Refactor Step 3: Fix threading context loss, enable real-time streaming, and optimize UI/UX flow.

- Core Fix: Implemented `patch_repo_context` (monkey patch) in `planner_service` to resolve session ID loss during multi-threaded tool execution.
- Service: Refactored `run_step3_team` to yield real-time events (reasoning, report stream, agent status) to the UI.
- UI/App: Updated `step3_wrapper` and `build_interface` wiring to ensure all 9 outputs are correctly mapped, fixing the frozen UI issue.
- UX: Enabled immediate visibility of Report/Map tabs upon planning start and added auto-switch logic to the Report tab.
- Cleanup: Synced agent lists (removed 'planner', confirmed 'presenter') across `app.py` and `results.py`."

app.py

@@ -15,6 +15,7 @@ from config import APP_TITLE, DEFAULT_SETTINGS
 from ui.theme import get_enhanced_css
 from ui.renderers import (
     create_agent_stream_output,
+    create_summary_card,
     create_agent_card_enhanced,
     get_reasoning_html_reversed,
     generate_chat_history_html_bubble
@@ -48,15 +49,39 @@ class LifeFlowAI:
 
     def _get_agent_outputs(self, active_agent: str = None, status: str = "idle", message: str = "Waiting") -> List[str]:
         """輔助函數：生成 Agent 卡片 HTML"""
-        agents = ['planner', 'scout', 'optimizer', 'validator', 'weather', 'traffic']
+
+        # 🔥 [關鍵修正] 更新為你的新 Agent 名單 (共 5 個)
+        # 必須與 create_team_area 中的順序完全一致！
+        agents = ['team', 'scout', 'optimizer', 'navigator', 'weatherman', 'presenter']
+
         outputs = []
         for agent in agents:
             if agent == active_agent:
                 outputs.append(create_agent_card_enhanced(agent, status, message))
             else:
                 outputs.append(create_agent_card_enhanced(agent, "idle", "On standby"))
+
         return outputs
 
+    def _update_task_summary(self, session: UserSession) -> str:
+        """統一計算並回傳 Summary Card HTML"""
+        tasks = session.task_list
+        if not tasks:
+            return create_summary_card(0, 0, 0)
+
+        high_priority = sum(1 for t in tasks if t.get("priority", "").upper() == "HIGH")
+
+        total_time = 0
+        for t in tasks:
+            dur_str = str(t.get("duration", "0"))
+            try:
+                val = int(dur_str.split()[0])
+                total_time += val
+            except (ValueError, IndexError):
+                pass
+
+        return create_summary_card(len(tasks), high_priority, total_time)
+
     # -------------------------------------------------------------------------
     # Step 1: Analyze (分析任務)
     # -------------------------------------------------------------------------
@@ -65,7 +90,7 @@ class LifeFlowAI:
 
         # 1. 輸入驗證
         if not user_input or not user_input.strip():
-            agent_outputs = self._get_agent_outputs("planner", "idle", "Waiting")
+            agent_outputs = self._get_agent_outputs("team", "idle", "Waiting")
             yield (
                 "<div style='color: #ef4444; font-weight: bold; padding: 5px;'>⚠️ Please describe your plans first!</div>",
                 gr.HTML(), gr.HTML(),
@@ -80,78 +105,90 @@ class LifeFlowAI:
 
         for event in iterator:
             evt_type = event.get("type")
-            agent_status = event.get("agent_status", ("planner", "idle", "Waiting"))
+            agent_status = event.get("agent_status", ("team", "idle", "Waiting"))
             agent_outputs = self._get_agent_outputs(*agent_status)
             current_session = event.get("session", session)
             reasoning_html = get_reasoning_html_reversed(current_session.reasoning_messages)
 
