feat: expose graph orchestration mode, fix pipeline multi-agent, add Feishu tools (Phase 3)

增强编排 + 飞书深度集成：
- Graph 模式：暴露 orchestrator._graph() 到 run() 方法，workflow_integration 支持 graph nodes/edges
- Pipeline 修复：多 Agent 按步骤轮转分配，不再只用 agents[0]
- 4个飞书操作工具: feishu_create_doc / feishu_create_calendar_event / feishu_search_contacts / feishu_send_approval
- 飞书 @mention→Goal：feishu/ orange WS handler 支持 "目标: xxx" 触发自动创建 Goal

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

backend/app/services/orange_ws_handler.py

@@ -137,6 +137,22 @@ def _make_llm_logger(db, agent_id: Optional[str] = None, user_id: Optional[str]
     return _log
 
 
+async def _handle_goal_creation(db, user_id: str, goal_title: str, open_id: str):
+    """橙子飞书消息中创建 Goal 并异步启动执行。"""
+    from app.services.goal_service import create_goal as svc_create_goal, update_goal
+    from app.tasks.goal_tasks import execute_goal_task
+
+    try:
+        goal = svc_create_goal(db=db, creator_id=user_id, title=goal_title, priority=5)
+        _reply_to_feishu(open_id, f"✅ 目标已创建: **{goal.title}**\n正在分解任务并启动执行...")
+        task = execute_goal_task.delay(str(goal.id))
+        logger.info("橙子触发 Goal 创建: goal_id=%s celery_task=%s", goal.id, task.id)
+        update_goal(db, str(goal.id), status="active")
+    except Exception as e:
+        logger.error("橙子 Goal 创建失败: %s", e)
+        _reply_to_feishu(open_id, f"创建目标失败: {e}")
+
+
 async def _handle_message_async(data):
     """异步处理橙子消息 — 固定使用橙子助手 Agent。"""
     open_id = _get_sender_open_id(data)
@@ -226,6 +242,15 @@ async def _handle_message_async(data):
             memory_scope_id=str(agent.id),
         )
 
+        # ── 目标/任务意图检测 ──
+        for trigger in ["创建目标:", "目标:", "new goal:", "goal:"]:
+            if text.lower().startswith(trigger.lower()):
+                goal_title = text[len(trigger):].strip()
+                if goal_title:
+                    await _handle_goal_creation(db, user.id, goal_title[:500], open_id)
+                    return
+                break
+
         on_llm_call = _make_llm_logger(db, agent_id=str(agent.id))
         runtime = AgentRuntime(config=config, on_llm_call=on_llm_call)
         result = await runtime.run(text)