diff --git a/docs/ios_setup/ios_setup.md b/docs/ios_setup/ios_setup.md
index a2b22fee..f8b25c46 100644
--- a/docs/ios_setup/ios_setup.md
+++ b/docs/ios_setup/ios_setup.md
@@ -10,7 +10,6 @@
 - iOS 设备（iPhone/iPad）
 - USB 数据线或同一 WiFi 网络
 
-
 ## WebDriverAgent 配置
 
 WebDriverAgent 是 iOS 自动化的核心组件，需要在 iOS 设备上运行。
@@ -28,7 +27,7 @@ cd WebDriverAgent
 
 1. 在 Xcode 中选中 `WebDriverAgent`，出现General、Signing&Capabilities等选项。
 2. 进入 `Signing & Capabilities` 选项卡
-3.   勾选 `Automatically manage signing`。在Team中选择自己的开发者账号
+3. 勾选 `Automatically manage signing`。在Team中选择自己的开发者账号
 4. 将 Bundle ID 改为唯一标识符，例如：`com.yourname.WebDriverAgentRunner`
 ![设置签名1](resources/ios0_WebDriverAgent0.png)
 
@@ -43,7 +42,7 @@ Mac和iPhone有USB和WiFi两种连接方式，建议通过USB方式，成功率
 #### 通过 WiFi 连接
 
 需要满足以下条件：
-1.  通过USB连接。在Finder中选中连接的IPhone，在“通用”中勾选"在 WiFi 中显示这台 iPhone"
+1. 通过USB连接。在Finder中选中连接的IPhone，在“通用”中勾选"在 WiFi 中显示这台 iPhone"
 2. Mac 与 iPhone 处于同一 WiFi 网络之下
 
 #### 具体步骤
@@ -52,7 +51,7 @@ Mac和iPhone有USB和WiFi两种连接方式，建议通过USB方式，成功率
 
 ![选择设备](resources/select-your-iphone-device.png)
 
-3. 长按"▶️"运行按钮，选择 "Test" 后开始编译并部署到你的 iPhone 上
+1. 长按"▶️"运行按钮，选择 "Test" 后开始编译并部署到你的 iPhone 上
 
 ![开始测试](resources/start-wda-testing.png)
 
@@ -85,6 +84,7 @@ brew install libimobiledevice
 # 设备检查
 idevice_id -ln
 ```
+
 2.使用xcodebuild安装WebAgent。命令行也需要进行“设备信任配置”，参考GUI模式下的方法。
 
 ```
@@ -95,6 +95,7 @@ xcodebuild -project WebDriverAgent.xcodeproj \
            -destination 'platform=iOS,name=YOUR_PHONE_NAME' \
            test
 ```
+
 这里，YOUR_PHONE_NAME可以在xcode的GUI中看到。
 WebDriverAgent 成功运行后，会在 Xcode 控制台输出类似以下信息：
 
diff --git a/ios.py b/ios.py
index 78dfc5ea..521d64f5 100755
--- a/ios.py
+++ b/ios.py
@@ -10,7 +10,7 @@
     PHONE_AGENT_MODEL: Model name (default: autoglm-phone-9b)
     PHONE_AGENT_MAX_STEPS: Maximum steps per task (default: 100)
     PHONE_AGENT_WDA_URL: WebDriverAgent URL (default: http://localhost:8100)
-    PHONE_AGENT_DEVICE_ID: iOS device UDID for multi-device setups
+    PHONE_AGENT_DEVICE_ID: iOS device UUID for multi-device setups
 """
 
 import argparse
@@ -99,9 +99,7 @@ def check_system_requirements(wda_url: str = "http://localhost:8100") -> bool:
             print("     4. Or connect via WiFi using device IP")
             all_passed = False
         else:
-            device_names = [
-                d.device_name or d.device_id[:8] + "..." for d in devices
-            ]
+            device_names = [d.device_name or d.device_id[:8] + "..." for d in devices]
             print(f"✅ OK ({len(devices)} device(s): {', '.join(device_names)})")
     except Exception as e:
         print("❌ FAILED")
@@ -261,7 +259,7 @@ def parse_args() -> argparse.Namespace:
     python ios.py --base-url http://localhost:8000/v1
 
     # Run with specific device
-    python ios.py --device-id <UDID>
+    python ios.py --device-id <UUID>
 
     # Use WiFi connection
     python ios.py --wda-url http://192.168.1.100:8100
@@ -315,7 +313,7 @@ def parse_args() -> argparse.Namespace:
         "-d",
         type=str,
         default=os.getenv("PHONE_AGENT_DEVICE_ID"),
-        help="iOS device UDID",
+        help="iOS device UUID",
     )
 
     parser.add_argument(
@@ -326,7 +324,9 @@ def parse_args() -> argparse.Namespace:
     )
 
     parser.add_argument(
-        "--list-devices", action="store_true", help="List connected iOS devices and exit"
+        "--list-devices",
+        action="store_true",
+        help="List connected iOS devices and exit",
     )
 
     parser.add_argument(
@@ -396,7 +396,7 @@ def handle_device_commands(args) -> bool:
                 name_info = device.device_name or "Unnamed"
 
                 print(f"  ✓ {name_info}")
-                print(f"    UDID: {device.device_id}")
+                print(f"    UUID: {device.device_id}")
                 print(f"    Model: {model_info}")
                 print(f"    OS: {ios_info}")
                 print(f"    Connection: {conn_type}")
@@ -474,9 +474,7 @@ def main():
 
     # Create configurations
     model_config = ModelConfig(
-        base_url=args.base_url,
-        model_name=args.model,
-        api_key=args.api_key
+        base_url=args.base_url, model_name=args.model, api_key=args.api_key
     )
 
     agent_config = IOSAgentConfig(
diff --git a/phone_agent/actions/handler.py b/phone_agent/actions/handler.py
index 0bef1c3a..a54850df 100644
--- a/phone_agent/actions/handler.py
+++ b/phone_agent/actions/handler.py
@@ -265,7 +265,7 @@ def _send_keyevent(self, keycode: str) -> None:
         # Handle HDC devices with HarmonyOS-specific keyEvent command
         if device_factory.device_type == DeviceType.HDC:
             hdc_prefix = ["hdc", "-t", self.device_id] if self.device_id else ["hdc"]
-            
+
             # Map common keycodes to HarmonyOS keyEvent codes
             # KEYCODE_ENTER (66) -> 2054 (HarmonyOS Enter key code)
             if keycode == "KEYCODE_ENTER" or keycode == "66":
@@ -283,7 +283,8 @@ def _send_keyevent(self, keycode: str) -> None:
                         # For now, only handle ENTER, other keys may need mapping
                         if "ENTER" in keycode:
                             _run_hdc_command(
-                                hdc_prefix + ["shell", "uitest", "uiInput", "keyEvent", "2054"],
+                                hdc_prefix
+                                + ["shell", "uitest", "uiInput", "keyEvent", "2054"],
                                 capture_output=True,
                                 text=True,
                             )
@@ -297,7 +298,8 @@ def _send_keyevent(self, keycode: str) -> None:
                     else:
                         # Assume it's a numeric code
                         _run_hdc_command(
-                            hdc_prefix + ["shell", "uitest", "uiInput", "keyEvent", str(keycode)],
+                            hdc_prefix
+                            + ["shell", "uitest", "uiInput", "keyEvent", str(keycode)],
                             capture_output=True,
                             text=True,
                         )
@@ -342,22 +344,101 @@ def parse_action(response: str) -> dict[str, Any]:
     Raises:
         ValueError: If the response cannot be parsed.