             if evt_type == "stream":
-                # Streaming: 保持 Input 顯示，Confirm 隱藏
+                html_output = create_agent_stream_output(event.get("stream_text", ""))
                 yield (
-                    create_agent_stream_output().replace("Ready to analyze...", event.get("stream_text", "")),
+                    html_output,
                     gr.HTML(), gr.HTML(), reasoning_html,
                     gr.update(visible=False), gr.update(visible=False), gr.HTML(),
                     f"Processing: {agent_status[2]}", gr.update(visible=True), *agent_outputs, current_session.to_dict()
                 )
             elif evt_type == "complete":
-                # Complete: 切換到 Confirm 頁面
+                summary_html = self._update_task_summary(current_session)
+
                 task_html = self.service.generate_task_list_html(current_session)
-                summary_html = f"<div class='summary-card'>Found {len(current_session.task_list)} tasks</div>"
-                final_agents = self._get_agent_outputs("planner", "complete", "Tasks ready")
+
+                final_text = event.get("stream_text", "Analysis complete!")
+                html_output = create_agent_stream_output(final_text)
+
+                final_agents = self._get_agent_outputs("team", "complete", "Tasks ready")
+
                 yield (
-                    create_agent_stream_output().replace("Ready...", event.get("stream_text", "")),
-                    gr.HTML(value=summary_html), gr.HTML(value=task_html), reasoning_html,
+                    html_output,
+                    gr.HTML(value=summary_html), # ✅ 更新 Summary Component
+                    gr.HTML(value=task_html),
+                    reasoning_html,
                     gr.update(visible=True), gr.update(visible=True),
                     generate_chat_history_html_bubble(current_session),
                     "✓ Tasks extracted", gr.update(visible=False), *final_agents, current_session.to_dict()
                 )
+
             elif evt_type == "error":
                 err_msg = event.get("message", "Unknown error")
-                error_agents = self._get_agent_outputs("planner", "idle", "Error")
+                error_agents = self._get_agent_outputs("team", "idle", "Error")
+
                 yield (
                     f"<div style='color:red'>Error: {err_msg}</div>", gr.HTML(), gr.HTML(), reasoning_html,
                     gr.update(visible=False), gr.update(visible=False), gr.HTML(),
                     f"Error: {err_msg}", gr.update(visible=True), *error_agents, current_session.to_dict()
                 )
 
+
     # -------------------------------------------------------------------------
     # Chat Modification (修改任務)
     # -------------------------------------------------------------------------
     def chat_wrapper(self, msg, session_data):
         session = UserSession.from_dict(session_data)
         iterator = self.service.modify_task_chat(msg, session)
+
         for event in iterator:
             current_session = event.get("session", session)
+
+            # 🔥 [Update 2] Step 2 Modify: 每次對話更新後，重新計算 Summary
+            new_summary_html = self._update_task_summary(current_session)
+
             yield (
-                generate_chat_history_html_bubble(current_session),
-                self.service.generate_task_list_html(current_session),
-                current_session.to_dict()
+                generate_chat_history_html_bubble(current_session),  # Output 1: Chat History
+                self.service.generate_task_list_html(current_session),  # Output 2: Task List
+                new_summary_html,  # Output 3: ✅ Summary Card (新增這個)
+                current_session.to_dict()  # Output 4: Session State
             )
 
     # -------------------------------------------------------------------------
     # Step 2 -> 3 Transition (驗證與過渡)
     # -------------------------------------------------------------------------
     def transition_to_planning(self, session_data):
-        """
-        Step 2 -> Step 3 過渡函數：
-        1. 檢查是否有任務 (防呆)
-        2. 若無任務：報錯並留在原位
-        3. 若有任務：隱藏 Confirm Area，顯示 Team Area，並執行 Step 2 (Search)
-        """
         session = UserSession.from_dict(session_data)
 
