'commit'

WeiXin/ChatMonitorAll.py

@@ -0,0 +1,361 @@
+# coding=utf-8
+import os
+import sys
+import logging
+import asyncio
+import hashlib
+import json
+import numpy as np
+
+import cv2
+
+# 添加项目根目录到 sys.path
+project_root = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+if project_root not in sys.path:
+    sys.path.append(project_root)
+
+from WeiXin import WxUtil
+from WeiXin.WxUtil import perform_input_action
+from Util.LlmUtil import get_llm_response
+from Util import Win32Patch
+
+# 配置日志
+log_dir = WxUtil.LOG_DIR
+if not os.path.exists(log_dir):
+    os.makedirs(log_dir)
+
+log_file_path = os.path.join(log_dir, "T2_ChatMonitor.log")
+
+# 设置 logger
+logger = logging.getLogger("T2_ChatMonitor")
+logger.setLevel(logging.INFO)
+
+if logger.hasHandlers():
+    logger.handlers.clear()
+
+file_handler = logging.FileHandler(log_file_path, encoding='utf-8', mode='w')
+file_handler.setFormatter(logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s'))
+logger.addHandler(file_handler)
+
+stream_handler = logging.StreamHandler()
+stream_handler.setFormatter(logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s'))
+logger.addHandler(stream_handler)
+
+logger.propagate = False
+logger.info(f"🚀 日志文件路径: {os.path.abspath(log_file_path)}")
+
+# 同时将 WxUtil 的日志也输出到同一个文件
+wx_logger = logging.getLogger("WxUtil")
+wx_logger.propagate = False  # 防止日志向上传递导致重复 (因为 WxUtil 中调用了 basicConfig)
+if not any(isinstance(h, logging.FileHandler) and os.path.abspath(h.baseFilename) == os.path.abspath(log_file_path) for h in wx_logger.handlers):
+    wx_logger.addHandler(file_handler)
+    wx_logger.addHandler(stream_handler) # 确保 WxUtil 也输出到控制台
+
+class ChatMonitorBot:
+    """
+    大张老师自动巡课系统 (CV版)
+    """
+    def __init__(self):
+        self.device = None
+        self.screenshot_path = os.path.join(WxUtil.OUTPUT_DIR, "T2_ChatMonitor_live_shot.jpg")
+        self.debug_view_path = os.path.join(WxUtil.OUTPUT_DIR, "T2_ChatMonitor_debug_view.jpg")
+        self.dialogue_log = []
+        self.input_pos = None
+        self.last_screen_hash = None
+        self.last_processed_msg_hash = None
+        # [User Requested] 移除持久化存储，只在内存中记录，重启即忘
+        self.processed_hashes = set()
+        # 新增：记录已处理消息的元数据 (sender, time_display, type) 用于防止空内容重试循环
+        self.processed_meta = set()
+        self.check_interval = 3  # 检查频率 (秒)
+        
+        self.persona = (
+            "你是一名1999年毕业、拥有27年一线教学经验的小学高级女教师，名叫‘大张老师’。你目前在‘长春市少惠林作文素养培养中心’工作。"
+            "你不仅是一位作文教学专家，更是一位心思细腻、能与家长共情的教育智者。"
+            "你的回复风格应该是：温柔、知性、亲切，就像一位邻家大姐姐在聊天。"
+            "【严格约束】：\n"
+            "1. 绝对禁止发散！绝对禁止幻觉！\n"
+            "2. 知道什么就说什么，不要乱讲话，不要自己编造内容！\n"
+            "3. 仅针对家长明确表达的内容进行回复。\n"
+            "4. 严禁使用列表格式。严禁使用‘首先、其次’等逻辑词。\n"
+            "5. 回复必须简练，字数严格控制在 50 字以内！\n"
+            "6. 对方问什么就答什么。例如问‘学校叫什么’，就只回答‘少惠林’，不要回复地址和电话！\n"
+            "如果涉及到校区信息，必须且只能使用以下真实数据：\n"
+            "- 单位/学校名称：长春市少惠林作文素养培养中心（简称：少惠林）\n"
+            "- 地址：南环城路与临河街交汇，TOUCH12街3楼325号\n"
+            "- 联系人：小张老师（电话：18686619970）\n"
+            "- 每学期开学招收小学三年级至六年级，初中七年级的学生入学，其它年段不招生。\n"
+        )
+
+    def _record_processed_hash(self, msg, msg_hash):
+        """记录已处理的消息哈希和元数据 (仅内存)"""
+        self.processed_hashes.add(msg_hash)
+        
+        # 记录元数据 (Sender, Time, Type)
+        if msg:
+            meta = (msg.get("sender", ""), msg.get("time_display", ""), msg.get("type", ""))
+            self.processed_meta.add(meta)
+
+        # 仅保留最近 100 条记录，防止无限增长
+        if len(self.processed_hashes) > 100:
+            # 简单丢弃旧的
+            temp = list(self.processed_hashes)[-100:]
+            self.processed_hashes = set(temp)
+        
+        if len(self.processed_meta) > 100:
+            temp_meta = list(self.processed_meta)[-100:]
+            self.processed_meta = set(temp_meta)
+
+    async def get_reply(self, last_message_text, context_text=""):
+        prompt = (
+            f"【教师人设】：{self.persona}\n\n"
+            f"【上下文对话内容】：\n{context_text}\n\n"
+            f"【最后一条待回复消息】：\n{last_message_text}\n\n"
+            "【任务要求】：\n"
+            "请作为大张老师回复家长。**必须且只能针对最后一条消息进行回复！**\n"
+            "参考上下文对话内容，确保回复逻辑连贯。\n"
+            "严禁发散，严禁编造家长没说过的情况。如果不清楚家长的意图，就温柔询问。\n"
+            "字数严格控制在 50 字以内。直接输出回复正文。"
+        )
+        
+        full_response = ""
+        async for chunk in get_llm_response(prompt, stream=False):
+            full_response += chunk
+        return full_response.strip().strip('"').strip('“').strip('”')
+
+    def step_1_prepare_env(self):
+        """步骤1: 环境准备"""
+        logger.info("--- [Step 1] 环境准备 ---")
+        WxUtil.setup_script_environment()
+        return True
+
+    def step_2_connect_device(self):
+        """步骤2: 连接设备"""
+        logger.info("--- [Step 2] 连接设备 ---")
+        self.device = WxUtil.connect_device()
+        if not self.device:
+            logger.error("❌ 设备连接失败，请检查手机是否连接且开启了调试模式")
+            return False
+        return True
+
+    def get_image_hash(self, file_path):
+        """计算图片的 MD5 哈希值 (忽略顶部 100 像素的状态栏)"""
+        if not os.path.exists(file_path):
+            return None
+        try:
+            # 使用 OpenCV 读取图片
+            img = cv2.imread(file_path)
+            if img is None:
+                # 如果读取失败，回退到文件哈希
+                with open(file_path, "rb") as f:
+                    return hashlib.md5(f.read()).hexdigest()
+            
+            # 裁剪掉顶部 150 像素 (状态栏/时间)
+            h, w = img.shape[:2]
+            if h > 150:
+                cropped_img = img[150:h, 0:w]
+            else:
+                cropped_img = img
+                
+            # 计算裁剪后数据的哈希
+            return hashlib.md5(cropped_img.tobytes()).hexdigest()
+        except Exception as e:
+            logger.error(f"计算哈希出错: {e}, 回退到文件哈希")
+            with open(file_path, "rb") as f:
+                return hashlib.md5(f.read()).hexdigest()
+
+    def get_stable_message_hash(self, msg):
+        """
+        计算消息的稳定哈希值（忽略坐标等易变字段）
+        仅包含: sender, content, time_display, type
+        """
+        if not msg:
+            return ""
+        
+        stable_data = {
+            "sender": msg.get("sender", ""),
+            "content": msg.get("content") or "", # 确保 None 转为空字符串
+            "time_display": msg.get("time_display", ""),
+            "type": msg.get("type", "")
+        }
+        
+        # 序列化并计算哈希
+        msg_str = json.dumps(stable_data, sort_keys=True, ensure_ascii=False)
+        return hashlib.md5(msg_str.encode('utf-8')).hexdigest()
+
+    async def run(self):
+        """
+        主运行循环
+        """
+        logger.info("🚀 正在启动 T2_ChatMonitor (Auto-Reply)...")
