"""消息监控 — 全列表指纹对比 + 窗口冻结检测 + ControlFromHandle 强制刷新""" import time import logging import threading import traceback from typing import Optional, Callable, List import uiautomation as auto import win32gui import win32con import win32api import win32process import re import ctypes import difflib import os from datetime import datetime import signal from PIL import Image, ImageGrab, ImageEnhance, ImageFilter from winrt.windows.media.ocr import OcrEngine from winrt.windows.globalization import Language from winrt.windows.graphics.imaging import SoftwareBitmap, BitmapPixelFormat logging.getLogger("comtypes").setLevel(logging.WARNING) logger = logging.getLogger("msgbot.monitor") # EasyOCR 备用引擎（可选依赖） try: import easyocr as _easyocr import numpy as _np _HAS_EASYOCR = True except (ImportError, OSError): _HAS_EASYOCR = False _easyocr = None _np = None logger.debug("[ocr] EasyOCR 未安装/不可用，仅使用 Windows OCR") class MessageMonitor: """多线程消息监控器""" def __init__(self, on_message: Optional[Callable[[str, str], None]] = None): self._running = False self._thread: Optional[threading.Thread] = None self._poll_interval = 2.0 self._on_message = on_message self._wechat_enabled = True self._qq_enabled = False # 全列表签名 & 完整文本列表 (用于新消息比对) self._last_sig: dict = {} self._last_texts: dict = {} # 最新消息文本（供 GUI 展示） self._last_msg: dict = {} # 轮询计数 self._poll_count: dict = {} # QQ 轮询跳过计数（每 3 周期 poll 一次） self._qq_skip_counter = 0 # 窗口句柄缓存（存 HWND，不存 Control 引用） self._cached_hwnd: dict = {} # 窗口对应进程 ID（验证窗口是否仍属同一进程） self._cached_pid: dict = {} # 连续相同签名计数（用于冻结检测） self._stale_count: dict = {} # 连续空读 self._empty_reads: dict = {} # 剪贴板原始文本（用于自动回复提取发送者） self._raw_clipboard: dict = {} # 发送中事件锁（防止自动回复时监控轮询干扰） self._sending_event = threading.Event() # 应用变体：qq→"qq_nt"或"tim" self._app_variant: dict = {} # 已处理消息的全局持久化集合（跨平台、跨轮次，防回环） self._seen_messages: set = set() # 已回复文本的永久集—永不裁切（剪贴板读到自己回复时必拦） self._replied_texts: set = set() # 上次回复时间戳（冷却期内跳过该平台 poll） self._last_reply_time: dict = {} # platform → timestamp # OCR 引擎（Windows 原生，lazy init） self._ocr_engine = None # EasyOCR 备用引擎（lazy init） self._easyocr_reader = None # 引擎选择计数器 self._ocr_engine_wins: dict = {"windows": 0, "easyocr": 0} # 上次 OCR 原始文本（用于增量检测） self._ocr_last_text: dict = {} # OCR 垃圾连续计数（连续垃圾超过阈值清缓存重找） self._ocr_garbage_count: int = 0 def set_poll_interval(self, seconds: float): self._poll_interval = max(0.5, seconds) def set_enabled_platforms(self, wechat: bool = True, qq: bool = False): self._wechat_enabled = wechat self._qq_enabled = qq def set_callback(self, cb: Callable[[str, str], None]): self._on_message = cb def start(self): if self._running: return self._running = True self._thread = threading.Thread(target=self._run_loop, daemon=True, name="monitor") self._thread.start() logger.info("消息监控已启动") def stop(self): self._running = False if self._thread and self._thread.is_alive(): self._thread.join(timeout=3) self._thread = None self._cached_hwnd.clear() self._cached_pid.clear() self._last_sig.clear() self._last_msg.clear() self._stale_count.clear() self._empty_reads.clear() self._poll_count.clear() logger.info("消息监控已停止") @property def is_running(self) -> bool: return self._running # ── 轮询主循环 ────────────────────────────────────────── def _run_loop(self): # 创建 Windows 消息队列（使 AttachThreadInput / SetForegroundWindow 在 daemon 线程生效） try: ctypes.windll.user32.PeekMessageW(None, 0, 0, 0, 1) # PM_NOREMOVE except Exception: pass heartbeat = 0 logger.debug("_run_loop 进入 while 循环") while self._running: try: if self._wechat_enabled and not self._sending_event.is_set(): self._poll("wechat") if self._qq_enabled: # QQ 降频：每 4 周期 poll 一次（~8s，OCR 计算略重） self._qq_skip_counter += 1 if self._qq_skip_counter >= 4 and not self._sending_event.is_set(): self._qq_skip_counter = 0 self._poll("qq") except Exception as e: logger.error(f"轮询异常: {e}\n{traceback.format_exc()}") heartbeat += 1 if heartbeat % 10 == 0: logger.info(f"[♡] 心跳 #{heartbeat}") time.sleep(self._poll_interval) # ── 窗口定位 ───────────────────────────────────────────── def _ensure_hwnd(self, platform: str) -> Optional[int]: """查找窗口并返回 HWND（不缓存 Control）。""" hwnd = self._cached_hwnd.get(platform) try: if hwnd: if win32gui.IsWindow(hwnd): return hwnd except Exception: pass self._cached_hwnd.pop(platform, None) self._cached_pid.pop(platform, None) if platform == "wechat": for name in ["微信", "WeChat"]: try: w = auto.WindowControl(Name=name) if w.Exists(maxSearchSeconds=0.5): rect = w.BoundingRectangle if rect.width() > 200 and rect.height() > 100: h = w.NativeWindowHandle # 确保存顶层窗口句柄（与 GetForegroundWindow 返回值一致） top = win32gui.GetAncestor(h, win32con.GA_ROOT) self._cached_hwnd[platform] = top self._cached_pid[platform] = win32process.GetWindowThreadProcessId(top)[1] logger.info(f"[{platform}] 窗口已定位: {name} handle={top}") return top except Exception: pass elif platform == "qq": # 1. Win32 FindWindow 直搜 TXGuiFoundation（旧版 QQ 主窗口） # ⚠️ 必须检查 IsWindowVisible：QQ NT 在托盘时有一个不可见的 TXGuiFoundation # 托盘辅助窗口（空标题, IsWindowVisible=False），FindWindow 可能返回这个。 # 不可见的窗口被缓存后，_poll 里 IsWindowVisible 检查会跳过所有后续 OCR。 try: hwnd = win32gui.FindWindow("TXGuiFoundation", None) if hwnd: top = win32gui.GetAncestor(hwnd, win32con.GA_ROOT) if win32gui.IsWindow(top) and win32gui.IsWindowVisible(top): rect = win32gui.GetWindowRect(top) if (rect[2] - rect[0]) > 200 and (rect[3] - rect[1]) > 100: title = win32gui.GetWindowText(top) self._app_variant[platform] = "tim" if "TIM" in title.upper() else "qq_nt" self._cached_hwnd[platform] = top self._cached_pid[platform] = win32process.GetWindowThreadProcessId(top)[1] logger.info(f"[{platform}] 窗口已定位(Win32 TXGuiFoundation): {title} top={top}") return top except Exception: pass # 2. Win32 EnumWindows 兜底 — 类名+标题搜索 # QQ NT (Electron) 类名 Chrome_WidgetWin_1, 传统 QQ 类名 TXGuiFoundation qq_titles = ("QQ", "TIM") def _enum_callback(hwnd, results): try: cls = win32gui.GetClassName(hwnd) title = win32gui.GetWindowText(hwnd) title_upper = title.upper() # 类名匹配（传统 QQ / TIM） if cls == "TXGuiFoundation": results.append(hwnd) return True # QQ NT Electron: 类名含 Chrome，标题含 QQ/TIM if "Chrome" in cls and ("QQ" in title_upper or "TIM" in title_upper): results.append(hwnd) return True # 标题完全匹配 QQ / TIM if title_upper in qq_titles: results.append(hwnd) return True except Exception: pass return True try: found = [] win32gui.EnumWindows(_enum_callback, found) logger.debug(f"[{platform}] EnumWindows 候选: {len(found)} windows: " f"{[(win32gui.GetWindowText(h)[:30], hex(h)) for h in found]}") for hwnd in found: top = win32gui.GetAncestor(hwnd, win32con.GA_ROOT) if not (win32gui.IsWindow(top) and win32gui.IsWindowVisible(top)): cls = win32gui.GetClassName(top) title = win32gui.GetWindowText(top) logger.debug(f"[{platform}] Enum 候选跳过(不可见): cls={cls} title={title[:30]} vis={win32gui.IsWindowVisible(top)}") continue rect = win32gui.GetWindowRect(top) if not ((rect[2] - rect[0]) > 200 and (rect[3] - rect[1]) > 100): cls = win32gui.GetClassName(top) title = win32gui.GetWindowText(top) logger.debug(f"[{platform}] Enum 候选跳过(尺寸): cls={cls} title={title[:30]} rect={rect}") continue cls = win32gui.GetClassName(top) title = win32gui.GetWindowText(top) title_upper = title.upper().strip() # _enum_callback 只加了 QQ/TIM 相关窗口，所以 found 里都是有效候选 # 优先选标题含 QQ/TIM 的（传统模式），其次任何 TXGuiFoundation 或 Chrome 都接受 if "TIM" in title_upper: self._app_variant[platform] = "tim" else: self._app_variant[platform] = "qq_nt" self._cached_hwnd[platform] = top self._cached_pid[platform] = win32process.GetWindowThreadProcessId(top)[1] if "Chrome" in cls: logger.info(f"[{platform}] 窗口已定位(Electron Chrome): {title} top={top}") else: logger.info(f"[{platform}] 窗口已定位(Win32 Enum): {title} top={top}") return top except Exception: pass # 3. UIA 按窗口名搜索（兜底，可能慢但不挂死） for name in ["QQ", "TIM"]: try: w = auto.WindowControl(Name=name) if w.Exists(maxSearchSeconds=0.3): rect = w.BoundingRectangle if rect.width() > 200 and rect.height() > 100: h = w.NativeWindowHandle top = win32gui.GetAncestor(h, win32con.GA_ROOT) self._cached_hwnd[platform] = top self._cached_pid[platform] = win32process.GetWindowThreadProcessId(top)[1] self._app_variant[platform] = "qq_nt" if name == "QQ" else "tim" logger.info(f"[{platform}] 窗口已定位(UIA 名称): {name} top={top}") return top except Exception: pass return None def _ensure_hwnd_with_timeout(self, platform: str, timeout: float = 12.0) -> Optional[int]: """_ensure_hwnd 包装 daemon 线程超时保护。 ThreadPoolExecutor 创建非 daemon 线程 → Ctrl+C 进程不退出。 改用 daemon thread + Event.wait(timeout)，daemon 线程不阻塞进程退出。 """ result = [None] done = threading.Event() def worker(): result[0] = self._ensure_hwnd(platform) done.set() t = threading.Thread(target=worker, daemon=True) t.start() if not done.wait(timeout=timeout): logger.warning(f"[{platform}] _ensure_hwnd 超时({timeout:.0f}s)，跳过本轮") return None return result[0] # ── 强制刷新 UIA ────────────────────────────────────────── def _fresh_window(self, platform: str): """从缓存 HWND 创建全新的 WindowControl（绕过 UIA 缓存）。""" hwnd = self._cached_hwnd.get(platform) if not hwnd: return None try: # 用 NativeWindowHandle 重建，保证 UIA 新缓存 win = auto.ControlFromHandle(hwnd) if win is not None: return win except Exception: pass return None def _find_qq_window_fresh(self): """通过类名重新查找 QQ/TIM 窗口，绕过 UIA 缓存（比 ControlFromHandle 更彻底）。""" try: # 先按窗口名称查找 for name in ["QQ", "TIM"]: try: w = auto.WindowControl(Name=name) if w.Exists(maxSearchSeconds=0.2): rect = w.BoundingRectangle if rect.width() > 200 and rect.height() > 100: return w except Exception: pass # 按 TXGuiFoundation 类名深度查找 root = auto.GetRootControl() for child in root.GetChildren(): try: if child.ClassName == "TXGuiFoundation": rect = child.BoundingRectangle if rect.width() > 200 and rect.height() > 100: return child except Exception: pass except Exception: pass return None def _force_uia_refresh(self, platform: str): """通过交互强制窗口更新 UIA 控件树。""" hwnd = self._cached_hwnd.get(platform) if not hwnd: return try: # 方式1: 窗口重绘 win32gui.InvalidateRect(hwnd, None, True) win32gui.UpdateWindow(hwnd) # QQ: 避免标题栏点击（不会干扰用户操作） if platform == "qq": win32gui.PostMessage(hwnd, win32con.WM_VSCROLL, win32con.SB_BOTTOM, 0) win32gui.PostMessage(hwnd, win32con.WM_KEYDOWN, win32con.VK_END, 0) win32gui.PostMessage(hwnd, win32con.WM_KEYUP, win32con.VK_END, 0) else: # 方式2: 向窗口发送非客户区点击（仅微信） win32gui.PostMessage(hwnd, 0x00A1, 2, 0) # WM_NCLBUTTONDOWN win32gui.PostMessage(hwnd, 0x00A2, 2, 0) # WM_NCLBUTTONUP logger.debug(f"[{platform}] UIA强制刷新") except Exception as e: logger.debug(f"[{platform}] 强制刷新失败: {e}") def _force_scroll_bottom(self, platform: str): """滚动窗口到底部，触发 UI 更新。""" hwnd = self._cached_hwnd.get(platform) if not hwnd: return try: # 1. WM_VSCROLL SB_BOTTOM win32gui.PostMessage(hwnd, win32con.WM_VSCROLL, win32con.SB_BOTTOM, 0) # 2. Ctrl+End 快捷键 win32gui.PostMessage(hwnd, win32con.WM_KEYDOWN, win32con.VK_CONTROL, 0) win32gui.PostMessage(hwnd, win32con.WM_KEYDOWN, win32con.VK_END, 0) win32gui.PostMessage(hwnd, win32con.WM_KEYUP, win32con.VK_END, 0) win32gui.PostMessage(hwnd, win32con.WM_KEYUP, win32con.VK_CONTROL, 0) # 3. InvalidateRect 强制重绘 win32gui.InvalidateRect(hwnd, None, True) except Exception as e: logger.debug(f"[{platform}] 滚动到底失败: {e}") def _ensure_foreground(self, hwnd, max_retries=5) -> bool: """确保窗口在前台，重试多次。失败返回 False，不强行发送。""" root_hwnd = win32gui.GetAncestor(hwnd, win32con.GA_ROOT) if not root_hwnd or not win32gui.IsWindow(root_hwnd): logger.warning(f"[前台] 无效句柄 {hwnd}") return False # 前置条件：最小化则恢复，隐藏则显示 if win32gui.IsIconic(root_hwnd): win32gui.ShowWindow(root_hwnd, win32con.SW_RESTORE) time.sleep(0.2) if not win32gui.IsWindowVisible(root_hwnd): win32gui.ShowWindow(root_hwnd, win32con.SW_SHOWNORMAL) time.sleep(0.2) # 允许任意进程 SetForegroundWindow（移除系统限制，与发送路径一致） try: ctypes.windll.user32.AllowSetForegroundWindow(-1) # ASFW_ANY except Exception: pass for i in range(max_retries): # Method 1: AttachThreadInput + SetForegroundWindow self._attach_set_foreground(root_hwnd) time.sleep(0.25) if self._check_foreground(root_hwnd): return True # Method 2: SwitchToThisWindow（权限要求低） try: ctypes.windll.user32.SwitchToThisWindow(root_hwnd, True) time.sleep(0.2) if self._check_foreground(root_hwnd): return True except Exception: pass # Method 3: SetWindowPos 置顶再取消（最后手段） if i >= 2: try: HWND_TOPMOST = -1 HWND_NOTOPMOST = -2 SWP_NOSIZE = 1 SWP_NOMOVE = 2 ctypes.windll.user32.SetWindowPos( root_hwnd, HWND_TOPMOST, 0, 0, 0, 0, SWP_NOMOVE | SWP_NOSIZE ) time.sleep(0.1) ctypes.windll.user32.SetWindowPos( root_hwnd, HWND_NOTOPMOST, 0, 0, 0, 0, SWP_NOMOVE | SWP_NOSIZE ) time.sleep(0.15) if self._check_foreground(root_hwnd): return True except Exception: pass time.sleep(0.15) logger.warning(f"无法将窗口 {hwnd} 带到前台，放弃操作") return False @staticmethod def _check_foreground(root_hwnd) -> bool: """检查窗口是否在前台（或其子窗口）。""" try: fg = win32gui.GetForegroundWindow() if fg == root_hwnd: return True if fg and win32gui.IsChild(root_hwnd, fg): return True except Exception: pass return False def _attach_set_foreground(self, hwnd): """AttachThreadInput + SetForegroundWindow，从 daemon 线程也能成功。""" try: current_tid = ctypes.windll.kernel32.GetCurrentThreadId() target_tid = ctypes.windll.user32.GetWindowThreadProcessId(hwnd, None) if ctypes.windll.user32.AttachThreadInput(current_tid, target_tid, True): try: win32gui.SetForegroundWindow(hwnd) except Exception: pass finally: ctypes.windll.user32.AttachThreadInput(current_tid, target_tid, False) return # AttachThreadInput 失败 → 裸 SetForegroundWindow try: win32gui.SetForegroundWindow(hwnd) except Exception: pass except Exception: try: win32gui.SetForegroundWindow(hwnd) except Exception: pass def _read_win32_texts(self, platform: str) -> list: """Win32 API 读窗口文本，递归枚举（depth 不限）。""" hwnd = self._cached_hwnd.get(platform) if not hwnd: return [] texts = [] def enum_callback(child_hwnd, _): try: text = win32gui.GetWindowText(child_hwnd) if text and len(text.strip()) >= 2: texts.append(text.strip()) except Exception: pass return True try: main_text = win32gui.GetWindowText(hwnd) if main_text and len(main_text.strip()) >= 2: texts.append(main_text.strip()) win32gui.EnumChildWindows(hwnd, enum_callback, None) except Exception: pass return texts # ── OCR 读屏（QQ 专用，Windows 原生 OCR） ────────────────── def _ocr_ensure_engine(self): """惰性初始化 OCR 引擎（中文简体）。""" if self._ocr_engine is None: try: self._ocr_engine = OcrEngine.try_create_from_language( Language('zh-CN') ) if self._ocr_engine: logger.info(f"[ocr] OCR引擎就绪: {self._ocr_engine.recognizer_language.display_name}") else: logger.warning("[ocr] 创建OCR引擎失败") except Exception as e: logger.error(f"[ocr] OCR引擎初始化异常: {e}") return self._ocr_engine @staticmethod def _normalize_ocr_text(text: str) -> str: """Windows OCR 给 CJK 字符间强加空格 → 去除这些空格使文本可去重。 "问 题 1 ： 你 答 ： 你 好" → "问题1：你答：你好" "你 好 he llo" → "你好 he llo"（ASCII 词内空格保留） 此外修复常见 Windows OCR 字符混淆错误（只在非 md 文本上运行， md 文档由 pre-split check 提前过滤，不受影响）。 """ # 所有紧邻 CJK/全角字符的空格全部去掉 # CJK 范围：汉字 (u4e00-) + 扩展 (u3400-) + 全角/兼容 (uff00-) text = re.sub( r'(?<=[\u4e00-\u9fff\u3400-\u4dbf\uff00-\uffef\u3000-\u303f])' r'\s+', '', text ) text = re.sub( r'\s+' r'(?=[\u4e00-\u9fff\u3400-\u4dbf\uff00-\uffef\u3000-\u303f])', '', text ) # ⭐ Windows OCR 字符混淆修复 # 《》→ <>：Minecraft 聊天大量使用 等尖括号，OCR 误读为中文书名号 text = text.replace('《', '<').replace('》', '>') # l→| 管道分隔符：|quiz| / |start| / |info| 误读为 lquizl / lstartl / linfol text = re.sub(r'(?<=\s)l([A-Za-z]\w*?)l(?=\s)', r'|\1|', text) # ．→ . 全角点→半角数字小数点 (0．45 → 0.45) text = re.sub(r'(\d)[．](\d)', r'\1.\2', text) return text.strip() @staticmethod def _is_ocr_console_garbage(text: str) -> bool: """OCR 输出的文本是否明显是控制台/Python 垃圾。 当 OCR 截错窗口（如读到终端窗口）时，文本包含： - Python 路径/模块名：comtypes, uiautomation, sklearn, jieba, winrt - 脚本文件名：monitor.py, main.py - 虚拟环境路径：.venv - 日志关键词：DEBUG, INFO, WARNING, msgbot - 控制台序列：←[, [32m, [0m 返回 True = 应当丢弃。 """ garbage_patterns = [ r'comtypes', r'uiautomation', r'sklearn', r'jieba', r'winrt\.', r'msgbot\.', r'monitor\.py', r'main\.py', r'\.venv', r'Loading model', r'Prefix dict', r'WARNING', r'UserWarning', r'Building prefix', r'←\[', r'\[32m', r'\[0m', r'\\n', ] for pat in garbage_patterns: if re.search(pat, text, re.IGNORECASE): return True # 全字母数字+标点无中文 → 肯定是非聊天内容 cjk_count = sum(1 for c in text if '\u4e00' <= c <= '\u9fff') if cjk_count == 0 and len(text.strip()) > 20: return True return False def _validate_qq_window(self, hwnd: int) -> bool: """验证窗口是否确实是 QQ/TIM 进程。 不尝试自动恢复隐藏/最小化窗口——交给剪贴板兜底的 _ensure_foreground 处理。 _ensure_foreground 有多重重试+SetForegroundWindow，比这里抢先 ShowWindow 更能正确弹出主窗口而非托盘预览。 """ if not win32gui.IsWindow(hwnd): return False if not win32gui.IsWindowVisible(hwnd): return False if win32gui.IsIconic(hwnd): return False # 验证进程名 try: _, pid = win32process.GetWindowThreadProcessId(hwnd) hproc = win32api.OpenProcess( 0x0400 | 0x0010, # PROCESS_QUERY_INFORMATION | PROCESS_VM_READ False, pid ) if hproc: exe_path = win32process.GetModuleFileNameEx(hproc, 0) win32api.CloseHandle(hproc) exe_name = exe_path.split('\\')[-1].lower() if exe_path else '' if exe_name in ('qq.exe', 'tim.exe'): return True except Exception: pass return False @staticmethod def _fuzzy_duplicate(new_text: str, seen_set: set, threshold: float = 0.80) -> bool: """检查 new_text 是否与 seen_set 中已有文本相似度超阈值。 OCR 输出有噪声（每次勖彐碎片、行首数字、尾部截断不同），精确匹配无法去重。 用 difflib.SequenceMatcher 的 quick_ratio() 做快速上界剪枝。 """ if not new_text or len(new_text) < 4: return False for existing in seen_set: if abs(len(existing) - len(new_text)) > max(len(existing), len(new_text)) * 0.4: # 长度差距超 40%，不可能相似（quick_ratio 上界 = 2*min/max < 阈值） continue ratio = difflib.SequenceMatcher(None, new_text, existing).quick_ratio() if ratio >= threshold: return True return False # ── 双 OCR 引擎 ────────────────────────────────────────── def _init_easyocr(self): """惰性初始化 EasyOCR（首次约 3-10 秒）。""" if not _HAS_EASYOCR or self._easyocr_reader is not None: return try: logger.info("[ocr] EasyOCR 初始化中（首次约 3-10 秒）...") self._easyocr_reader = _easyocr.Reader(['ch_sim', 'en'], gpu=False) logger.info("[ocr] EasyOCR 就绪（CPU 模式）") except Exception as e: logger.error(f"[ocr] EasyOCR 初始化失败: {e}") self._easyocr_reader = None @staticmethod def _score_ocr_text(text: str) -> float: """为 OCR 结果质量打分（越高越好）。对比引擎用。""" if not text or len(text.strip()) < 5: return 0.0 score = 0.0 # 1) 有效 CJK 数量（越多 = 识别越好） cjk = sum(1 for c in text if '\u4e00' <= c <= '\u9fff') score += cjk * 0.5 # 2) Unicode 垃圾字符扣分（OCR 常见误识碎片） junk = set('勖彐丬亻俨冉勹刂冫孑亠丷') junk_count = sum(1 for c in text if c in junk) score -= junk_count * 3 # 3) CJK 间空格比例扣分（Windows OCR 特有毛病） # "你 好" vs "你好" — 空格越多说明识别越差 total = len(text) spaces = text.count(' ') if total > 0 and cjk > 0: space_penalty = (spaces / total) * 10 score -= space_penalty # 4) 时间戳加分 = 消息结构更完整 timestamps = len(re.findall(r'\d{1,2}:\d{2}', text)) score += min(timestamps * 2, 6) return max(score, 0.0) def _pick_ocr_text(self, win_text: str, easy_text: str, image: Image.Image) -> str: """双引擎对比，选质量更高的结果。