     """
-    print(f"Parsing action: {response}")
+    if not response or not response.strip():
+        return {"_metadata": "finish", "message": "Model returned an empty action."}
+
     try:
         response = response.strip()
+
+        # 1. Try to extract do(...) or finish(...) using regex if it's wrapped in other text
+        do_match = re.search(r"do\(.*?\)", response, re.DOTALL)
+        finish_match = re.search(r"finish\(.*?\)", response, re.DOTALL)
+
+        if do_match:
+            action_str = do_match.group(0)
+            # Special handling for Type action with text that might contain special characters
+            if 'action="Type"' in action_str or 'action="Type_Name"' in action_str:
+                if "text=" in action_str:
+                    try:
+                        # Try to extract text between quotes more robustly
+                        text_part = action_str.split("text=", 1)[1]
+                        # Find the first and last quote
+                        first_quote = text_part.find('"')
+                        last_quote = text_part.rfind('"')
+                        if (
+                            first_quote != -1
+                            and last_quote != -1
+                            and first_quote < last_quote
+                        ):
+                            text = text_part[first_quote + 1 : last_quote]
+                            # Extract action type
+                            action_type = (
+                                "Type" if 'action="Type"' in action_str else "Type_Name"
+                            )
+                            return {
+                                "_metadata": "do",
+                                "action": action_type,
+                                "text": text,
+                            }
+                    except Exception:
+                        pass  # Fallback to AST if regex fails
+
+            # Standard do(...) parsing using AST
+            try:
+                # Clean up the string for AST
+                clean_str = (
+                    action_str.replace("\n", "\\n")
+                    .replace("\r", "\\r")
+                    .replace("\t", "\\t")
+                )
+                tree = ast.parse(clean_str, mode="eval")
+                if isinstance(tree.body, ast.Call):
+                    call = tree.body
+                    action = {"_metadata": "do"}
+                    for keyword in call.keywords:
+                        key = keyword.arg
+                        try:
+                            value = ast.literal_eval(keyword.value)
+                            action[key] = value
+                        except (ValueError, SyntaxError):
+                            # Fallback for non-literal values (though model should only output literals)
+                            if isinstance(keyword.value, ast.Constant):
+                                action[key] = keyword.value.value
+                            else:
+                                # Last resort: raw string representation
+                                action[key] = str(keyword.value)
+                    return action
+            except (SyntaxError, ValueError) as e:
+                print(f"AST parsing failed for {action_str}: {e}")
+                # If it's a simple do(action="Home") but AST failed, try one more manual parse
+                if 'action="Home"' in action_str:
+                    return {"_metadata": "do", "action": "Home"}
+
+        if finish_match:
+            action_str = finish_match.group(0)
+            # Simple extraction for finish(message="...")