+        
+        # 定义 JSON 序列化辅助函数
+        def numpy_serializer(obj):
+            if isinstance(obj, np.integer):
+                return int(obj)
+            if isinstance(obj, np.floating):
+                return float(obj)
+            if isinstance(obj, np.ndarray):
+                return obj.tolist()
+            raise TypeError(f"Type {type(obj)} not serializable")
+        
+        # 1. 环境准备
+        if not self.step_1_prepare_env(): return
+        if not self.step_2_connect_device(): return
+
+        # [User Requested] 移除首屏概念，直接进入监控循环
+        # 以前说过什么都不管了，只关注最后一条
+        logger.info("🚀 启动完成，直接进入实时监控阶段...")
+        
+        # 3. 进入循环阶段
+        while True:
+            try:
+                # A. 截图并计算哈希
+                self.device.screenshot(self.screenshot_path)
+                current_screen_hash = self.get_image_hash(self.screenshot_path)
+                
+                # B. 如果屏幕无变化，则跳过识别
+                if current_screen_hash == self.last_screen_hash:
+                    await asyncio.sleep(self.check_interval)
+                    continue
+                
+                self.last_screen_hash = current_screen_hash
+                logger.info("📸 屏幕发生变化，正在分析...")
+
+                # C. 分析最新图片
+                dialogue_log, input_pos = await WxUtil.analyze_chat_image(
+                    self.screenshot_path, 
+                    self.debug_view_path, 
+                    device=self.device,
+                    process_strategy="UNREAD",  # 监控阶段：只处理带红点的新语音
+                    restore_processed_voice=False # 不还原状态，防止红点未消导致重复转换
+                )
+                
+                if not dialogue_log:
+                    logger.info("😴 未识别到有效消息")
+                    await asyncio.sleep(self.check_interval)
+                    continue
+                
+                logger.info(f"📊 当前识别到 {len(dialogue_log)} 条消息，最后一条: {dialogue_log[-1]}")
+                
+                # 更新当前对话日志（可用于上下文参考）
+                self.dialogue_log = dialogue_log
+                self.input_pos = input_pos
+                
+                # D. 只关注最后一条消息
+                last_msg = dialogue_log[-1]
+                # 计算稳定哈希（忽略坐标变化）
+                current_msg_hash = self.get_stable_message_hash(last_msg)
+                
+                # E. 判断是否需要回复 (对方发送且非重复消息)
+                sender = last_msg.get('sender', '')
+                
+                # Check if hash is already processed (in-memory only)
+                is_processed = current_msg_hash in self.processed_hashes
+                
+                # Log only if it changed from last *in-memory* check to avoid spam
+                if is_processed and current_msg_hash != self.last_processed_msg_hash:
+                     # logger.info(f"🚫 [监控] 消息哈希已存在于历史记录中，跳过回复 (Hash: {current_msg_hash})")
+                     self.last_processed_msg_hash = current_msg_hash
+
+                if not is_processed and current_msg_hash != self.last_processed_msg_hash:
+                    if sender != "我":
+                        event_shot = WxUtil.get_next_debug_path("event_new_msg")
+                        self.device.screenshot(event_shot)
+                        logger.info(f"💡 [监控] 发现新消息: {last_msg}，保存现场截图: {event_shot}")
+                        
+                        # 获取上下文文本 (格式化为 Sender: Content)
+                        context_text = "\n".join([f"{m.get('time_display', '') + ' ' if m.get('time_display') else ''}{m.get('sender')}: {m.get('content')}" for m in dialogue_log[:-1]])
+                        last_content = last_msg.get('content') or ""
+                        
+                        # 兜底逻辑：如果最后一条是语音且内容为空（可能因无红点未被 UNREAD 策略处理），尝试强制转换
+                        if last_msg.get('type') == 'voice' and not last_content.strip():
+                             logger.info("⚠️ [监控] 最后一条语音消息未获取到内容（可能已读无红点），尝试强制转换...")