         # 1. 🛑 防呆檢查
         if not session.task_list or len(session.task_list) == 0:
-            agent_outputs = self._get_agent_outputs("planner", "idle", "No tasks")
+            agent_outputs = self._get_agent_outputs("team", "idle", "No tasks")
             reasoning = get_reasoning_html_reversed(session.reasoning_messages)
 
             return (
-                gr.update(visible=True),  # task_confirm_area (保持顯示)
+                gr.update(visible=True),  # task_confirm_area
                 gr.update(visible=True),  # chat_input_area
-                gr.update(visible=False),  # team_area
-                gr.update(),  # tabs
+                gr.update(visible=False), # team_area
+                gr.update(),              # tabs
+                gr.update(),              # 🔥 [新增] report_tab (維持現狀)
+                gr.update(),              # 🔥 [新增] map_tab (維持現狀)
                 reasoning,
-                "⚠️ Error: No tasks to plan! Please add tasks.",  # Status Bar
+                "⚠️ Error: No tasks to plan! Please add tasks.",
                 *agent_outputs,
                 session.to_dict()
             )
@@ -164,10 +201,12 @@ class LifeFlowAI:
         agent_outputs = self._get_agent_outputs("scout", "working", "Searching POIs...")
 
         return (
-            gr.update(visible=False),  # task_confirm_area (隱藏)
-            gr.update(visible=False),  # chat_input_area
-            gr.update(visible=True),  # team_area (顯示)
-            gr.update(selected="ai_conversation_tab"),  # 切換 Tab
+            gr.update(visible=False),          # task_confirm_area
+            gr.update(visible=False),          # chat_input_area
+            gr.update(visible=True),           # team_area
+            gr.update(selected="report_tab"), # tabs (預設看 Log)
+            gr.update(visible=True),           # 🔥 [關鍵修改] report_tab 設為可見！
+            gr.update(visible=False),           # 🔥 [關鍵修改] map_tab 設為可見！
             reasoning_html,
             "🗺️ Scout is searching locations...",
             *agent_outputs,
@@ -179,17 +218,38 @@ class LifeFlowAI:
     # -------------------------------------------------------------------------
     def step3_wrapper(self, session_data):
         session = UserSession.from_dict(session_data)
-        if not session.task_list:  # 二次防呆
-            yield ("", session.to_dict())
+
+        # 1. 防呆回傳 (9個值)
+        if not session.task_list:
+            agent_outputs = self._get_agent_outputs("team", "idle", "No tasks")
+            # yield 順序: Report, Reasoning, Agents(6個), Session
+            yield ("", "", *agent_outputs, session.to_dict())
             return
 
         iterator = self.service.run_step3_team(session)
+        current_report = "⏳ Generating plan..."
+
         for event in iterator:
+            evt_type = event.get("type")
             current_session = event.get("session", session)
-            if event["type"] == "complete":
-                yield (event.get("report_html", ""), current_session.to_dict())
-            elif event["type"] == "error":
-                yield (f"Error: {event.get('message')}", current_session.to_dict())
+
+            # 準備 UI Data
+            reasoning_html = get_reasoning_html_reversed(current_session.reasoning_messages)
+            agent_status = event.get("agent_status", ("team", "working", "Processing..."))
+            agent_outputs = self._get_agent_outputs(*agent_status)
+
+            # 更新 Report 暫存
+            if evt_type == "report_stream":
+                current_report = event.get("content", current_report)
+
+            # 🔥 統一回傳 (確保任何事件都更新所有 UI)
+            # 必須是 9 個值：Report, Reasoning, Agent1...6, Session
+            yield (
+                current_report,
+                reasoning_html,
+                *agent_outputs,
+                current_session.to_dict()
+            )
 