日志记录引擎胜负统计。""" if not easy_text: return win_text if not win_text: return easy_text win_score = self._score_ocr_text(win_text) easy_score = self._score_ocr_text(easy_text) # 引擎选择统计 chosen = "windows" if win_score >= easy_score else "easyocr" self._ocr_engine_wins[chosen] = self._ocr_engine_wins.get(chosen, 0) + 1 total = sum(self._ocr_engine_wins.values()) ratio = self._ocr_engine_wins.get(chosen, 0) / max(total, 1) logger.debug( f"[ocr] 引擎对比: win={win_score:.1f} vs easy={easy_score:.1f} → 选{chosen} " f"(胜率{ratio:.0%}, {total}轮)" ) return win_text if chosen == "windows" else easy_text def _ocr_read_qq(self) -> list: """OCR 截屏 QQ 消息区域 -> 双引擎（Windows OCR + EasyOCR）提取文本。 包含窗口验证 + 垃圾过滤 + 连续垃圾清缓存。 """ hwnd = self._cached_hwnd.get("qq") if not hwnd: return [] if not self._validate_qq_window(hwnd): self._ocr_garbage_count += 1 if self._ocr_garbage_count == 1: # 首次失败时详细日志 visible = win32gui.IsWindowVisible(hwnd) if win32gui.IsWindow(hwnd) else False iconic = win32gui.IsIconic(hwnd) if win32gui.IsWindow(hwnd) else False logger.debug(f"[qq] OCR: 窗口不可用(visible={visible}, iconic={iconic})") if self._ocr_garbage_count >= 3: logger.info("[qq] OCR: 窗口不合法连续3次，清缓存") self._cached_hwnd.pop("qq", None) self._cached_pid.pop("qq", None) self._app_variant.pop("qq", None) self._ocr_garbage_count = 0 return [] engine = self._ocr_ensure_engine() if not engine: return [] try: rect = win32gui.GetWindowRect(hwnd) if not rect: return [] left, top, right, bottom = rect win_w = right - left win_h = bottom - top # QQ NT 窗口布局: ~0-5% 标签栏, ~5-30% 联系人/群列表, ~30-55% 会话列表, ~55-100% 消息面板 # 先切 QQ 到前台再截屏，避免 CMD 窗口遮挡→OCR 读到终端文字 self._ensure_foreground(hwnd) time.sleep(0.15) # 点一下消息面板区域设置焦点（TIM 默认焦点在会话列表，VK_END 只会滚会话列表） try: click_x = int(win_w * 0.72) # 消息面板中心偏右 click_y = int(win_h * 0.50) lparam = win32api.MAKELONG(click_x, click_y) win32gui.PostMessage(hwnd, win32con.WM_LBUTTONDOWN, win32con.MK_LBUTTON, lparam) win32gui.PostMessage(hwnd, win32con.WM_LBUTTONUP, 0, lparam) logger.debug(f"[qq] 点击消息面板({click_x},{click_y})设置焦点") time.sleep(0.15) except Exception as e: logger.debug(f"[qq] 点击消息面板失败: {e}") # Ctrl+End 滚动到底部，只截最新消息（避开历史堆积+bot 旧回复） VK_CONTROL = 0x11 VK_END = 0x23 # 方式A: keybd_event（窗口在前台时生效） ctypes.windll.user32.keybd_event(VK_CONTROL, 0, 0, 0) ctypes.windll.user32.keybd_event(VK_END, 0, 0, 0) time.sleep(0.05) ctypes.windll.user32.keybd_event(VK_END, 0, 2, 0) ctypes.windll.user32.keybd_event(VK_CONTROL, 0, 2, 0) # 方式B: PostMessage 兜底（keybd_event 可能被 TIM 忽略） win32gui.PostMessage(hwnd, win32con.WM_KEYDOWN, win32con.VK_END, 0) win32gui.PostMessage(hwnd, win32con.WM_KEYUP, win32con.VK_END, 0) time.sleep(0.3) # TIM 窗口布局: ~5-30% 左栏, ~30-50% 会话列表(含文档预览), ~50-100% 消息面板 # 退回 c9acd7e 裁剪: 25% 覆盖消息面板两侧(对方+自己消息都可见) # 右边界 78% 排除群成员侧栏(~80%+), md 过滤负责拦截 bot 的 md 格式回复 cap_left = int(left + win_w * 0.25) cap_top = top + max(80, int(win_h * 0.06)) # 跳过标题栏+群名头栏 cap_right = int(left + win_w * 0.78) # 跳过右栏群成员列表(~80%+) 和滚动条 cap_bottom = bottom - max(100, int(win_h * 0.10)) # 输入框上方 if cap_right <= cap_left or cap_bottom <= cap_top: return [] im = ImageGrab.grab(bbox=(cap_left, cap_top, cap_right, cap_bottom)) if im.width < 50 or im.height < 50: return [] # 保存截屏到程序根目录用于调试（保留最近 20 张） try: script_dir = os.path.dirname(os.path.abspath(__file__)) ts = datetime.now().strftime('%Y%m%d_%H%M%S') path = os.path.join(script_dir, f"ocr_debug_{ts}.png") im.save(path) # 清理旧文件：只保留最近 20 张 _clean_old_ocr_debug(script_dir, keep=20) except Exception as e: logger.debug(f"[ocr] 保存截屏失败: {e}") # 图像预处理：提升 Windows OCR 识别准确率 # Windows OCR 对低 DPI/未处理截图识别差——CJK 空格、字符混淆 # 2x LANCZOS 放大 + 锐化 + 对比度增强 try: im = im.resize((im.width * 2, im.height * 2), Image.LANCZOS) im = ImageEnhance.Sharpness(im).enhance(1.8) im = ImageEnhance.Contrast(im).enhance(1.4) logger.debug(f"[ocr] 预处理: resize 2x + sharpen(1.8) + contrast(1.4)") except Exception as e: logger.debug(f"[ocr] 预处理失败(跳过): {e}") # ── 双引擎 OCR ────────────────────────────────────── # Windows OCR if im.mode != 'RGBA': im = im.convert('RGBA') sb = SoftwareBitmap(BitmapPixelFormat.RGBA8, im.width, im.height) sb.copy_from_buffer(im.tobytes()) win_result = engine.recognize_async(sb).get() win_text = win_result.text.strip() # EasyOCR 备用 easy_text = "" if _HAS_EASYOCR: self._init_easyocr() if self._easyocr_reader: try: img_np = _np.array(im) easy_results = self._easyocr_reader.readtext(img_np) easy_text = ' '.join([r[1] for r in easy_results]).strip() except Exception as e: logger.debug(f"[ocr] EasyOCR 识别失败: {e}") if win_text and easy_text: logger.debug(f"[ocr] Windows OCR: {win_text[:100]!r}...") logger.debug(f"[ocr] EasyOCR: {easy_text[:100]!r}...") raw_text = self._pick_ocr_text(win_text, easy_text, im) if not raw_text: return [] if self._is_ocr_console_garbage(raw_text): self._ocr_garbage_count += 1 if self._ocr_garbage_count >= 3: logger.info("[qq] OCR: 垃圾连续3次，清缓存") self._cached_hwnd.pop("qq", None) self._cached_pid.pop("qq", None) self._app_variant.pop("qq", None) self._ocr_garbage_count = 0 return [] self._ocr_garbage_count = 0 self._ocr_last_text["qq"] = raw_text ocr_lines = [l.strip() for l in raw_text.split('\n') if l.strip()] # ⭐ debug: 记录 OCR 原始行内容，定位过滤导致空白的原因 for i, rl in enumerate(ocr_lines): logger.info(f"[ocr] raw[{i}] raw_line<{rl!r}>") # QQ NT OCR 将所有可见消息合并为一段无换行文本，消息间以时间戳分隔 # 按时间戳分割恢复独立消息，避免整坨变成一条"新消息" expanded = [] for i, rl in enumerate(ocr_lines): before = rl rl = self._normalize_ocr_text(rl) logger.debug(f"[ocr] stage[{i}] normalize: len={len(before)} -> {rl[:120]!r}") # 先清理 & 碎片（OCR 把时间戳右边缘的 : 读成 &） # 必须在时间戳分割前做，否则 "13：5&07" 无法被时间戳正则捕获 rl = rl.replace('&', '') # 把全角 ：统一成半角 :，让后续正则统一匹配 rl = rl.replace('：', ':') logger.debug(f"[ocr] stage[{i}] clean: len={len(rl)} -> {rl[:120]!r}") # 分割"msg1 14:24 msg2 13:27 msg3" → ["msg1", "msg2", "msg3"] # 使用更宽松的 \d{1,2}（允许单数字），因为 & 移除后 "5&07"→"507" # 打破的时间戳如 "13:5 07" 也尽量捕获 rl = re.sub(r'\s+', ' ', rl) # 多重空格压为一个 # ⭐ 分割前 md 预检：rl 中 md 语法完整（链接、标题等），分割后碎片会丢失语法 if self._is_markdown_document(rl): logger.debug(f"[ocr] stage[{i}] markdown document detected pre-split, skip line") continue parts = re.split(r'(\d{1,2}:\d{1,2}(?::\d{1,2})?)', rl) logger.debug(f"[ocr] stage[{i}] split -> {len(parts)} parts: {[p[:60] for p in parts if p.strip()]}") buf = '' for part in parts: if re.match(r'^\d{1,2}:\d{1,2}', part): # 遇到时间戳 → 前面的 buf 为一条完整消息 if buf.strip(): expanded.append(buf.strip()) buf = '' else: buf += part if buf.strip(): expanded.append(buf.strip()) ocr_lines = expanded texts = [] for line in ocr_lines: # 注: OCR 行已在分割时 normalize+全角:转换+&清理，此处不再重复 if not line: continue # ⭐ QQ NT 界面固定噪声：整行匹配直接跳过 if line in ("查看更多", "收起", "选择会话开始聊天"): continue # 连续纯标点+符号（emoji、装饰字符等） if re.match(r'^[✿♡♥❤★☆✦✧❀❁❂❃❄❅❆❇❈❉❊❋➠➡➢➣➤.