+            message = ""
+            if 'message="' in action_str:
+                parts = action_str.split('message="', 1)[1].rsplit('"', 1)
+                if len(parts) >= 1:
+                    message = parts[0]
+            elif "message='" in action_str:
+                parts = action_str.split("message='", 1)[1].rsplit("'", 1)
+                if len(parts) >= 1:
+                    message = parts[0]
+
+            return {"_metadata": "finish", "message": message}
+
+        # Legacy/Fallback behavior
         if response.startswith('do(action="Type"') or response.startswith(
             'do(action="Type_Name"'
         ):
-            text = response.split("text=", 1)[1][1:-2]
-            action = {"_metadata": "do", "action": "Type", "text": text}
-            return action
-        elif response.startswith("do"):
             # Use AST parsing instead of eval for safety
             try:
                 # Escape special characters (newlines, tabs, etc.) for valid Python syntax
-                response = response.replace('\n', '\\n')
-                response = response.replace('\r', '\\r')
-                response = response.replace('\t', '\\t')
+                response = response.replace("\n", "\\n")
+                response = response.replace("\r", "\\r")
+                response = response.replace("\t", "\\t")
 
                 tree = ast.parse(response, mode="eval")
                 if not isinstance(tree.body, ast.Call):
@@ -381,10 +462,16 @@ def parse_action(response: str) -> dict[str, Any]:
                 "message": response.replace("finish(message=", "")[1:-2],
             }
         else:
-            raise ValueError(f"Failed to parse action: {response}")
+            # If all parsing attempts fail, treat the entire response as a message for 'finish'
+            # This is more robust than crashing with a ValueError
+            action = {"_metadata": "finish", "message": response}
         return action
     except Exception as e:
-        raise ValueError(f"Failed to parse action: {e}")
+        # Final fallback: return the original response if possible
+        return {
+            "_metadata": "finish",
+            "message": f"Parsing failed: {str(e)}. Raw response: {response}",
+        }
 
 
 def do(**kwargs) -> dict[str, Any]:
diff --git a/phone_agent/actions/handler_ios.py b/phone_agent/actions/handler_ios.py
index c37f50d9..11c977b3 100644
--- a/phone_agent/actions/handler_ios.py
+++ b/phone_agent/actions/handler_ios.py
@@ -129,9 +129,7 @@ def _handle_launch(self, action: dict, width: int, height: int) -> ActionResult:
         if not app_name:
             return ActionResult(False, False, "No app name specified")
 
-        success = launch_app(
-            app_name, wda_url=self.wda_url, session_id=self.session_id
-        )
+        success = launch_app(app_name, wda_url=self.wda_url, session_id=self.session_id)
         if success:
             return ActionResult(True, False)
         return ActionResult(False, False, f"App not found: {app_name}")
diff --git a/phone_agent/adb/connection.py b/phone_agent/adb/connection.py
index b723ca91..a132821f 100644
--- a/phone_agent/adb/connection.py
+++ b/phone_agent/adb/connection.py
@@ -109,7 +109,9 @@ def disconnect(self, address: str | None = None) -> tuple[bool, str]:
             if address:
                 cmd.append(address)
 
-            result = subprocess.run(cmd, capture_output=True, text=True, encoding="utf-8", timeout=5)
+            result = subprocess.run(
+                cmd, capture_output=True, text=True, encoding="utf-8", timeout=5
+            )
 
             output = result.stdout + result.stderr
             return True, output.strip() or "Disconnected"
@@ -241,7 +243,9 @@ def enable_tcpip(
                 cmd.extend(["-s", device_id])
             cmd.extend(["tcpip", str(port)])
 
-            result = subprocess.run(cmd, capture_output=True, text=True, encoding="utf-8", timeout=10)
+            result = subprocess.run(
+                cmd, capture_output=True, text=True, encoding="utf-8", timeout=10
+            )
 
             output = result.stdout + result.stderr
 
@@ -270,7 +274,9 @@ def get_device_ip(self, device_id: str | None = None) -> str | None:
                 cmd.extend(["-s", device_id])
             cmd.extend(["shell", "ip", "route"])
 
-            result = subprocess.run(cmd, capture_output=True, text=True, encoding="utf-8", timeout=5)
+            result = subprocess.run(
+                cmd, capture_output=True, text=True, encoding="utf-8", timeout=5
+            )
 
             # Parse IP from route output
             for line in result.stdout.split("\n"):
diff --git a/phone_agent/adb/device.py b/phone_agent/adb/device.py
index 995336a1..81a21616 100644
--- a/phone_agent/adb/device.py
+++ b/phone_agent/adb/device.py
@@ -22,7 +22,10 @@ def get_current_app(device_id: str | None = None) -> str:
     adb_prefix = _get_adb_prefix(device_id)
 
     result = subprocess.run(
-        adb_prefix + ["shell", "dumpsys", "window"], capture_output=True, text=True, encoding="utf-8"
+        adb_prefix + ["shell", "dumpsys", "window"],
+        capture_output=True,
+        text=True,
+        encoding="utf-8",
     )
     output = result.stdout
     if not output:
diff --git a/phone_agent/agent.py b/phone_agent/agent.py
index 36427917..d65405a8 100644
--- a/phone_agent/agent.py
+++ b/phone_agent/agent.py
@@ -1,8 +1,10 @@
 """Main PhoneAgent class for orchestrating phone automation."""
 
 import json
+import os
 import traceback
 from dataclasses import dataclass
+from datetime import datetime
 from typing import Any, Callable
 
 from phone_agent.actions import ActionHandler
@@ -171,9 +173,6 @@ def _execute_step(
         # Get model response
         try:
             msgs = get_messages(self.agent_config.lang)
-            print("\n" + "=" * 50)
-            print(f"💭 {msgs['thinking']}:")
-            print("-" * 50)
             response = self.model_client.request(self._context)
         except Exception as e:
             if self.agent_config.verbose:
@@ -195,8 +194,7 @@ def _execute_step(
             action = finish(message=response.action)
 
         if self.agent_config.verbose:
-            # Print thinking process
-            print("-" * 50)
+            # Print parsed action
             print(f"🎯 {msgs['action']}:")
             print(json.dumps(action, ensure_ascii=False, indent=2))
             print("=" * 50 + "\n")
diff --git a/phone_agent/agent_ios.py b/phone_agent/agent_ios.py
index a3b20d9f..7cbe9cec 100644
--- a/phone_agent/agent_ios.py
+++ b/phone_agent/agent_ios.py
@@ -20,7 +20,7 @@ class IOSAgentConfig:
     max_steps: int = 100
     wda_url: str = "http://localhost:8100"
     session_id: str | None = None
-    device_id: str | None = None  # iOS device UDID
+    device_id: str | None = None  # iOS device UUID
     lang: str = "cn"
     system_prompt: str | None = None
     verbose: bool = True
@@ -214,13 +214,8 @@ def _execute_step(
             action = finish(message=response.action)
 
         if self.agent_config.verbose:
-            # Print thinking process
+            # Print parsed action
             msgs = get_messages(self.agent_config.lang)
-            print("\n" + "=" * 50)
-            print(f"💭 {msgs['thinking']}:")
-            print("-" * 50)
-            print(response.thinking)
-            print("-" * 50)
             print(f"🎯 {msgs['action']}:")
             print(json.dumps(action, ensure_ascii=False, indent=2))
             print("=" * 50 + "\n")
diff --git a/phone_agent/config/apps.py b/phone_agent/config/apps.py
index 874592da..0db93e15 100644
--- a/phone_agent/config/apps.py
+++ b/phone_agent/config/apps.py
@@ -29,7 +29,7 @@
     "12306": "com.MobileTicket",
     "去哪儿": "com.Qunar",
     "去哪儿旅行": "com.Qunar",
-    "滴滴出行": "com.sdu.didi.psnger",
+    "滴滴出行": "com.sdu.did.psnger",
     # Video & Entertainment
     "bilibili": "tv.danmaku.bili",
     "抖音": "com.ss.android.ugc.aweme",
@@ -224,4 +224,4 @@ def list_supported_apps() -> list[str]:
     Returns:
         List of app names.