     # -------------------------------------------------------------------------
     # Step 4 Wrapper (Optimization & Results)
@@ -198,7 +258,7 @@ class LifeFlowAI:
         session = UserSession.from_dict(session_data)
         if not session.task_list:  # 二次防呆
             default_map = create_animated_map()
-            agent_outputs = self._get_agent_outputs("planner", "idle", "No tasks")
+            agent_outputs = self._get_agent_outputs("team", "idle", "No tasks")
             return (
                 "", "", "", default_map,
                 gr.update(), gr.update(), "⚠️ Planning aborted",
@@ -310,7 +370,12 @@ class LifeFlowAI:
             chat_event = chat_send.click(
                 fn=self.chat_wrapper,
                 inputs=[chat_input, session_state],
-                outputs=[chat_history_output, task_list_display, session_state]
+                outputs=[
+                    chat_history_output,  # 1. 對話框
+                    task_list_display,  # 2. 任務列表
+                    task_summary_display,  # 🔥 [Critical Fix] 3. 摘要卡片 (必須加這行!)
+                    session_state  # 4. Session
+                ]
             ).then(fn=lambda: "", outputs=[chat_input])
 
             # ---------------------------------------------------------------------
@@ -321,15 +386,28 @@ class LifeFlowAI:
                 fn=self.transition_to_planning,
                 inputs=[session_state],
                 outputs=[
-                    task_confirm_area, chat_input_area, team_area, tabs,
-                    reasoning_output, status_bar, *agent_displays, session_state
+                    task_confirm_area,
+                    chat_input_area,
+                    team_area,
+                    tabs,
+                    report_tab,  # 🔥 [新增] 這裡要接 report_tab
+                    map_tab,  # 🔥 [新增] 這裡要接 map_tab
+                    reasoning_output,
+                    status_bar,
+                    *agent_displays,
+                    session_state
                 ]
             )
 
             step3_event = step2_event.then(
                 fn=self.step3_wrapper,
                 inputs=[session_state],
-                outputs=[report_output, session_state]
+                outputs=[
+                    report_output,  # 1. 報告 Tab
+                    reasoning_output,  # 2. AI Conversation Log (這裡漏掉了!)
+                    *agent_displays,  # 3-8. Agent Grid (這裡也漏掉了!)
+                    session_state  # 9. Session
+                ]
             )
 
             step3_ui_update = step3_event.then(

services/planner_service.py

@@ -7,6 +7,7 @@ import time
 import uuid
 from datetime import datetime
 from typing import Generator, Dict, Any, Tuple, Optional
+from contextlib import contextmanager
 
 # 導入 Core & UI 模組
 from core.session import UserSession
@@ -43,6 +44,44 @@ from src.infra.logger import get_logger
 
 logger = get_logger(__name__)
 
+
+@contextmanager
+def patch_repo_context(session_id: str):
+    """
+    暴力解決方案：
+    在執行期間，暫時將 poi_repo 的 save/add 方法「綁架」，
+    強迫它們在執行前，先設定當前的 Thread Session。
+    """
+
+    # 1. 備份原始方法
+    original_save = getattr(poi_repo, 'save', None)
+    original_add = getattr(poi_repo, 'add', None)
+
+    # 2. 定義「綁架」後的方法
+    def patched_save(*args, **kwargs):
+        set_session_id(session_id)
+        # 呼叫原始方法
+        if original_save:
+            return original_save(*args, **kwargs)
+        elif original_add:
+            return original_add(*args, **kwargs)
+
+    # 3. 執行替換
+    if original_save:
+        setattr(poi_repo, 'save', patched_save)
+    if original_add:
+        setattr(poi_repo, 'add', patched_save)
+
+    try:
+        yield
+    finally:
+        # 4. 恢復原始方法
+        if original_save:
+            setattr(poi_repo, 'save', original_save)
+        if original_add:
+            setattr(poi_repo, 'add', original_add)
+
+
 class PlannerService:
     """
     PlannerService 封裝了所有的業務流程 (Step 1-4)。
@@ -264,10 +303,11 @@ class PlannerService:
                     accumulated_response += content
                     if "@@@" not in accumulated_response:  # 簡單過濾 JSON 標記
                         displayed_text += content
-                        # 🔥 確保每個 Chunk 都觸發 UI 更新
+
+                        formatted_text = displayed_text.replace("\n", "<br/>")
                         yield {
                             "type": "stream",
-                            "stream_text": displayed_text,
+                            "stream_text": formatted_text,
                             "agent_status": ("planner", "working", "Thinking..."),
                             "session": session
                         }
@@ -286,6 +326,7 @@ class PlannerService:
                 session.task_list = self._convert_task_list_to_ui_format(task_list_data)
             except Exception as e:
                 logger.error(f"Failed to parse task_list: {e}")
+                print(json_data)
                 session.task_list = []
 
