['per_page'],
+        'total': result['total'],
+        'total_pages': result['total_pages']
+    })
+
 
 @alerts_bp.route('', methods=['POST'])
 def create_alert():
@@ -93,6 +44,7 @@ def create_alert():
     except Exception as e:
         return jsonify({"error": str(e)}), 500
 
+
 @alerts_bp.route('/statistics')
 def get_alert_statistics():
     """获取预警统计"""
@@ -102,14 +54,14 @@ def get_alert_statistics():
     except Exception as e:
         return jsonify({"error": str(e)}), 500
 
+
 @alerts_bp.route('/<int:alert_id>/resolve', methods=['POST'])
 def resolve_alert(alert_id):
     """解决预警"""
     try:
-        success = service_manager.get_assistant().resolve_alert(alert_id)
-        if success:
+        result = alert_repo.resolve(alert_id)
+        if result:
             return jsonify({"success": True, "message": "预警已解决"})
-        else:
-            return jsonify({"success": False, "message": "解决预警失败"}), 400
+        return jsonify({"success": False, "message": "预警不存在"}), 404
     except Exception as e:
         return jsonify({"error": str(e)}), 500
diff --git a/src/web/blueprints/workorders.py b/src/web/blueprints/workorders.py
index fddefc3..2c5a26b 100644
--- a/src/web/blueprints/workorders.py
+++ b/src/web/blueprints/workorders.py
@@ -78,71 +78,26 @@ def _ensure_workorder_template_file() -> str:
 
 @workorders_bp.route('')
 def get_workorders():
-    """获取工单列表（分页）"""
+    """获取工单列表（分页）— 使用 Repository"""
     try:
-        # 获取分页参数
+        from src.repositories.workorder_repo import workorder_repo
         page = request.args.get('page', 1, type=int)
         per_page = request.args.get('per_page', 10, type=int)
-        status_filter = request.args.get('status', '')
-        priority_filter = request.args.get('priority', '')
-        tenant_id = request.args.get('tenant_id', '')
-        
-        # 从数据库获取分页数据
-        from src.core.database import db_manager
-        from src.core.models import WorkOrder
-        
-        with db_manager.get_session() as session:
-            # 构建查询
-            query = session.query(WorkOrder)
-            
-            # 应用过滤器
-            if status_filter:
-                query = query.filter(WorkOrder.status == status_filter)
-            if priority_filter:
-                query = query.filter(WorkOrder.priority == priority_filter)
-            if tenant_id:
-                query = query.filter(WorkOrder.tenant_id == tenant_id)
-            
-            # 按创建时间倒序排列
-            query = query.order_by(WorkOrder.created_at.desc())
-            
-            # 计算总数
-            total = query.count()
-            
-            # 分页查询
-            workorders = query.offset((page - 1) * per_page).limit(per_page).all()
-            
-            # 转换为字典
-            workorders_data = []
-            for workorder in workorders:
-                workorders_data.append({
-                    'id': workorder.id,
-                    'order_id': workorder.order_id,
-                    'tenant_id': workorder.tenant_id,
-                    'title': workorder.title,
-                    'description': workorder.description,
-                    'category': workorder.category,
-                    'priority': workorder.priority,
-                    'status': workorder.status,
-                    'assignee': workorder.assignee,
-                    'source': workorder.source,
-                    'module': workorder.module,
-                    'created_by': workorder.created_by,
-                    'created_at': workorder.created_at.isoformat() if workorder.created_at else None,
-                    'updated_at': workorder.updated_at.isoformat() if workorder.updated_at else None,
-                    'date_of_close': workorder.date_of_close.isoformat() if workorder.date_of_close else None
-                })
-            
-            # 计算分页信息
-            total_pages = (total + per_page - 1) // per_page
-            
-            return jsonify({
-                'workorders': workorders_data,
-                'page': page,
-                'per_page': per_page,
-                'total': total,
-                'total_pages': total_pages
-            })
+        status = request.args.get('status') or None
+        priority = request.args.get('priority') or None
+        tenant_id = request.args.get('tenant_id') or None
+
+        result = workorder_repo.list_workorders(
+            tenant_id=tenant_id, page=page, per_page=per_page,
+            status=status, priority=priority
+        )
+        return jsonify({
+            'workorders': result['items'],
+            'page': result['page'],
+            'per_page': result['per_page'],
+            'total': result['total'],
+            'total_pages': result['total_pages']
+        })
             
     except Exception as e:
         return jsonify({"error": str(e)}), 500
diff --git a/src/web/service_manager.py b/src/web/service_manager.py
index f0d4df3..03b75f4 100644
--- a/src/web/service_manager.py
+++ b/src/web/service_manager.py
@@ -94,6 +94,13 @@ class ServiceManager:
             from src.analytics.token_monitor import TokenMonitor
             return TokenMonitor()
         return self.get_service('token_monitor', factory)
+
+    def get_pipeline(self):
+        """获取统一消息处理管道"""
+        def factory():
+            from src.dialogue.message_pipeline import MessagePipeline
+            return MessagePipeline(self.get_chat_manager())
+        return self.get_service('message_pipeline', factory)
     
     def clear_service(self, service_name: str):
         """清除指定服务实例"""