     """
-    return list(APP_PACKAGES.keys())
\ No newline at end of file
+    return list(APP_PACKAGES.keys())
diff --git a/phone_agent/config/apps_harmonyos.py b/phone_agent/config/apps_harmonyos.py
index 5ffb52c5..76018a7f 100644
--- a/phone_agent/config/apps_harmonyos.py
+++ b/phone_agent/config/apps_harmonyos.py
@@ -21,7 +21,6 @@
     "com.tencent.videohm": "AppAbility",
     "com.ximalaya.ting.xmharmony": "MainBundleAbility",
     "com.zhihu.hmos": "PhoneAbility",
-
     # Huawei system apps
     "com.huawei.hmos.browser": "MainAbility",
     "com.huawei.hmos.calculator": "com.huawei.hmos.calculator.CalculatorAbility",
@@ -40,7 +39,6 @@
     "com.huawei.hmos.soundrecorder": "MainAbility",
     "com.huawei.hmos.vassistant": "AiCaptionServiceExtAbility",
     "com.huawei.hmos.wallet": "MainAbility",
-
     # Huawei services
     "com.huawei.hmsapp.appgallery": "MainAbility",
     "com.huawei.hmsapp.books": "MainAbility",
@@ -49,7 +47,6 @@
     "com.huawei.hmsapp.music": "MainAbility",
     "com.huawei.hmsapp.thememanager": "MainAbility",
     "com.huawei.hmsapp.totemweather": "com.huawei.hmsapp.totemweather.MainAbility",
-
     # OHOS system apps
     "com.ohos.callui": "com.ohos.callui.ServiceAbility",
     "com.ohos.contacts": "com.ohos.contacts.MainAbility",
@@ -85,7 +82,7 @@
     "12306": "com.chinarailway.ticketingHM",
     # "去哪儿": "com.Qunar",  # 未在 hdc 列表中找到
     # "去哪儿旅行": "com.Qunar",  # 未在 hdc 列表中找到
-    "滴滴出行": "com.sdu.didi.hmos.psnger",
+    "滴滴出行": "com.sdu.did.hmos.psnger",
     # Video & Entertainment
     "bilibili": "yylx.danmaku.bili",
     "抖音": "com.ss.hm.ugc.aweme",
@@ -124,7 +121,6 @@
     # "星穹铁道": "com.miHoYo.hkrpg",  # 未在 hdc 列表中找到
     # "崩坏：星穹铁道": "com.miHoYo.hkrpg",  # 未在 hdc 列表中找到
     # "恋与深空": "com.papegames.lysk.cn",  # 未在 hdc 列表中找到
-    
     # HarmonyOS 第三方应用
     "百度": "com.baidu.baiduapp",
     "阿里巴巴": "com.alibaba.wireless_hmos",
@@ -149,7 +145,6 @@
     "国家税务总局": "cn.gov.chinatax.gt4.hm",
     "建设银行": "com.ccb.mobilebank.hm",
     "快手极速版": "com.kuaishou.hmnebula",
-    
     # HarmonyOS 系统应用 - 工具类
     "浏览器": "com.huawei.hmos.browser",
     "计算器": "com.huawei.hmos.calculator",
@@ -169,12 +164,10 @@
     "截屏": "com.huawei.hmos.screenshot",
     "笔记": "com.huawei.hmos.notepad",
     "备忘录": "com.huawei.hmos.notepad",
-    
     # HarmonyOS 系统应用 - 媒体类
     "相册": "com.huawei.hmos.photos",
     "图库": "com.huawei.hmos.photos",
     # "视频": "com.huawei.hmos.mediaplayer",  # 未在 hdc 列表中找到，但有 com.huawei.hmsapp.himovie
-    
     # HarmonyOS 系统应用 - 通讯类
     "联系人": "com.ohos.contacts",
     "通讯录": "com.ohos.contacts",
@@ -182,7 +175,6 @@
     "信息": "com.ohos.mms",
     "电话": "com.ohos.callui",
     "拨号": "com.ohos.callui",
-    
     # HarmonyOS 系统应用 - 设置类
     "设置": "com.huawei.hmos.settings",
     "系统设置": "com.huawei.hmos.settings",
@@ -191,7 +183,6 @@
     "Android  System Settings": "com.huawei.hmos.settings",
     "Android-System-Settings": "com.huawei.hmos.settings",
     "Settings": "com.huawei.hmos.settings",
-    
     # HarmonyOS 系统应用 - 生活服务
     "健康": "com.huawei.hmos.health",
     "运动健康": "com.huawei.hmos.health",
@@ -202,7 +193,6 @@
     "智慧生活": "com.huawei.hmos.ailife",
     "智能助手": "com.huawei.hmos.vassistant",
     "小艺": "com.huawei.hmos.vassistant",
-    
     # HarmonyOS 服务
     "应用市场": "com.huawei.hmsapp.appgallery",
     "华为应用市场": "com.huawei.hmsapp.appgallery",
diff --git a/phone_agent/hdc/__init__.py b/phone_agent/hdc/__init__.py
index 9b06993f..59c68d17 100644
--- a/phone_agent/hdc/__init__.py
+++ b/phone_agent/hdc/__init__.py
@@ -1,9 +1,9 @@
 """HDC utilities for HarmonyOS device interaction."""