             # 🛡️ 檢查 2: Planner 是否回傳空列表
@@ -360,7 +401,11 @@ class PlannerService:
                 if chunk.event == RunEvent.run_content:
                     content = chunk.content
                     accumulated_response += content
-                    # 可選：在這裡也可以做 stream update 讓 chat bubble 動態出現文字
+                    if "@@@" not in accumulated_response:  # 簡單過濾 JSON 標記
+                        accumulated_response += content
+                        formatted_text = accumulated_response.replace("\n", "<br/>")
+                        session.chat_history[-1]["message"] += formatted_text
+                        yield {"type": "update_history", "session": session}
 
             json_data = "{" + accumulated_response.split("{", maxsplit=1)[-1]
             json_data = json_data.replace("`", "").replace("@", "").replace("\\", " ").replace("\n", " ")
@@ -407,8 +452,6 @@ class PlannerService:
             set_session_id(session.session_id)
             logger.info(f"🔄 [Step 2] Session Context Set: {session.session_id}")
 
-        self._add_reasoning(session, "team", "🚀 Core Team activated")
-        self._add_reasoning(session, "scout", "Searching for POIs...")
         return {"session": session}
 
     # ================= Step 3: Run Core Team =================
@@ -417,51 +460,107 @@ class PlannerService:
         try:
             session = self._get_live_session(session)
 
-            # 🔥 修正 2: 確保 Session ID 在這個執行緒中生效
             if session.session_id:
                 set_session_id(session.session_id)
                 logger.info(f"🔄 [Step 3] Session Context Set: {session.session_id}")
-            else:
-                logger.warning("⚠️ [Step 3] No Session ID found!")
 
             if not session.task_list:
                 yield {"type": "error", "message": "No tasks to plan.", "session": session}
                 return
 
-            if session.planner_agent is None:
-                raise ValueError("Agents not initialized. Please run Step 1 first.")
-
-            # 準備輸入
+            # 準備 Task List String
             task_list_input = session.planner_agent.get_session_state().get("task_list")
-            if isinstance(task_list_input, dict) or isinstance(task_list_input, list):
-                task_list_str = json.dumps(task_list_input, indent=2, ensure_ascii=False)
-            else:
-                task_list_str = str(task_list_input)
+            task_list_str = json.dumps(task_list_input, indent=2, ensure_ascii=False) if isinstance(task_list_input, (
+            dict, list)) else str(task_list_input)
 
             self._add_reasoning(session, "team", "🎯 Multi-agent collaboration started")
-            yield {"type": "start", "session": session}
-
-            # 執行 Team Run
-            # 注意：如果 Agno 內部使用 ThreadPool，Context 還是可能遺失。
-            # 但既然之前能跑，代表 Agno 應該是在當前執行緒或兼容的 Context 下運行工具的。
-            team_stream = session.core_team.run(
-                f"Plan this trip: {task_list_str}",
-                stream=True, stream_events=True, session_id=session.session_id
-            )
 
