From 0316afa299b641be06709f6c0b4def6afeb15797 Mon Sep 17 00:00:00 2001
From: wxg0103 <727495428@qq.com>
Date: Wed, 9 Jul 2025 20:34:34 +0800
Subject: [PATCH] refactor: enhance HTML tag removal in text processing to
 exclude audio, video, and image tags

---
 apps/common/utils/common.py | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/apps/common/utils/common.py b/apps/common/utils/common.py
index 0b9732b5c..6d12e803a 100644
--- a/apps/common/utils/common.py
+++ b/apps/common/utils/common.py
@@ -116,6 +116,9 @@ def markdown_to_plain_text(md: str) -> str:
     text = re.sub(r'\n{2,}', '\n', text)
     # 使用正则表达式去除所有 HTML 标签
     text = re.sub(r'<[^>]+>', '', text)
+    # 先移除特定媒体标签（优先级高于通用HTML标签移除）
+    text = re.sub(r'<(audio|video)[^>]*>.*?</\1>', '', text, flags=re.DOTALL)  # 匹配音频/视频标签
+    text = re.sub(r'<img[^>]*>', '', text)  # 匹配图片标签
     # 去除多余的空白字符（包括换行符、制表符等）
     text = re.sub(r'\s+', ' ', text)
     # 去除表单渲染