·•●○◦◎◉◍◐◑◒◓◔◕◖◗◢◣◤◥☺☻☼♠♣♦♤♧♡♢♔♕♖♗♘♙🕐-🕧\s]{1,3}$', line): continue # 去噪：去掉行首未读红点数字（如 "0"）和行尾删除按钮误识别（勖彐 等） # 否则同一消息每次 OCR 输出微小差异 → 精确去重失效 → 重复触发 line = re.sub(r'^0+', '', line) line = re.sub(r'[勖彐"\'\s\u2000-\u206f]+$', '', line) if not line: continue # 过滤 KB 预览文字（行起始）: "问题1：你叫什么名字答：你好我叫123" # 注意：行中 KB 预览在分割前 rl 阶段已按整行丢弃，此处只过滤以问题开头的残留 if re.match(r'^问题\d+\s*[:：]', line): continue # 过滤纯时间戳垃圾：整段只有数字、冒号、空格（如 "13:5 07 14:2 15"） if re.match(r'^[\d:\s]+$', line): continue # 半角全角数字混杂的界面碎片（如 "1.0s 0．35后监控应用设置"） line = re.sub(r'^\d+[\.．]\d+s?\s*\d*[．\.]?\d*', '', line) line = re.sub(r'^[\d．\.\s]{2,20}[一-鿿]{1,5}?[，。、]', '', line) if not line: continue if self._is_system_message(line): continue # 群成员侧栏/搜索残留（QQ/TIM 右栏成员列表） if re.search(r'群成员|索群成员|搜索成员', line): logger.debug(f"[qq] OCR 过滤: 群成员侧栏 line={line[:60]!r}") continue # 聊天面板顶部群名头栏：群名（角色〕/ 【群名】/ 群名[标签] # 特有签名：使用 `〕` `」` `】` 等非日常聊天中出现的闭合括号 if re.match(r'^[^）】〕」]{2,40}[（【〔「][^）】〕」]{1,15}[）】〕」]', line): logger.debug(f"[qq] OCR 过滤: 群名头栏 line={line[:60]!r}") continue # 纯中文短文本（1 字符）— UI 噪声碎片 if re.match(r'^[一-鿿]{1}$', line): continue if re.match(r'^\d{1,2}[:：\s]\d{1,2}', line): continue if re.match(r'^\d{4}[/-]\d{1,2}', line): continue texts.append(line) logger.debug(f"[qq] OCR -> {len(ocr_lines)} raw, {len(texts)} filtered, " f"last=\"{texts[-1][:50] if texts else ''}\"") return texts except Exception as e: logger.debug(f"[qq] OCR失败: {e}") return [] def _read_via_clipboard_once(self, platform: str) -> list: """剪贴板法单次执行（安全键盘导航：Enter + Ctrl+End + Ctrl+A/C，无 Esc）。 去重由 _poll 的 _replied_texts + _seen_messages 负责，本函数只过滤并返回。 回复回环防御由 _paste_and_send 写入 _replied_texts 承担。 """ hwnd = self._cached_hwnd.get(platform) if not hwnd or not win32gui.IsWindow(hwnd): return [] import win32clipboard as wc old = None try: wc.OpenClipboard() if wc.IsClipboardFormatAvailable(wc.CF_UNICODETEXT): old = wc.GetClipboardData(wc.CF_UNICODETEXT) wc.CloseClipboard() except Exception: pass texts = [] raw_lines = [] # 预初始化，防止 content 为空时 UnboundLocalError try: # 确保窗口在前台（失败则返回空，避免发送到错误窗口） # QQ 剪贴板优先路径需要更可靠的前台保证（与发送路径 retries 一致） if not self._ensure_foreground(hwnd, max_retries=5): return [] # PostMessage 点击消息区域（设焦点，不全局，不比鼠标移动） rect = win32gui.GetWindowRect(hwnd) if rect: left, top, right, bottom = rect w = right - left h = bottom - top # QQ NT 消息面板在 ~55-95% 宽度区域 click_x = int(w * 0.72) # 消息面板中部 click_y = int(h * 0.60) # 消息面板中部偏下 lparam = win32api.MAKELONG(click_x, click_y) win32gui.PostMessage(hwnd, win32con.WM_LBUTTONDOWN, win32con.MK_LBUTTON, lparam) win32gui.PostMessage(hwnd, win32con.WM_LBUTTONUP, 0, lparam) time.sleep(0.2) # 滚动到底部（PostMessage 安全，不切焦点，确保新消息可见） self._force_scroll_bottom(platform) time.sleep(0.2) # TIM/QQ NT: 安全键盘导航（无 Esc，不关窗口） # 每步前校验目标窗口仍在焦点，防止 keybd_event 泄露到控制台 VK_CONTROL = 0x11 variant = self._app_variant.get(platform, "") if variant in ("tim", "qq_nt"): VK_RETURN = 0x0D VK_END = 0x23 # Enter 激活当前对话 if self._fg_is_target(hwnd): ctypes.windll.user32.keybd_event(VK_RETURN, 0, 0, 0) time.sleep(0.05) ctypes.windll.user32.keybd_event(VK_RETURN, 0, 2, 0) time.sleep(0.3) # Ctrl+End 滚动到底部（双保险：PostMessage + keybd_event） if self._fg_is_target(hwnd): ctypes.windll.user32.keybd_event(VK_CONTROL, 0, 0, 0) ctypes.windll.user32.keybd_event(VK_END, 0, 0, 0) time.sleep(0.05) ctypes.windll.user32.keybd_event(VK_END, 0, 2, 0) ctypes.windll.user32.keybd_event(VK_CONTROL, 0, 2, 0) time.sleep(0.3) # Ctrl+A (全选) — 仅当目标窗口仍在焦点 VK_A = 0x41 if self._fg_is_target(hwnd): ctypes.windll.user32.keybd_event(VK_CONTROL, 0, 0, 0) ctypes.windll.user32.keybd_event(VK_A, 0, 0, 0) time.sleep(0.05) ctypes.windll.user32.keybd_event(VK_A, 0, 2, 0) ctypes.windll.user32.keybd_event(VK_CONTROL, 0, 2, 0) time.sleep(0.15) # Ctrl+C (复制) VK_C = 0x43 if self._fg_is_target(hwnd): ctypes.windll.user32.keybd_event(VK_CONTROL, 0, 0, 0) ctypes.windll.user32.keybd_event(VK_C, 0, 0, 0) time.sleep(0.05) ctypes.windll.user32.keybd_event(VK_C, 0, 2, 0) ctypes.windll.user32.keybd_event(VK_CONTROL, 0, 2, 0) time.sleep(0.3) wc.OpenClipboard() if wc.IsClipboardFormatAvailable(wc.CF_UNICODETEXT): content = wc.GetClipboardData(wc.CF_UNICODETEXT) if content: raw_lines = [l.strip() for l in content.strip().split('\n') if l.strip()] self._raw_clipboard[platform] = content wc.CloseClipboard() logger.debug(f"[{platform}] 剪贴板读取 → {len(raw_lines)} lines") # ── 过滤：只返回"像用户消息"的行 ── sender_pattern = re.compile(r'^[\u4e00-\u9fff·A-Za-z0-9_\-]{1,10}$') for line in raw_lines: # 跳过：系统消息、纯时间、纯数字、无意义短行 if self._is_system_message(line): continue # 跳过：控制台/Python 垃圾（剪贴板截错窗口内容） if self._is_ocr_console_garbage(line): continue # 跳过：单行发送者名（1-10汉字/字母/数字/下划线） if sender_pattern.match(line): continue # 跳过：纯时间格式（连字符/斜杠分格日期） if re.match(r'^\d{1,2}[:：\s]\d{2}', line): continue if re.match(r'^\d{4}[/-]\d{1,2}', line): continue # 跳过：纯 emoji / 纯标点（短于 6 个字符且只有符号） clean = re.sub('[~!@#$%^&*()_+\\-=\\[\\]{};:\\\'",.<>/?·！@#￥%……&*（）—+。，、；：""\'\'【】｛｝「」『』《》【】 \\t\\n\\r\\u2000-\\u206f\\u2190-\\u2bff\\U0001F000-\\U0001FFFF]', '', line) if len(clean) == 0 and len(line) <= 6: continue # 通过全部检查 → 保留 texts.append(line) logger.debug(f"[{platform}] 剪贴板过滤后 → {len(texts)} 条用户消息") except Exception as e: logger.debug(f"[{platform}] 剪贴板失败: {e}") finally: if old is not None: try: wc.OpenClipboard() wc.EmptyClipboard() wc.SetClipboardText(old, wc.CF_UNICODETEXT) wc.CloseClipboard() except Exception: pass return texts def get_raw_clipboard(self, platform: str) -> str: """获取最近一次剪贴板读取的原始文本。""" return self._raw_clipboard.get(platform, "") @staticmethod def _parse_last_sender(raw_text: str) -> Optional[str]: """从剪贴板原始文本中提取最后一条消息的发送者。 