-            report_content = ""
-            for event in team_stream:
-                if event.event in [TeamRunEvent.run_content]:
-                    report_content += event.content
-                elif event.event == "tool_call":
-                    tool_name = event.tool_call.get('function', {}).get('name', 'unknown')
-                    self._add_reasoning(session, "team", f"🔧 Tool: {tool_name}")
-                    yield {"type": "reasoning_update", "session": session}
-                elif event.event == TeamRunEvent.run_completed:
-                    self._add_reasoning(session, "team", "🎉 Completed")
+            # 🔥 初始狀態
+            yield {
+                "type": "reasoning_update",
+                "session": session,
+                "agent_status": ("team", "working", "Analyzing tasks...")
+            }
+
+            # 🔥 [CRITICAL FIX] 使用 Patch Context 包裹執行區塊
+            with patch_repo_context(session.session_id):
+
+                team_stream = session.core_team.run(
+                    f"Plan this trip: {task_list_str}",
+                    stream=True, stream_events=True, session_id=session.session_id
+                )
+
+
+                start_time = time.perf_counter()
+                report_content = ""
+                # 🔥 Event Loop: 捕捉事件並 yield 給 UI
+                for event in team_stream:
 
+                    # 1. 報告內容 stream
+                    if event.event == RunEvent.run_content and event.agent_id == "presenter":
+                        report_content += event.content
+                        yield {
+                            "type": "report_stream",
+                            "content": report_content,
+                            "session": session
+                        }
+
+                    # 2. Team Delegate (分派任務)
+                    if event.event == TeamRunEvent.tool_call_started:
+                        tool_name = event.tool.tool_name
+                        if "delegate_task_to_member" in tool_name:
+                            member_id = event.tool.tool_args.get("member_id", "unknown")
+                            msg = f"Delegating to {member_id}..."
+                            self._add_reasoning(session, "team", f"👉 {msg}")
+
+                            yield {
+                                "type": "reasoning_update",
+                                "session": session,
+                                "agent_status": ("team", "working", msg)
+                            }
+                        else:
+                            self._add_reasoning(session, "team", f"🔧 Tool: {tool_name}")
+
+                    # 3. Member Tool Start (成員開始工作)
+                    elif event.event == RunEvent.tool_call_started:
+                        member_id = event.agent_id
+                        tool_name = event.tool.tool_name
+                        self._add_reasoning(session, member_id, f"Using tool: {tool_name}...")
+
+                        yield {
+                            "type": "reasoning_update",
+                            "session": session,
+                            "agent_status": (member_id, "working", f"Running {tool_name}...")
+                        }
+
+                    # 4. Member Tool End (成員工作結束)
+                    elif event.event == RunEvent.tool_call_completed:
+                        member_id = event.agent_id
+                        self._add_reasoning(session, member_id, "✅ Task completed")
+
+                        yield {
+                            "type": "reasoning_update",
+                            "session": session,
+                            "agent_status": (member_id, "idle", "Done")
+                        }
+
+                    # 5. Team Complete
+                    elif event.event == TeamRunEvent.run_completed:
+                        self._add_reasoning(session, "team", "🎉 Planning process finished")
+
+                logger.info(f"Total tokens: {event.metrics.total_tokens}")
+                logger.info(f"Input tokens: {event.metrics.input_tokens}")
+                logger.info(f"Output tokens: {event.metrics.output_tokens}")
+                logger.info(f"Run time (s): {time.perf_counter() - start_time}")
+
+            # 迴圈結束
             report_html = f"## 🎯 Planning Complete\n\n{report_content}"
-            yield {"type": "complete", "report_html": report_html, "session": session}
+            yield {
+                "type": "complete",
+                "report_html": report_html,
+                "session": session,
+                "agent_status": ("team", "complete", "Finished")
+            }
 
         except Exception as e:
             logger.error(f"Team run error: {e}", exc_info=True)

src/agent/setting/team.py

@@ -49,7 +49,7 @@ instructions="""
     - 🛑 **CRITICAL**: You MUST use the ID starting with `final_itinerary_`. DO NOT reuse `navigation_result_`.
     
     #### STEP 6: FINISH
-    - **Action**: Output the text returned by `Presenter` verbatim.
+    - **Action**: Only Return `Completed Task`.
     