 
 from phone_agent.hdc.connection import (
-    HDCConnection,
     ConnectionType,
     DeviceInfo,
+    HDCConnection,
     list_devices,
     quick_connect,
     set_hdc_verbose,
diff --git a/phone_agent/hdc/connection.py b/phone_agent/hdc/connection.py
index 15809f84..d1a29899 100644
--- a/phone_agent/hdc/connection.py
+++ b/phone_agent/hdc/connection.py
@@ -9,7 +9,6 @@
 
 from phone_agent.config.timing import TIMING_CONFIG
 
-
 # Global flag to control HDC command output
 _HDC_VERBOSE = os.getenv("HDC_VERBOSE", "false").lower() in ("true", "1", "yes")
 
@@ -32,7 +31,7 @@ def _run_hdc_command(cmd: list, **kwargs) -> subprocess.CompletedProcess:
 
     if _HDC_VERBOSE and result.returncode != 0:
         print(f"[HDC] Command failed with return code {result.returncode}")
-        if hasattr(result, 'stderr') and result.stderr:
+        if hasattr(result, "stderr") and result.stderr:
             print(f"[HDC] Error: {result.stderr}")
 
     return result
@@ -150,11 +149,13 @@ def disconnect(self, address: str | None = None) -> tuple[bool, str]:
                             [self.hdc_path, "tdisconn", device.device_id],
                             capture_output=True,
                             text=True,
-                            timeout=5
+                            timeout=5,
                         )
                 return True, "Disconnected all remote devices"
 
-            result = _run_hdc_command(cmd, capture_output=True, text=True, encoding="utf-8", timeout=5)
+            result = _run_hdc_command(
+                cmd, capture_output=True, text=True, encoding="utf-8", timeout=5
+            )
 
             output = result.stdout + result.stderr
             return True, output.strip() or "Disconnected"
@@ -278,7 +279,9 @@ def enable_tcpip(
                 cmd.extend(["-t", device_id])
             cmd.extend(["tmode", "port", str(port)])
 
-            result = _run_hdc_command(cmd, capture_output=True, text=True, encoding="utf-8", timeout=10)
+            result = _run_hdc_command(
+                cmd, capture_output=True, text=True, encoding="utf-8", timeout=10
+            )
 
             output = result.stdout + result.stderr
 
@@ -307,7 +310,9 @@ def get_device_ip(self, device_id: str | None = None) -> str | None:
                 cmd.extend(["-t", device_id])
             cmd.extend(["shell", "ifconfig"])
 
-            result = _run_hdc_command(cmd, capture_output=True, text=True, encoding="utf-8", timeout=5)
+            result = _run_hdc_command(
+                cmd, capture_output=True, text=True, encoding="utf-8", timeout=5
+            )
 
             # Parse IP from ifconfig output
             for line in result.stdout.split("\n"):
@@ -339,9 +344,7 @@ def restart_server(self) -> tuple[bool, str]:
         """
         try:
             # Kill server
-            _run_hdc_command(
-                [self.hdc_path, "kill"], capture_output=True, timeout=5
-            )
+            _run_hdc_command([self.hdc_path, "kill"], capture_output=True, timeout=5)
 
             time.sleep(TIMING_CONFIG.connection.server_restart_delay)
 
diff --git a/phone_agent/hdc/device.py b/phone_agent/hdc/device.py
index 9d693b42..dcef40a0 100644
--- a/phone_agent/hdc/device.py
+++ b/phone_agent/hdc/device.py
@@ -1,6 +1,7 @@
 """Device control utilities for HarmonyOS automation."""
 
 import os
+import re
 import subprocess
 import time
 from typing import List, Optional, Tuple
@@ -8,7 +9,7 @@
 from phone_agent.config.apps_harmonyos import APP_ABILITIES, APP_PACKAGES
 from phone_agent.config.timing import TIMING_CONFIG
 from phone_agent.hdc.connection import _run_hdc_command
-import re
+
 
 def get_current_app(device_id: str | None = None) -> str:
     """
@@ -27,7 +28,7 @@ def get_current_app(device_id: str | None = None) -> str:
         hdc_prefix + ["shell", "aa", "dump", "-l"],
         capture_output=True,
         text=True,
-        encoding="utf-8"
+        encoding="utf-8",
     )
     output = result.stdout
     # print(output)
@@ -51,7 +52,7 @@ def get_current_app(device_id: str | None = None) -> str:
     for line in lines:
         # Track the current mission's bundle name
         if "app name [" in line:
-            match = re.search(r'\[([^\]]+)\]', line)
+            match = re.search(r"\[([^\]]+)\]", line)
             if match:
                 current_bundle = match.group(1)
 
@@ -71,9 +72,9 @@ def get_current_app(device_id: str | None = None) -> str:
             if package == foreground_bundle:
                 return app_name
         # If bundle is found but not in our known apps, return the bundle name
-        print(f'Bundle is found but not in our known apps: {foreground_bundle}')
+        print(f"Bundle is found but not in our known apps: {foreground_bundle}")
         return foreground_bundle
-    print(f'No bundle is found')
+    print(f"No bundle is found")
     return "System Home"
 
 
@@ -97,7 +98,7 @@ def tap(
     # HarmonyOS uses uitest uiInput click
     _run_hdc_command(
         hdc_prefix + ["shell", "uitest", "uiInput", "click", str(x), str(y)],
-        capture_output=True
+        capture_output=True,
     )
     time.sleep(delay)
 
@@ -122,7 +123,7 @@ def double_tap(
     # HarmonyOS uses uitest uiInput doubleClick
     _run_hdc_command(
         hdc_prefix + ["shell", "uitest", "uiInput", "doubleClick", str(x), str(y)],
-        capture_output=True
+        capture_output=True,
     )
     time.sleep(delay)
 
@@ -226,7 +227,7 @@ def back(device_id: str | None = None, delay: float | None = None) -> None:
     # HarmonyOS uses uitest uiInput keyEvent Back
     _run_hdc_command(
         hdc_prefix + ["shell", "uitest", "uiInput", "keyEvent", "Back"],
-        capture_output=True
+        capture_output=True,
     )
     time.sleep(delay)
 