TIM/QQ 剪贴板格式通常是: [发送者] [时间] [消息] (空行) [发送者] [时间] [消息] """ if not raw_text: return None blocks = re.split(r'\n\s*\n', raw_text.strip()) if not blocks: return None last_block = blocks[-1].strip() if not last_block: return None lines = [l.strip() for l in last_block.split('\n') if l.strip()] if not lines: return None candidate = lines[0] # 排除时间格式 if re.match(r'^\d{1,2}[:：]\d{2}', candidate): return None if re.match(r'^\d{4}[/-]\d{1,2}', candidate): return None if len(candidate) > 30: return None return candidate def _search_contact(self, platform: str, contact: str) -> bool: """在 TIM/QQ 搜索框中搜索联系人并切换到对应会话。""" hwnd = self._cached_hwnd.get(platform) if not hwnd: return False try: win = auto.ControlFromHandle(hwnd) search = win.EditControl(Name="搜索", maxSearchSeconds=0.5) if not search.Exists(maxSearchSeconds=0.3): logger.warning(f"[{platform}] 找不到搜索框") return False # 恢复窗口并前置（失败则放弃搜索） if win32gui.IsIconic(hwnd): win32gui.ShowWindow(hwnd, win32con.SW_RESTORE) if not self._ensure_foreground(hwnd): logger.warning(f"[{platform}] 无法前置窗口，放弃搜索联系人") return False time.sleep(0.2) # SetFocus 到搜索框（不移动鼠标） search.SetFocus() time.sleep(0.2) # 全选已有内容并输入联系人名 auto.SendKeys('{Ctrl}a') time.sleep(0.1) auto.SendKeys(contact) time.sleep(0.5) # Enter 打开第一个结果 auto.SendKeys('{Enter}') time.sleep(0.5) logger.info(f"[{platform}] 已搜索并切换到联系人: {contact}") return True except Exception as e: logger.error(f"[{platform}] 搜索联系人失败: {e}") return False # ── SendInput 键盘模拟 ───────────────────────────────── @staticmethod def _key_down(vk: int): """SendInput KEYBDINPUT keydown.""" ctypes.windll.user32.keybd_event(vk, 0, 0, 0) @staticmethod def _key_up(vk: int): """SendInput KEYBDINPUT keyup (KEYEVENTF_KEYUP=2).""" ctypes.windll.user32.keybd_event(vk, 0, 2, 0) @staticmethod def _send_key_press(vk: int, delay: float = 0.05): """SendInput single key press (down + up).""" ctypes.windll.user32.keybd_event(vk, 0, 0, 0) time.sleep(delay) ctypes.windll.user32.keybd_event(vk, 0, 2, 0) @staticmethod def _send_ctrl_combo(vk: int): """Send Ctrl+: Ctrl down → key down → key up → Ctrl up.""" VK_CONTROL = 0x11 ctypes.windll.user32.keybd_event(VK_CONTROL, 0, 0, 0) ctypes.windll.user32.keybd_event(vk, 0, 0, 0) ctypes.windll.user32.keybd_event(vk, 0, 2, 0) ctypes.windll.user32.keybd_event(VK_CONTROL, 0, 2, 0) @staticmethod def _post_key(hwnd, vk, delay=0.05): """PostMessage 键盘到指定窗口（不全局，安全）。""" win32gui.PostMessage(hwnd, win32con.WM_KEYDOWN, vk, 0) time.sleep(delay) win32gui.PostMessage(hwnd, win32con.WM_KEYUP, vk, 0) @staticmethod def _post_ctrl_combo(hwnd, vk, delay=0.05): """PostMessage Ctrl+组合键到指定窗口（不泄露到控制台）。""" VK_CONTROL = 0x11 win32gui.PostMessage(hwnd, win32con.WM_KEYDOWN, VK_CONTROL, 0) win32gui.PostMessage(hwnd, win32con.WM_KEYDOWN, vk, 0) time.sleep(delay) win32gui.PostMessage(hwnd, win32con.WM_KEYUP, vk, 0) win32gui.PostMessage(hwnd, win32con.WM_KEYUP, VK_CONTROL, 0) def _fg_is_target(self, hwnd) -> bool: """校验目标窗口仍在焦点，防止 keybd_event 泄露到控制台。""" try: fg = win32gui.GetForegroundWindow() root = win32gui.GetAncestor(hwnd, win32con.GA_ROOT) return fg == root or (fg and win32gui.IsChild(root, fg)) except Exception: return False @staticmethod def _is_system_message(text: str) -> bool: """过滤机器人/系统消息，不作为用户消息处理。""" if not text: return True # 系统前缀/标记 system_markers = [ "✅", "❌", "📝", "🔍", "📄", "➡", "♡", "✿", "[WECHAT]", "[QQ]", "[wechat]", "[qq]", ] for m in system_markers: if text.startswith(m): return True # ⭐ QQ NT 界面噪声文字（按钮/提示/碎片，非用户消息） qq_ui_noise = [ "查看更多", "可查看在其他设备产生的聊天记录", "收藏", "转发", "多选", "引用", "回复", "选择会话开始聊天", ] for n in qq_ui_noise: if text == n: return True # 系统事件模板 — 注意：用 startswith 而非 in，避免混合文本（OCR 多行合一）误杀 system_patterns = [ "已自动回复", "知识库中未找到", "找到相关结果", "监控已启动", "监控已暂停", "设置已保存", "🤖", "找不到联系人", "当前会话回复", "自动回复触发", "匹配度", # KB 文档内容格式（机器人回复格式），OCR 可能把用户消息+KB 回复混成一行 # 只有全行以"答："开头才过滤，内含"答："的混合文本不过滤 "答：", # 纯系统命令反馈 "PS ", "❯ ", "$ ", ] for p in system_patterns: if text.startswith(p): return True # 纯时间格式（不含消息内容） if re.match(r'^\d{1,2}:\d{2}', text): return True return False @staticmethod def _is_markdown_document(text: str) -> bool: """检测是否为 Markdown 格式的文档内容。 Bot 回复用 md 格式（# 标题、[链接](url)、列表等），用户日常对话不用。 借此区分 bot 旧回复和历史堆积 vs 用户新消息。 """ if not text or len(text) < 10: return False # 行首是 # 标题（# ## ### #### 等） if re.match(r'^#{1,6}\s', text): return True # 包含 markdown 链接 [text](url) if re.search(r'\[.+?\]\(.+?\)', text): return True # 包含 .md 文件引用（如 "gin.md)" 分割后断开的链接） if re.search(r'\S+\.md\)', text): return True # 行首是 - 或 * 列表项（标准 md 列表） if re.match(r'^[\-\*]\s', text): return True # 包含代码块标记 ``` 或 `code` if re.search(r'`{1,3}', text): return True # 连续 3+ 行包含 markdown 特征：| 表格、--- 分隔线、> 引用 pipes = text.count('|') if pipes >= 3: return True # 内联 # # 标题标记（如 "# # 必装依赖" → 分割后 `# #` 或 `# #`） if re.search(r'#\s*#', text): return True return False def _paste_and_send(self, platform: str, text: str) -> bool: """粘贴文本到输入框并发送（剪贴板 + SendInput）。""" hwnd = self._cached_hwnd.get(platform) if not hwnd: return False import win32clipboard as wc # 保存剪贴板 old = None try: wc.OpenClipboard() if wc.IsClipboardFormatAvailable(wc.CF_UNICODETEXT): old = wc.GetClipboardData(wc.CF_UNICODETEXT) wc.CloseClipboard() except Exception: pass try: # 设置答案到剪贴板 wc.OpenClipboard() wc.EmptyClipboard() wc.SetClipboardText(text, wc.CF_UNICODETEXT) wc.CloseClipboard() # ── 更激进的窗口唤醒 ── # 1) 允许任意进程 SetForegroundWindow（移除系统限制） try: ctypes.windll.user32.AllowSetForegroundWindow(-1) # ASFW_ANY except Exception: pass # 2) 恢复/显示窗口 root_hwnd = win32gui.GetAncestor(hwnd, win32con.GA_ROOT) if win32gui.IsIconic(root_hwnd): win32gui.ShowWindow(root_hwnd, win32con.SW_RESTORE) time.sleep(0.3) if not win32gui.IsWindowVisible(root_hwnd): win32gui.ShowWindow(root_hwnd, win32con.SW_SHOWNORMAL) time.sleep(0.3) # 3) 试图 SetForegroundWindow + Alt 键锁（让当前线程获得用户输入资格） win32gui.SetForegroundWindow(root_hwnd) VK_MENU = 0x12 KEYEVENTF_KEYUP = 0x0002 ctypes.windll.user32.keybd_event(VK_MENU, 0, 0, 0) # Alt down ctypes.windll.user32.keybd_event(VK_MENU, 0, KEYEVENTF_KEYUP, 0) # Alt up time.sleep(0.2) # 4) 再 SetForegroundWindow + SwitchToThisWindow win32gui.SetForegroundWindow(root_hwnd) ctypes.windll.user32.SwitchToThisWindow(root_hwnd, True) time.sleep(0.3) # 5) 验证 foreground — 不成功则用 SendMessage 兜底 fg_now = win32gui.GetForegroundWindow() fg_ok = (fg_now == root_hwnd) or (fg_now and win32gui.IsChild(root_hwnd, fg_now)) if not fg_ok: logger.warning(f"[{platform}] 前台唤醒失败(fg={fg_now}, target={root_hwnd})，SendMessage 兜底") # SendMessage 方式：直接发消息给窗口，不依赖前台 rect = win32gui.GetWindowRect(hwnd) if rect: left, top, right, bottom = rect w = right - left h = bottom - top # QQ NT 输入框在右侧消息面板底部 click_x = int(w * 0.72) click_y = int(h * 0.92) lparam = win32api.MAKELONG(click_x, click_y) win32gui.SendMessage(hwnd, win32con.WM_LBUTTONDOWN, win32con.MK_LBUTTON, lparam) win32gui.SendMessage(hwnd, win32con.WM_LBUTTONUP, 0, lparam) time.sleep(0.1) # Ctrl+A (全选) win32gui.PostMessage(hwnd, win32con.WM_KEYDOWN, 0x41, 0 | (1 << 