     ### 🛡️ EXCEPTION HANDLING
     - **Valid ID Patterns**: `scout_result_`, `optimization_result_`, `navigation_result_`, `final_itinerary_`.

ui/components/results.py

@@ -9,29 +9,28 @@ import gradio as gr
 
 
 def create_team_area(create_agent_card_func):
-    """創建 AI Team 展示區域 (改為 Grid)"""
+    """創建 AI Team 展示區域"""
     with gr.Group(visible=False) as team_area:
         gr.Markdown("### 🤖 Agent Status")
-        # 使用自定義 CSS grid wrapper
-        with gr.Column():
-            # 這裡我們需要一個容器來放入所有的 HTML 卡片，或者將它們合併為一個 HTML 輸出
-            # 為了簡單起見，我們將 agent_displays 保持為 list，但在前端用 flex/grid 呈現
-            # 注意：為了達到最佳 grid 效果，建議在 app.py 中將 agent updates 合併為一個 HTML string
-            # 但為了不改動邏輯，我們這裡運用 CSS (.agent-grid) 包裹
-
-            agent_container_start = gr.HTML('<div class="agent-grid">')
-            agent_displays = []
-            # 這裡對應 app.py 的邏輯，會生成多個 HTML 組件
-            for agent_key in ['planner', 'scout', 'optimizer', 'validator', 'weather', 'traffic']:
-                # 初始狀態
-                card_html = create_agent_card_func(agent_key, "idle", "Waiting")
-                agent_displays.append(gr.HTML(value=card_html, show_label=False))
-
-            agent_container_end = gr.HTML('</div>')
-
-
-            for display in agent_displays:
-                display.elem_classes = ["agent-grid-item"]
+
+        # 使用 HTML 容器
+        agent_container_start = gr.HTML('<div class="agent-grid">')
+        agent_displays = []
+
+        # 🔥 [關鍵修正] 這裡的名單必須跟 app.py 的 _get_agent_outputs 一模一樣！
+        target_agents = ['team', 'scout', 'optimizer', 'navigator', 'weatherman', 'presenter']
+
+        for agent_key in target_agents:
+            # 初始狀態
+            card_html = create_agent_card_func(agent_key, "idle", "Waiting")
+            # 存入 list，這會成為 build_interface 中的 outputs 元件
+            agent_displays.append(gr.HTML(value=card_html, show_label=False))
+
+        agent_container_end = gr.HTML('</div>')
+
+        # 套用 CSS Class
+        for display in agent_displays:
+            display.elem_classes = ["agent-grid-item"]
 
     return team_area, agent_displays
 

ui/renderers.py

@@ -5,12 +5,40 @@ LifeFlow AI - UI Renderers
 from datetime import datetime
 from config import AGENTS_INFO
 
-def create_agent_stream_output() -> str:
-    """創建 Agent 串流輸出 HTML"""
-    return """
+
+# ui/renderers.py
+
+def create_agent_stream_output(text: str = None) -> str:
+    """創建 Agent 串流輸出 HTML (支援動態傳入文字)"""
+
+    # 如果沒有傳入文字，顯示預設值
+    if not text:
+        display_content = 'Ready to analyze your tasks...<span class="stream-cursor"></span>'
+    else:
+        # 這裡直接使用傳入的 text
+        display_content = f'{text}<span class="stream-cursor"></span>'
+
+    return f"""
     <div class="stream-container">
-        <div class="stream-text">Ready to analyze your tasks...<span class="stream-cursor"></span></div>
+        <div class="stream-text" style="white-space: pre-wrap; line-height: 1.6; min-height: 60px;">{display_content}</div>
     </div>
+    <style>
+        .stream-text {{
+            font-family: 'Inter', sans-serif;
+            color: #334155;
+            font-size: 1.05rem;
+        }}
+        .stream-cursor {{
+            display: inline-block;
+            width: 8px;
+            height: 18px;
+            background-color: #6366f1;
+            margin-left: 5px;
+            animation: blink 1s infinite;
+            vertical-align: sub;
+        }}
+        @keyframes blink {{ 0%, 100% {{ opacity: 1; }} 50% {{ opacity: 0; }} }}
+    </style>
     """