@@ -247,7 +248,7 @@ def home(device_id: str | None = None, delay: float | None = None) -> None:
     # HarmonyOS uses uitest uiInput keyEvent Home
     _run_hdc_command(
         hdc_prefix + ["shell", "uitest", "uiInput", "keyEvent", "Home"],
-        capture_output=True
+        capture_output=True,
     )
     time.sleep(delay)
 
@@ -306,5 +307,6 @@ def _get_hdc_prefix(device_id: str | None) -> list:
         return ["hdc", "-t", device_id]
     return ["hdc"]
 
+
 if __name__ == "__main__":
     print(get_current_app())
diff --git a/phone_agent/hdc/input.py b/phone_agent/hdc/input.py
index 920cf7dd..45a7d2ce 100644
--- a/phone_agent/hdc/input.py
+++ b/phone_agent/hdc/input.py
@@ -25,8 +25,8 @@ def type_text(text: str, device_id: str | None = None) -> None:
     hdc_prefix = _get_hdc_prefix(device_id)
 
     # Handle multi-line text by splitting on newlines
-    if '\n' in text:
-        lines = text.split('\n')
+    if "\n" in text:
+        lines = text.split("\n")
         for i, line in enumerate(lines):
             if line:  # Only process non-empty lines
                 # Escape special characters for shell
diff --git a/phone_agent/hdc/screenshot.py b/phone_agent/hdc/screenshot.py
index 332d198c..d0d2a437 100644
--- a/phone_agent/hdc/screenshot.py
+++ b/phone_agent/hdc/screenshot.py
@@ -10,6 +10,7 @@
 from typing import Tuple
 
 from PIL import Image
+
 from phone_agent.hdc.connection import _run_hdc_command
 
 
@@ -56,7 +57,11 @@ def get_screenshot(device_id: str | None = None, timeout: int = 10) -> Screensho
 
         # Check for screenshot failure (sensitive screen)
         output = result.stdout + result.stderr
-        if "fail" in output.lower() or "error" in output.lower() or "not found" in output.lower():
+        if (
+            "fail" in output.lower()
+            or "error" in output.lower()
+            or "not found" in output.lower()
+        ):
             # Try method 2: snapshot_display (older versions or different devices)
             result = _run_hdc_command(
                 hdc_prefix + ["shell", "snapshot_display", "-f", remote_path],
diff --git a/phone_agent/model/client.py b/phone_agent/model/client.py
index 72377a61..f4e4c485 100644
--- a/phone_agent/model/client.py
+++ b/phone_agent/model/client.py
@@ -81,9 +81,20 @@ def request(self, messages: list[dict[str, Any]]) -> ModelResponse:
 
         raw_content = ""
         buffer = ""  # Buffer to hold content that might be part of a marker
-        action_markers = ["finish(message=", "do(action="]
+
+        # Action markers that signal the end of the thinking process
+        action_markers = ["<answer>", "finish(message=", "do(action="]
+        # Tags to hide from the output stream
+        hide_tags = ["<think>", "</think>"]
+
         in_action_phase = False  # Track if we've entered the action phase
         first_token_received = False
+        lang = self.config.lang
+
+        # Print thinking header
+        print("\n" + "=" * 50)
+        print(f"💭 {get_message('thinking', lang)}:")
+        print("-" * 50)
 
         for chunk in stream:
             if len(chunk.choices) == 0:
@@ -103,7 +114,12 @@ def request(self, messages: list[dict[str, Any]]) -> ModelResponse:
 
                 buffer += content
 
-                # Check if any marker is fully present in buffer
+                # Hide tags from stream
+                for tag in hide_tags:
+                    if tag in buffer:
+                        buffer = buffer.replace(tag, "")
+
+                # Check if any action marker is fully present in buffer
                 marker_found = False
                 for marker in action_markers:
                     if marker in buffer:
@@ -123,10 +139,10 @@ def request(self, messages: list[dict[str, Any]]) -> ModelResponse:
                 if marker_found:
                     continue  # Continue to collect remaining content
 
-                # Check if buffer ends with a prefix of any marker
+                # Check if buffer ends with a prefix of any marker or tag
                 # If so, don't print yet (wait for more content)
                 is_potential_marker = False
-                for marker in action_markers:
+                for marker in action_markers + hide_tags:
                     for i in range(1, len(marker)):
                         if buffer.endswith(marker[:i]):
                             is_potential_marker = True
@@ -147,7 +163,6 @@ def request(self, messages: list[dict[str, Any]]) -> ModelResponse:
 
         # Print performance metrics
         lang = self.config.lang
-        print()
         print("=" * 50)
         print(f"⏱️  {get_message('performance_metrics', lang)}:")
         print("-" * 50)
@@ -178,11 +193,11 @@ def _parse_response(self, content: str) -> tuple[str, str]:
         Parse the model response into thinking and action parts.
 
         Parsing rules:
-        1. If content contains 'finish(message=', everything before is thinking,
+        1. If content contains XML tags <think> and <answer>, use them.
+        2. If content contains 'finish(message=', everything before is thinking,
            everything from 'finish(message=' onwards is action.
-        2. If rule 1 doesn't apply but content contains 'do(action=',
-           everything before is thinking, everything from 'do(action=' onwards is action.
-        3. Fallback: If content contains '<answer>', use legacy parsing with XML tags.
+        3. If content contains 'do(action=', everything before is thinking,
+           everything from 'do(action=' onwards is action.
         4. Otherwise, return empty thinking and full content as action.
 
         Args:
@@ -191,25 +206,45 @@ def _parse_response(self, content: str) -> tuple[str, str]:
         Returns:
             Tuple of (thinking, action).