29)) # Ctrl 修饰 win32gui.PostMessage(hwnd, win32con.WM_KEYUP, 0x41, 0) time.sleep(0.05) # Ctrl+V (粘贴) win32gui.PostMessage(hwnd, win32con.WM_KEYDOWN, 0x56, 0 | (1 << 29)) win32gui.PostMessage(hwnd, win32con.WM_KEYUP, 0x56, 0) time.sleep(0.1) # Enter (发送) win32gui.PostMessage(hwnd, win32con.WM_KEYDOWN, 0x0D, 0) win32gui.PostMessage(hwnd, win32con.WM_KEYUP, 0x0D, 0) time.sleep(0.5) logger.info(f"[{platform}] 已回复(SendMessage): {text[:60]}") else: return False else: # ── 前台成功：PostMessage 点击输入框 + keybd_event 发送 ── rect = win32gui.GetWindowRect(hwnd) if rect: left, top, right, bottom = rect w = right - left h = bottom - top click_x = int(w * 0.72) click_y = int(h * 0.92) lparam = win32api.MAKELONG(click_x, click_y) win32gui.PostMessage(hwnd, win32con.WM_LBUTTONDOWN, win32con.MK_LBUTTON, lparam) win32gui.PostMessage(hwnd, win32con.WM_LBUTTONUP, 0, lparam) time.sleep(0.2) # keybd_event 发送（窗口已在前台） VK_A = 0x41 VK_V = 0x56 VK_BACK = 0x08 VK_RETURN = 0x0D self._send_ctrl_combo(VK_A) time.sleep(0.05) self._key_down(VK_BACK) self._key_up(VK_BACK) time.sleep(0.05) self._send_ctrl_combo(VK_V) time.sleep(0.3) self._key_down(VK_RETURN) self._key_up(VK_RETURN) time.sleep(0.5) logger.info(f"[{platform}] 已回复: {text[:60]}") # 记录回复文本到去重集 # 将回复文本加入 _replied_texts（永久集，永不裁切，必拦）和 seen reply_text = text.strip() self._replied_texts.add(reply_text) self._seen_messages.add(reply_text) # 添加换行符归一化变体（跨频道转发时会替换 \n → 空格） if '\n' in reply_text: self._replied_texts.add(reply_text.replace('\n', ' ')) self._replied_texts.add(reply_text.replace('\n', ' ').rstrip('.')) self._seen_messages.add(reply_text.replace('\n', ' ')) self._seen_messages.add(reply_text.replace('\n', ' ').rstrip('.')) for part in reply_text.split('\n'): part = part.strip() if part and len(part) > 3: self._replied_texts.add(part) self._seen_messages.add(part) self._last_reply_time[platform] = time.time() return True except Exception as e: logger.error(f"[{platform}] 回复失败: {e}\n{traceback.format_exc()}") return False finally: # 恢复原始剪贴板 if old is not None: try: wc.OpenClipboard() wc.EmptyClipboard() wc.SetClipboardText(old, wc.CF_UNICODETEXT) wc.CloseClipboard() except Exception: pass # ── 核心：读消息 ────────────────────────────────────────── def _read_messages(self, window, platform: str = "") -> list: """每轮重新定位 ListControl。""" if platform == "qq": # QQ NT: OCR 读屏优先（零依赖，不切焦点，不碰剪贴板） texts = self._ocr_read_qq() if texts: return texts # OCR 空 → 剪贴板兜底（窗口置前→滚底→全选→复制，内部自带重试） texts = self._read_via_clipboard_once(platform) if texts and len(texts) > 5: return texts # 剪贴板也空 → UIA 最后兜底（可能返回陈旧数据） fresh_win = self._find_qq_window_fresh() if fresh_win: texts = self._read_message_list(fresh_win) if texts: return texts texts = self._read_conversation_list(fresh_win) if texts: return texts try: rect = fresh_win.BoundingRectangle boundary = rect.left + rect.width() * 0.40 for lst in self._list_controls(fresh_win): try: cr = lst.BoundingRectangle if cr.left >= boundary: texts = self._extract_last_line(lst) if texts: logger.debug(f"[S-C] ListControl(x={cr.left}) → {len(texts)} items") return texts except Exception: pass except Exception: pass try: texts = self._collect_filtered_texts(fresh_win, "qq", max_depth=5) if texts: logger.debug(f"[S-D] 全窗文本 → {len(texts)} items") return texts except Exception: pass # Win32 兜底 texts = self._read_win32_texts(platform) if texts: return texts return [] # 微信: 优先读"消息"（ScrollPattern 滚底后实时） texts = self._read_message_list(window) if texts: return texts # 兜底：会话列表 texts = self._read_conversation_list(window) if texts: return texts try: rect = window.BoundingRectangle boundary = rect.left + rect.width() * 0.40 for lst in self._list_controls(window): try: cr = lst.BoundingRectangle if cr.left >= boundary: texts = self._extract_last_line(lst) if texts: logger.debug(f"[S-C] ListControl(x={cr.left}) → {len(texts)} items") return texts except Exception: pass except Exception: pass try: texts = self._collect_filtered_texts(window, "wechat", max_depth=5) if texts: logger.debug(f"[S-D] 全窗文本 → {len(texts)} items, last=\"{texts[-1][:50]}\"") return texts except Exception: pass return [] def _read_message_list(self, window) -> list: # 只接受窗口右半侧（>45%）的 ListControl，避开侧栏/文件列表用同名控件 try: wr = window.BoundingRectangle boundary = wr.left + wr.width() * 0.45 except Exception: boundary = 0 for name in ["消息", "Message"]: try: lst = window.ListControl(Name=name) if lst.Exists(maxSearchSeconds=0.2): try: cr = lst.BoundingRectangle if cr.left < boundary: continue # 左侧的"消息"列表（文件/笔记预览）跳过 except Exception: pass self._scroll_to_bottom(lst) texts = self._extract_last_line(lst) if texts: logger.debug(f"[S-A] ListControl({name}) → {len(texts)} items, last=\"{texts[-1][:50]}\"") return texts except Exception: pass return [] def _read_conversation_list(self, window) -> list: try: wr = window.BoundingRectangle boundary = wr.left + wr.width() * 0.45 except Exception: boundary = 0 for name in ["会话", "聊天", "Chat"]: try: lst = window.ListControl(Name=name) if lst.Exists(maxSearchSeconds=0.2): try: cr = lst.BoundingRectangle if cr.left < boundary: continue except Exception: pass self._scroll_to_bottom(lst) texts = self._extract_second_line(lst) if texts: logger.debug(f"[S-B] 会话({name}) → {len(texts)} items, last=\"{texts[-1][:50]}\"") return texts except Exception: pass return [] def _scroll_to_bottom(self, lst): """滚动 ListControl 到底部，触发应用加载新消息。""" try: scroll = lst.GetScrollPattern() if scroll: scroll.SetScrollPercent(100, 0) except Exception: pass @staticmethod def _extract_last_line(lst) -> list: texts = [] try: for item in lst.GetChildren(): try: name = item.Name if name and name.strip(): lines = [l.strip() for l in name.split("\n") if l.strip()] if lines: texts.append(lines[-1]) except Exception: pass except Exception: pass return texts @staticmethod def _extract_deep_text(lst) -> list: """深度递归提取 ListControl 内所有控件的 Name 文本（最多3层）。""" texts = [] def walk(ctrl, depth): if depth > 3: return try: name = ctrl.Name if name and name.strip(): texts.append(name.strip()) except Exception: pass try: for child in ctrl.GetChildren(): walk(child, depth + 1) except Exception: pass try: for item in lst.GetChildren(): walk(item, 0) except Exception: pass return texts @staticmethod def _extract_second_line(lst) -> list: texts = [] try: for item in lst.GetChildren(): try: name = item.Name if name and