+                             # 强制使用 LAST 策略重试，并且不还原状态（保持文字展开），防止下次循环因无内容再次触发重试
+                             dialogue_log_retry, _ = await WxUtil.analyze_chat_image(
+                                self.screenshot_path,
+                                self.debug_view_path,
+                                device=self.device,
+                                process_strategy="LAST",
+                                restore_processed_voice=False
+                             )
+                             if dialogue_log_retry:
+                                 # 更新引用
+                                 self.dialogue_log = dialogue_log_retry
+                                 dialogue_log = dialogue_log_retry
+                                 last_msg = dialogue_log[-1]
+                                 last_content = last_msg.get('content') or ""
+                                 logger.info(f"🔄 [重试] 强制转换后内容: {last_content}")
+                                 
+                                 # 重新构建哈希
+                                 current_msg_hash = self.get_stable_message_hash(last_msg)
+                                 # 再次检查是否已处理 (因为内容变了，哈希变了)
+                                 if current_msg_hash in self.processed_hashes:
+                                     logger.info(f"🚫 [重试] 转换后发现该消息已处理，跳过。")
+                                     self.last_processed_msg_hash = current_msg_hash
+                                     # 跳过本次循环的剩余部分
+                                     continue
+                        
+                        # 生成回复
+                        reply = await self.get_reply(last_content, context_text)
+                        
+                        if reply:
+                            logger.info(f"🤖 [监控] LLM 建议回复: {reply}")
+                            if self.input_pos:
+                                logger.info(f"⚡ [监控] 执行自动回复...")
+                                # input_pos 是 ((x,y), box) 格式，取第一个元素坐标点
+                                target_pos = self.input_pos[0] if isinstance(self.input_pos, (list, tuple)) and len(self.input_pos) == 2 and isinstance(self.input_pos[0], (list, tuple)) else self.input_pos
+                                # 简单兼容处理：如果 input_pos[0] 是 tuple/list 且 input_pos[1] 是 None/box，则取 input_pos[0]
+                                if isinstance(self.input_pos, (list, tuple)) and len(self.input_pos) == 2 and isinstance(self.input_pos[0], (list, tuple)):
+                                     target_pos = self.input_pos[0]
+                                
+                                # [User Request] 语音转文字未关闭 -> 顺序不对
+                                # 在回复发送前，尝试关闭语音转文字（如果最后一条是语音）
+                                # 这样截图看起来更干净，且符合用户期望的顺序
+                                if last_msg.get('type') == 'voice':
+                                    logger.info("🧹 [监控] 尝试关闭语音转文字显示...")