         """
-        # Rule 1: Check for finish(message=
+        # Rule 1: XML tag parsing (highest priority if tags exist)
+        if "<answer>" in content:
+            thinking = ""
+            if "<think>" in content and "</think>" in content:
+                thinking_parts = content.split("<think>", 1)[1].split("</think>", 1)
+                thinking = thinking_parts[0].strip()
+            elif "<think>" in content:
+                thinking = (
+                    content.split("<think>", 1)[1].split("<answer>", 1)[0].strip()
+                )
+            else:
+                thinking = content.split("<answer>", 1)[0].strip()
+
+            action_part = content.split("<answer>", 1)[1]
+            if "</answer>" in action_part:
+                action = action_part.split("</answer>", 1)[0].strip()
+            else:
+                action = action_part.strip()
+
+            return thinking, action
+
+        # Rule 2: Check for finish(message=
         if "finish(message=" in content:
             parts = content.split("finish(message=", 1)
             thinking = parts[0].strip()
             action = "finish(message=" + parts[1]
+            # Clean up trailing tags if any
+            if "</answer>" in action:
+                action = action.split("</answer>", 1)[0].strip()
             return thinking, action
 
-        # Rule 2: Check for do(action=
+        # Rule 3: Check for do(action=
         if "do(action=" in content:
             parts = content.split("do(action=", 1)
             thinking = parts[0].strip()
             action = "do(action=" + parts[1]
-            return thinking, action
-
-        # Rule 3: Fallback to legacy XML tag parsing
-        if "<answer>" in content:
-            parts = content.split("<answer>", 1)
-            thinking = parts[0].replace("<think>", "").replace("</think>", "").strip()
-            action = parts[1].replace("</answer>", "").strip()
+            # Clean up trailing tags if any
+            if "</answer>" in action:
+                action = action.split("</answer>", 1)[0].strip()
             return thinking, action
 
         # Rule 4: No markers found, return content as action
diff --git a/phone_agent/xctest/connection.py b/phone_agent/xctest/connection.py
index deb29369..8a5373f2 100644
--- a/phone_agent/xctest/connection.py
+++ b/phone_agent/xctest/connection.py
@@ -17,7 +17,7 @@ class ConnectionType(Enum):
 class DeviceInfo:
     """Information about a connected iOS device."""
 
-    device_id: str  # UDID
+    device_id: str  # UUID
     status: str
     connection_type: ConnectionType
     model: str | None = None
@@ -76,23 +76,23 @@ def list_devices(self) -> list[DeviceInfo]:
 
             devices = []
             for line in result.stdout.strip().split("\n"):
-                udid = line.strip()
-                if not udid:
+                uuid = line.strip()
+                if not uuid:
                     continue
 
                 # Determine connection type (network devices have specific format)
                 conn_type = (
                     ConnectionType.NETWORK
-                    if "-" in udid and len(udid) > 40
+                    if "-" in uuid and len(uuid) > 40
                     else ConnectionType.USB
                 )
 
                 # Get detailed device info
-                device_info = self._get_device_details(udid)
+                device_info = self._get_device_details(uuid)
 
                 devices.append(
                     DeviceInfo(
-                        device_id=udid,
+                        device_id=uuid,
                         status="connected",
                         connection_type=conn_type,
                         model=device_info.get("model"),
@@ -112,19 +112,19 @@ def list_devices(self) -> list[DeviceInfo]:
             print(f"Error listing devices: {e}")
             return []
 
-    def _get_device_details(self, udid: str) -> dict[str, str]:
+    def _get_device_details(self, uuid: str) -> dict[str, str]:
         """
         Get detailed information about a specific device.
 
         Args:
-            udid: Device UDID.
+            uuid: Device UUID.
 
         Returns:
             Dictionary with device details.
         """
         try:
             result = subprocess.run(
-                ["ideviceinfo", "-u", udid],
+                ["ideviceinfo", "-u", uuid],
                 capture_output=True,
                 text=True,
                 timeout=5,
@@ -154,7 +154,7 @@ def get_device_info(self, device_id: str | None = None) -> DeviceInfo | None:
         Get detailed information about a device.
 
         Args:
-            device_id: Device UDID. If None, uses first available device.
+            device_id: Device UUID. If None, uses first available device.
 
         Returns:
             DeviceInfo or None if not found.
@@ -178,7 +178,7 @@ def is_connected(self, device_id: str | None = None) -> bool:
         Check if a device is connected.
 
         Args:
-            device_id: Device UDID to check. If None, checks if any device is connected.
+            device_id: Device UUID to check. If None, checks if any device is connected.
 
         Returns:
             True if connected, False otherwise.
@@ -211,9 +211,7 @@ def is_wda_ready(self, timeout: int = 2) -> bool:
             )
             return response.status_code == 200
         except ImportError:
-            print(
-                "Error: requests library not found. Install it: pip install requests"
-            )
+            print("Error: requests library not found. Install it: pip install requests")
             return False
         except Exception:
             return False
@@ -276,7 +274,7 @@ def pair_device(self, device_id: str | None = None) -> tuple[bool, str]:
         Pair with an iOS device (required for some operations).
 
         Args:
-            device_id: Device UDID. If None, uses first available device.
+            device_id: Device UUID. If None, uses first available device.
 
         Returns:
             Tuple of (success, message).
@@ -309,7 +307,7 @@ def get_device_name(self, device_id: str | None = None) -> str | None:
         Get the device name.
 
         Args:
-            device_id: Device UDID. If None, uses first available device.
+            device_id: Device UUID. If None, uses first available device.