name.strip(): lines = [l.strip() for l in name.split("\n") if l.strip()] if len(lines) >= 2: texts.append(lines[1]) elif lines: texts.append(lines[-1]) except Exception: pass except Exception: pass return texts @staticmethod def _list_controls(window): results = [] try: for ctrl in window.GetChildren(): results.extend(_walk_list_controls(ctrl, 0)) except Exception: pass return results @staticmethod def _collect_filtered_texts(window, platform: str = "", max_depth=14) -> list: """深度遍历 + 位置过滤（排除搜索框/标题栏区域）。""" items = [] # [(text, rect_top, rect_left, depth)] def walk(ctrl, depth): if depth > max_depth: return try: cr = ctrl.BoundingRectangle if cr.width() < 20 or cr.height() < 20: return except Exception: return try: name = ctrl.Name if name and name.strip(): t = name.strip() if len(t) >= 2 and not t.isdigit(): items.append((t, cr.top, cr.left, depth)) except Exception: pass try: for child in ctrl.GetChildren(): walk(child, depth + 1) except Exception: pass try: for child in window.GetChildren(): walk(child, 1) except Exception: pass if not items: return [] try: w_rect = window.BoundingRectangle w_width = w_rect.width() w_height = w_rect.height() except Exception: w_width = 1390 w_height = 953 if platform == "qq": filtered = [ t for t, top, left, d in items if top > w_height * 0.08 ] seen = set() deduped = [] for t in filtered: if t not in seen: seen.add(t) deduped.append(t) return deduped return [t for t, _, _, _ in items] # ── 单次轮询 ───────────────────────────────────────────── def _poll(self, platform: str): # 自动回复发送中跳过轮询（防止窗口焦点被抢） if self._sending_event.is_set(): return # 冷却期：回复后 5 秒内跳过 poll，防止反馈回环 last_reply = self._last_reply_time.get(platform, 0.0) if time.time() - last_reply < 5.0: return counter = self._poll_count.get(platform, 0) + 1 self._poll_count[platform] = counter # 1. 确保窗口句柄（带超时保护，防止 UIA COM 调用挂死不返回） hwnd = self._ensure_hwnd_with_timeout(platform, timeout=12.0) if not hwnd: if counter % 10 == 0 or (counter == 1 and platform == "qq"): logger.debug(f"[{platform}] 窗口未找到") return # QQ 在托盘隐藏（IsWindowVisible=False）时跳过本轮。 # 任何尝试恢复隐藏窗口的操作（SW_SHOWNORMAL 等）在 QQ NT 上都会触发托盘预览弹窗， # 不会恢复主聊天窗口。用户保持 QQ 窗口最小化到任务栏即可。 if platform == "qq" and hwnd and win32gui.IsWindow(hwnd) and not win32gui.IsWindowVisible(hwnd): return # 2. 从 HWND 创建全新 Control（绕过缓存） win = self._fresh_window(platform) if not win: logger.debug(f"[{platform}] 创建窗口控件失败") return # 3. 读消息 texts = self._read_messages(win, platform) if not texts: empty = self._empty_reads.get(platform, 0) + 1 self._empty_reads[platform] = empty if counter % 5 == 0: logger.debug(f"[{platform}] 空读 #{empty}") if empty >= 5: self._cached_hwnd.pop(platform, None) self._cached_pid.pop(platform, None) self._empty_reads[platform] = 0 logger.info(f"[{platform}] 连续空读，清窗口缓存重试") return self._empty_reads[platform] = 0 # 冻结计数器（不在 early-return 里定义，保证所有路径都能访问） stale = 0 # 4a. 全文本签名快速跳过 — OCR 输出与上轮完全相同 → 无变化，跳过 sig = "||".join([t.strip() for t in texts if t.strip()]) if sig == self._last_sig.get(platform, ""): stale = self._stale_count.get(platform, 0) + 1 self._stale_count[platform] = stale return # 4. 用全局 _seen_messages 去重（跨平台、跨轮次防回环） # 精确匹配 + 模糊匹配（OCR 噪声导致精确匹配失效） seen = self._seen_messages replied = self._replied_texts new_items = [] for t in texts: ts = t.strip() if not ts: continue # _replied_texts 优先检查（永不裁切，100% 拦住自己的回复） if ts in replied: continue # _replied_texts 模糊匹配：OCR 噪声导致精确匹配失效（长文本尤甚） if self._fuzzy_duplicate(ts, replied, threshold=0.85): logger.debug(f"[{platform}] 回复模糊跳过: {ts[:50]}") continue if ts in seen: continue # 模糊匹配：OCR 输出每次微小差异（勖彐碎片、行首数字、尾部截断） # 精确匹配命中不了 → 用 difflib 相似度去重 if self._fuzzy_duplicate(ts, seen, threshold=0.80): logger.debug(f"[{platform}] 模糊重复跳过: {ts[:50]}") continue # 超长文本（>200字符）跳过——大概率是堆积历史或 bot 自己的回复，不是单条新消息 if len(ts) > 200: logger.debug(f"[{platform}] 超长文本跳过({len(ts)}c): {ts[:50]}...") continue if self._is_system_message(ts): continue # Markdown 文档过滤：bot 回复以 md 格式为主（#标题、[链接](url)、-列表项）， # 用户日常对话不用 md 语法。拦截 md 格式文本避免 bot 旧回复被重新处理。 if self._is_markdown_document(ts): logger.debug(f"[{platform}] md文档跳过: {ts[:50]}") continue # 纯装饰符号（单字符非汉字非字母数字）跳过 if len(ts) == 1 and not re.match(r'[\u4e00-\u9fff\w]', ts): continue # 只有符号/空格/emoji 的非消息文本（<3 字符且长度等于标点去掉后长度） clean = re.sub(r'[\s\u2000-\u206f\u2190-\u2bff\U0001F000-\U0001FFFF✿♡♥❤★☆]', '', ts) if len(ts) <= 3 and len(clean) == 0: continue seen.add(ts) new_items.append(ts) # _seen_messages 上限 500 条，超了裁到 300 # 防止集合无限膨胀 → 新消息模糊匹配时被历史噪声误杀 if len(seen) > 500: # 保留后 300 条（最新） self._seen_messages = set(list(seen)[-300:]) # 单次 poll 最多处理 3 条新消息，防止系统消息海量涌入触发回环 MAX_NEW_ITEMS = 3 new_items = new_items[:MAX_NEW_ITEMS] # 过滤纯时间/日期文本 → 不应进入 KB 匹配（OCR 时间戳分割遗留的噪音片段） # 这些文本在 KB 子串匹配阶段会命中文件内任何含相同数字的内容 new_items = [ it for it in new_items if not re.match(r'^\d{1,2}[:：]\d{2}(?:[:：]\d{2})?$', it.strip()) and not re.match(r'^\d{4}[/-]\d{1,2}', it.strip()) ] # 更新 last_texts 与签名 self._last_texts[platform] = list(texts) self._last_sig[platform] = sig self._stale_count[platform] = 0 if new_items: # 有纯粹的新文本 → 逐一报告 for item in new_items: self._last_msg[platform] = item logger.info(f"[{platform}] 新消息: {item[:60]}") if self._on_message: self._on_message(platform, item) if platform == "qq": if stale == 3: logger.debug(f"[QQ] 冻结 #{stale}，强制刷新") self._force_uia_refresh(platform) elif stale == 6: logger.debug(f"[QQ] 冻结 #{stale}，重建窗口引用") self._cached_hwnd.pop(platform, None) self._cached_pid.pop(platform, None) else: if stale == 5: logger.debug(f"[{platform}] 冻结 #{stale}，尝试强制刷新") self._force_uia_refresh(platform) elif stale == 10: logger.debug(f"[{platform}] 冻结 #{stale}，重建窗口引用") self._cached_hwnd.pop(platform, None) self._cached_pid.pop(platform, None) def _clean_old_ocr_debug(dir_path: str, keep: int = 20): """清理 ocr_debug_*.png 文件，只保留最近 keep 张。""" try: files = [f for f in os.listdir(dir_path) if f.startswith("ocr_debug_") and f.endswith(".png")] if len(files) <= keep: return files.sort() # 按字母序 = 时间序 for f in files[:-keep]: os.remove(os.path.join(dir_path, f)) except Exception: pass def _walk_list_controls(ctrl, depth): if depth > 6: return [] results = [] try: cname = ctrl.ControlTypeName except Exception: return results if cname == "ListControl": results.append(ctrl) return results try: for child in ctrl.GetChildren(): results.extend(_walk_list_controls(child, depth + 1)) except Exception: pass return results