+                                    try:
+                                        # 点击语音消息中心，关闭文字
+                                        cx, cy = last_msg['center']
+                                        WxUtil.safe_device_click(self.device, cx, cy)
+                                        # 稍微等待界面刷新
+                                        await asyncio.sleep(1)
+                                    except Exception as e:
+                                        logger.warning(f"关闭语音转文字失败: {e}")
+
+                                perform_input_action(self.device, target_pos, reply)
+                                
+                                # 发送后截图留存
+                                reply_sent_shot = WxUtil.get_next_debug_path("event_reply_sent")
+                                self.device.screenshot(reply_sent_shot)
+                                logger.info(f"✅ [监控] 回复已发送，保存发送后截图: {reply_sent_shot}")
+                                
+                                self._record_processed_hash(last_msg, current_msg_hash)
+                                self.last_processed_msg_hash = current_msg_hash
+                            else:
+                                logger.error("❌ 未找到输入框位置，无法发送回复")
+                        else:
+                            logger.info("⚪ [监控] LLM 认为无需回复")
+                            self._record_processed_hash(last_msg, current_msg_hash)
+                            self.last_processed_msg_hash = current_msg_hash
+                    else:
+                         # 是我发的消息，更新哈希，不再处理
+                         self.last_processed_msg_hash = current_msg_hash
+
+                await asyncio.sleep(self.check_interval)
+
+            except Exception as e:
+                logger.error(f"Error in monitoring loop: {e}", exc_info=True)
+                await asyncio.sleep(self.check_interval)
+
+async def run_main():
+    """
+    运行自动巡课机器人
+    """
+    bot = ChatMonitorBot()
+    await bot.run()
+
+if __name__ == "__main__":
+    # 应用 Win32 补丁
+    Win32Patch.patch()
+    asyncio.run(run_main())

WeiXin/T2_Identify.py

@@ -0,0 +1,126 @@
+# coding=utf-8
+import os
+import sys
+import logging
+import cv2
+import json
+
+# Add project root
+project_root = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+if project_root not in sys.path:
+    sys.path.append(project_root)
+
+from WeiXin import WxUtil
+
+# Setup Logger
+logger = logging.getLogger("T2_Identify")
+logger.setLevel(logging.INFO)
+formatter = logging.Formatter('%(asctime)s - %(levelname)s - %(message)s')
+
+# Console Handler
+ch = logging.StreamHandler()
+ch.setFormatter(formatter)
+logger.addHandler(ch)
+
+# File Handler
+log_file = os.path.join(WxUtil.LOG_DIR, "T2_Identify.log")
+fh = logging.FileHandler(log_file, mode='w', encoding='utf-8')
+fh.setFormatter(formatter)
+logger.addHandler(fh)
+
+def main():
+    logger.info("--- T2: Identify Elements ---")
+    
+    # 1. Connect
+    d = WxUtil.connect_device()
+    if not d:
+        logger.error("Device connection failed.")
+        return
+
+    # 2. Screenshot/Image Source
+    if len(sys.argv) > 1:
+        screenshot_path = sys.argv[1]
+        logger.info(f"Using provided image: {screenshot_path}")
+    else:
+        screenshot_path = os.path.join(WxUtil.OUTPUT_DIR, "T2_Screenshot.jpg")
+        logger.info("Taking screenshot...")
+        try:
+            d.screenshot(screenshot_path)
+            logger.info(f"Screenshot saved to: {screenshot_path}")
+        except Exception as e:
+             logger.error(f"Failed to take screenshot: {e}")
+             # If screenshot fails, try to use existing one for debugging if it exists
+             if os.path.exists(screenshot_path):
+                 logger.warning(f"Using existing screenshot at {screenshot_path}")
+             else:
+                 return
+
+    # 3. Analyze (Using the internal _scan function to see raw detection)
+    logger.info("Scanning image for elements...")
