From b08363551ac50f05b736e446bbe5a256baf26c0d Mon Sep 17 00:00:00 2001
From: youpaishidifu <youpaishidifu@git.ypshidifu.cn>
Date: Fri, 14 Aug 2026 10:10:16 +0800
Subject: [PATCH] =?UTF-8?q?fix(monitor):=20Poll=E9=98=B6=E6=AE=B5=E4=B9=9F?=
 =?UTF-8?q?=E6=94=BE=E5=AE=BD=E5=81=8F=E6=97=81=E8=BF=87=E6=BB=A4?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

与OCR阶段保持一致，从"包含偏旁字符就丢弃"改为"偏旁字符占比>10%才丢弃"
防止OCR误识别（默讠/亻尔）导致有效消息被丢弃
---
 monitor.py | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/monitor.py b/monitor.py
index beb948e..2f87d18 100644
--- a/monitor.py
+++ b/monitor.py
@@ -7025,8 +7025,12 @@ class MessageMonitor:
             # === 偏旁碎片过滤：包含偏旁字符就丢弃 ===
             # 偏旁字符（U+2E80-U+2FDF）在正常中文聊天中不会出现
             # OCR 拆字碎片如"亻牛""彡"只要出现就是碎片
-            if any(_is_radical_char(ch) for ch in ts):
-                logger.debug(f"[{platform}] Poll过滤: 包含偏旁字符(丢弃) ts={ts!r}")
+            # ⭐ 2026-08-14 修复：放宽过滤，仅当偏旁字符占比>10%时才丢弃
+            #   原因：OCR 误识别会把"默认"识别成"默讠"、把"个"识别成"亻尔"
+            #   这些是单个偏旁的 OCR 噪声，不应因此丢弃整行
+            radical_count = sum(1 for ch in ts if _is_radical_char(ch))
+            if radical_count > 0 and radical_count / max(len(ts), 1) > 0.1:
+                logger.debug(f"[{platform}] Poll过滤: 偏旁字符过多(占比{radical_count/max(len(ts),1)*100:.0f}%) ts={ts!r}")
                 continue
             # === 系统配置内容过滤：SMTP/邮件地址/代码片段不是用户消息 ===
             if re.search(r'@example|smtp|\.py\b|\.bat\b|\.sh\b|//.*\.\w{2,4}\b', ts, re.IGNORECASE):