 
         Returns:
             Device name string or None if not found.
diff --git a/phone_agent/xctest/device.py b/phone_agent/xctest/device.py
index 49fc379c..a4be177a 100644
--- a/phone_agent/xctest/device.py
+++ b/phone_agent/xctest/device.py
@@ -6,7 +6,8 @@
 
 from phone_agent.config.apps_ios import APP_PACKAGES_IOS as APP_PACKAGES
 
-SCALE_FACTOR = 3 # 3 for most modern iPhone 
+SCALE_FACTOR = 3  # 3 for most modern iPhone
+
 
 def _get_wda_session_url(wda_url: str, session_id: str | None, endpoint: str) -> str:
     """
@@ -102,7 +103,12 @@ def tap(
                     "id": "finger1",
                     "parameters": {"pointerType": "touch"},
                     "actions": [
-                        {"type": "pointerMove", "duration": 0, "x": x / SCALE_FACTOR, "y": y / SCALE_FACTOR},
+                        {
+                            "type": "pointerMove",
+                            "duration": 0,
+                            "x": x / SCALE_FACTOR,
+                            "y": y / SCALE_FACTOR,
+                        },
                         {"type": "pointerDown", "button": 0},
                         {"type": "pause", "duration": 0.1},
                         {"type": "pointerUp", "button": 0},
@@ -151,7 +157,12 @@ def double_tap(
                     "id": "finger1",
                     "parameters": {"pointerType": "touch"},
                     "actions": [
-                        {"type": "pointerMove", "duration": 0, "x": x / SCALE_FACTOR, "y": y / SCALE_FACTOR},
+                        {
+                            "type": "pointerMove",
+                            "duration": 0,
+                            "x": x / SCALE_FACTOR,
+                            "y": y / SCALE_FACTOR,
+                        },
                         {"type": "pointerDown", "button": 0},
                         {"type": "pause", "duration": 100},
                         {"type": "pointerUp", "button": 0},
@@ -209,7 +220,12 @@ def long_press(
                     "id": "finger1",
                     "parameters": {"pointerType": "touch"},
                     "actions": [
-                        {"type": "pointerMove", "duration": 0, "x": x / SCALE_FACTOR, "y": y / SCALE_FACTOR},
+                        {
+                            "type": "pointerMove",
+                            "duration": 0,
+                            "x": x / SCALE_FACTOR,
+                            "y": y / SCALE_FACTOR,
+                        },
                         {"type": "pointerDown", "button": 0},
                         {"type": "pause", "duration": duration_ms},
                         {"type": "pointerUp", "button": 0},
diff --git a/phone_agent/xctest/input.py b/phone_agent/xctest/input.py
index 9c8d8656..52d5733f 100644
--- a/phone_agent/xctest/input.py
+++ b/phone_agent/xctest/input.py
@@ -49,11 +49,16 @@ def type_text(
 
         # Send text to WDA
         response = requests.post(
-            url, json={"value": list(text), "frequency": frequency}, timeout=30, verify=False
+            url,
+            json={"value": list(text), "frequency": frequency},
+            timeout=30,
+            verify=False,
         )
 
         if response.status_code not in (200, 201):
-            print(f"Warning: Text input may have failed. Status: {response.status_code}")
+            print(
+                f"Warning: Text input may have failed. Status: {response.status_code}"
+            )
 
     except ImportError:
         print("Error: requests library required. Install: pip install requests")
@@ -86,11 +91,15 @@ def clear_text(
 
         if response.status_code == 200:
             data = response.json()
-            element_id = data.get("value", {}).get("ELEMENT") or data.get("value", {}).get("element-6066-11e4-a52e-4f735466cecf")
+            element_id = data.get("value", {}).get("ELEMENT") or data.get(
+                "value", {}
+            ).get("element-6066-11e4-a52e-4f735466cecf")
 
             if element_id:
                 # Clear the element
-                clear_url = _get_wda_session_url(wda_url, session_id, f"element/{element_id}/clear")
+                clear_url = _get_wda_session_url(
+                    wda_url, session_id, f"element/{element_id}/clear"
+                )
                 requests.post(clear_url, timeout=10, verify=False)
                 return
 
@@ -259,7 +268,10 @@ def set_pasteboard(
         url = f"{wda_url.rstrip('/')}/wda/setPasteboard"
 
         requests.post(
-            url, json={"content": text, "contentType": "plaintext"}, timeout=10, verify=False
+            url,
+            json={"content": text, "contentType": "plaintext"},
+            timeout=10,
+            verify=False,
         )
 
     except ImportError:
diff --git a/phone_agent/xctest/screenshot.py b/phone_agent/xctest/screenshot.py
index bbf4bdd6..c52351d3 100644
--- a/phone_agent/xctest/screenshot.py
+++ b/phone_agent/xctest/screenshot.py
@@ -33,7 +33,7 @@ def get_screenshot(
     Args:
         wda_url: WebDriverAgent URL.
         session_id: Optional WDA session ID.
-        device_id: Optional device UDID (for idevicescreenshot fallback).
+        device_id: Optional device UUID (for idevicescreenshot fallback).
         timeout: Timeout in seconds for screenshot operations.
 
     Returns:
@@ -103,14 +103,12 @@ def _get_screenshot_wda(
     return None
 
 
-def _get_screenshot_idevice(
-    device_id: str | None, timeout: int
-) -> Screenshot | None:
+def _get_screenshot_idevice(device_id: str | None, timeout: int) -> Screenshot | None:
     """
     Capture screenshot using idevicescreenshot (libimobiledevice).
 
     Args:
-        device_id: Optional device UDID.
+        device_id: Optional device UUID.
         timeout: Timeout in seconds.
 
     Returns:
@@ -126,9 +124,7 @@ def _get_screenshot_idevice(
             cmd.extend(["-u", device_id])
         cmd.append(temp_path)
 
-        result = subprocess.run(
-            cmd, capture_output=True, text=True, timeout=timeout
-        )
+        result = subprocess.run(cmd, capture_output=True, text=True, timeout=timeout)
 
         if result.returncode == 0 and os.path.exists(temp_path):
             # Read and encode image
@@ -217,7 +213,7 @@ def get_screenshot_png(
     Args:
         wda_url: WebDriverAgent URL.
         session_id: Optional WDA session ID.
-        device_id: Optional device UDID.
+        device_id: Optional device UUID.
 
     Returns:
         PNG bytes or None if failed.