+    messages, debug_img, chat_title = WxUtil._scan_chat_messages(screenshot_path)
+
+    # 4. Output Raw Detection
+    logger.info(f"Chat Title: {chat_title}")
+    logger.info(f"--- Raw Detection ({len(messages)} items) ---")
+    
+    for i, msg in enumerate(messages):
+        # Format for readability
+        content = msg.get('content', '')
+        if content is None: content = "[None]"
+        
+        type_str = msg.get('type', 'unknown')
+        sender_str = msg.get('sender', 'unknown')
+        center = msg.get('center', (0,0))
+        y = msg.get('y', 0)
+        time_disp = msg.get('time_display', 'N/A')
+        
+        logger.info(f"[{i}] {type_str:<8} | {sender_str:<4} | y={y:<4} | time={time_disp} | {content}")
+
+    # 5. Merging Logic (Simulate what the bot SHOULD do)
+    logger.info("--- Merged Dialogues (Simulated) ---")
+    merged_msgs = []
+    if messages:
+        current_msg = messages[0].copy()
+        last_y = current_msg['y'] # Track Y of the last element in the current merge block
+        
+        for i in range(1, len(messages)):
+            next_msg = messages[i]
+            
+            # Merge Criteria:
+            # 1. Same Sender
+            # 2. Same Type (text)
+            # 3. Vertically close (y_diff < 100) relative to the LAST line, not the first
+            # 4. Not a timestamp
+            
+            is_same_sender = (current_msg['sender'] == next_msg['sender'])
+            is_text = (current_msg['type'] == 'text' and next_msg['type'] == 'text')
+            
+            # Use last_y to compare with next_msg['y']
+            y_diff = next_msg['y'] - last_y
+            is_close = (y_diff < 120) # Tolerance for line spacing (slightly increased)
+            
+            if is_same_sender and is_text and is_close:
+                # Merge content
+                current_msg['content'] += " " + next_msg['content']
+                # Update last_y to this message's y
+                last_y = next_msg['y']
+            else:
+                merged_msgs.append(current_msg)
+                current_msg = next_msg.copy()
+                last_y = current_msg['y']
+        
+        merged_msgs.append(current_msg)
+        
+    for i, msg in enumerate(merged_msgs):
+        content = msg.get('content', '')
+        sender = msg.get('sender', 'unknown')
+        logger.info(f"[{i}] {sender}: {content}")
+
+    # 6. Save Debug Image
+    debug_path = os.path.join(WxUtil.OUTPUT_DIR, "T2_Identify_Debug.jpg")
+    if debug_img is not None:
+        cv2.imwrite(debug_path, debug_img)
+        logger.info(f"Debug image saved to: {debug_path}")
+
+if __name__ == "__main__":
+    main()

WeiXin/T3_Voice.py

@@ -0,0 +1,85 @@
+# coding=utf-8
+import os
+import sys
+import logging
+import asyncio
+import cv2
+
+# Add project root
+project_root = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+if project_root not in sys.path:
+    sys.path.append(project_root)
+
+from WeiXin import WxUtil
+
+# Setup Logger
+logger = logging.getLogger("T3_Voice")
+
+# Configure Root Logger to ensure all logs (including WxUtil) go to file
+root_logger = logging.getLogger()
+root_logger.setLevel(logging.INFO)
+
+# Ensure Log Directory exists
+if hasattr(WxUtil, 'LOG_DIR'):
+    log_dir = WxUtil.LOG_DIR
+else:
+    log_dir = os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))), "Logs")
+
+if not os.path.exists(log_dir):
+    os.makedirs(log_dir)
+
+# File Handler
+log_file = os.path.join(log_dir, "T3_Voice.log")
+fh = logging.FileHandler(log_file, mode='w', encoding='utf-8')
+formatter = logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s')
+fh.setFormatter(formatter)
+root_logger.addHandler(fh)
+
+# Console Handler (if not already added by WxUtil)
+if not any(isinstance(h, logging.StreamHandler) for h in root_logger.handlers):
+    ch = logging.StreamHandler()
+    ch.setFormatter(formatter)
+    root_logger.addHandler(ch)
+
+async def main():
+    logger.info("--- T3: Voice Processing Test ---")
+    
+    # 1. Connect
+    d = WxUtil.connect_device()
+    if not d:
+        logger.error("Device connection failed.")
+        return
+
+    # 2. Screenshot
+    screenshot_path = os.path.join(WxUtil.OUTPUT_DIR, "T3_Screenshot.jpg")
+    debug_path = os.path.join(WxUtil.OUTPUT_DIR, "T3_Debug.jpg")
+    
+    logger.info("Taking screenshot...")
+    d.screenshot(screenshot_path)
+    logger.info("Initial screenshot taken.")
+
+    # 3. Use analyze_chat_image with strategy='ALL' to test conversion
+    # This simulates the real bot behavior for voice messages
+    logger.info("Running WxUtil.analyze_chat_image (Strategy=ALL)...")
+    
+    # This function handles the click -> convert -> rescan loop internally
+    final_messages, _ = await WxUtil.analyze_chat_image(
+        screenshot_path, 
+        debug_path, 
+        device=d, 
+        process_strategy="ALL"
+    )
+
+    # 4. Output Result
+    logger.info("--- Final Voice Results ---")
+    voice_msgs = [m for m in final_messages if m['type'] == 'voice']
+    if not voice_msgs:
+        logger.info("No voice messages found.")
+    else:
+        for i, msg in enumerate(voice_msgs):
+            content = msg.get('content')
+            converted_status = "YES" if msg.get('is_converted') else "NO"
+            logger.info(f"Voice [{i}] Sender: {msg['sender']} | Converted: {converted_status} | Content: {content}")
+
+if __name__ == "__main__":
+    asyncio.run(main())

WeiXin/T4_Text.py

@@ -0,0 +1,104 @@
+# coding=utf-8
+import os
+import sys
+import logging
+import cv2
+
+# Add project root
+project_root = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+if project_root not in sys.path:
+    sys.path.append(project_root)
+
+from WeiXin import WxUtil
+
+# Setup Logger
+logger = logging.getLogger("T4_Text")
+
+# Configure Root Logger to ensure all logs (including WxUtil) go to file
+root_logger = logging.getLogger()
+root_logger.setLevel(logging.INFO)
+
+# Ensure Log Directory exists
+if hasattr(WxUtil, 'LOG_DIR'):
+    log_dir = WxUtil.LOG_DIR
+else:
+    log_dir = os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))), "Logs")
+
+if not os.path.exists(log_dir):
+    os.makedirs(log_dir)
+
+# File Handler
+log_file = os.path.join(log_dir, "T4_Text.log")
+fh = logging.FileHandler(log_file, mode='w', encoding='utf-8')
+formatter = logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s')
+fh.setFormatter(formatter)
+root_logger.addHandler(fh)
+
+# Console Handler (if not already added by WxUtil)
+if not any(isinstance(h, logging.StreamHandler) for h in root_logger.handlers):
+    ch = logging.StreamHandler()
+    ch.setFormatter(formatter)
+    root_logger.addHandler(ch)
+
+def main():
+    logger.info("--- T4: Text Processing Test ---")
+    
+    # 1. Connect
+    d = WxUtil.connect_device()
+    if not d:
+        logger.error("Device connection failed.")
+        return
+
+    # 2. Screenshot
+    screenshot_path = os.path.join(WxUtil.OUTPUT_DIR, "T4_Screenshot.jpg")
+    logger.info("Taking screenshot...")
+    d.screenshot(screenshot_path)
+    
+    # 3. Scan
+    logger.info("Scanning image...")
+    messages, debug_img, _ = WxUtil._scan_chat_messages(screenshot_path)
+    
+    # 4. Filter and Output
+    logger.info("--- Raw Text Messages ---")
+    text_msgs = [m for m in messages if m['type'] == 'text']
+    
+    if not text_msgs:
+        logger.info("No text messages found.")
+    else:
+        for i, msg in enumerate(text_msgs):
+            logger.info(f"Text [{i}] Sender: {msg['sender']} | y={msg['y']} | Content: {msg['content']}")
+
+    # 5. Merging Logic (Same as T2)
+    logger.info("--- Merged Text Dialogues ---")
+    merged_msgs = []
+    if text_msgs:
+        current_msg = text_msgs[0].copy()
+        last_y = current_msg['y']
+        
+        for i in range(1, len(text_msgs)):
+            next_msg = text_msgs[i]
+            
+            is_same_sender = (current_msg['sender'] == next_msg['sender'])
+            y_diff = next_msg['y'] - last_y
+            is_close = (y_diff < 120) 
+            
+            if is_same_sender and is_close:
+                current_msg['content'] += " " + next_msg['content']
+                last_y = next_msg['y']
+            else:
+                merged_msgs.append(current_msg)
+                current_msg = next_msg.copy()
+                last_y = current_msg['y']
+        
+        merged_msgs.append(current_msg)
+    
+    for i, msg in enumerate(merged_msgs):
+        logger.info(f"[{i}] {msg['sender']}: {msg['content']}")
+
+    # Save debug
+    debug_path = os.path.join(WxUtil.OUTPUT_DIR, "T4_Debug.jpg")
+    cv2.imwrite(debug_path, debug_img)
+    logger.info(f"Debug image saved to: {debug_path}")
+
+if __name__ == "__main__":
+    main()

WeiXin/T5_SendText.py

@@ -0,0 +1,80 @@
+# coding=utf-8
+import os
+import sys
+import logging
+import time
+
+# Add project root
+project_root = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+if project_root not in sys.path:
+    sys.path.append(project_root)
+
+from WeiXin import WxUtil
+
+# Setup Logger
+logger = logging.getLogger("T5_SendText")
+
+# Configure Root Logger to ensure all logs go to file
+root_logger = logging.getLogger()
+root_logger.setLevel(logging.INFO)
+
+# Ensure Log Directory exists
+if hasattr(WxUtil, 'LOG_DIR'):
+    log_dir = WxUtil.LOG_DIR
+else:
+    log_dir = os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))), "Logs")
+
+if not os.path.exists(log_dir):
+    os.makedirs(log_dir)
+
+# File Handler
+log_file = os.path.join(log_dir, "T5_SendText.log")
+fh = logging.FileHandler(log_file, mode='w', encoding='utf-8')
+formatter = logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s')
+fh.setFormatter(formatter)
+root_logger.addHandler(fh)
+
+# Console Handler
+if not any(isinstance(h, logging.StreamHandler) for h in root_logger.handlers):
+    ch = logging.StreamHandler()
+    ch.setFormatter(formatter)
+    root_logger.addHandler(ch)
+
+def main():
+    logger.info("--- T5: Text Input and Send Test ---")
+    
+    # 1. Connect
+    d = WxUtil.connect_device()
+    if not d:
+        logger.error("Device connection failed.")
+        return
+
+    # 2. Prepare text
+    test_text = "我是AI返回的文本测试"
+    logger.info(f"Test Content: {test_text}")
+
+    # 3. Find Input Box and Send (Using perform_input_action which encapsulates the steps)
+    # We pass auto_send=True to click the send button
+    # The function will:
+    #   - Ensure keyboard mode
+    #   - Find input box (UI Tree -> Template -> Fallback)
+    #   - Input text
+    #   - Find and click Send button
+    
+    logger.info("Step 1 & 2: Finding input box and writing text...")
+    
+    # We can get a rough position first for logging, though perform_input_action is more robust
+    screenshot_path = os.path.join(WxUtil.OUTPUT_DIR, "T5_Initial_Check.jpg")
+    d.screenshot(screenshot_path)
+    input_pos, _ = WxUtil.find_input_box_center(screenshot_path)
+    logger.info(f"Initial target input position: {input_pos}")
+
+    success = WxUtil.perform_input_action(d, input_pos, test_text, auto_send=True)
+    
+    if success:
+        logger.info("✅ Step 3: Send action completed successfully.")
+    else:
+        logger.error("❌ Send action failed.")
+
+if __name__ == "__main__":
+    main()