Add ContentGuardian - Content Audit Agent with Hanyu Xinxie Style

README.md

@@ -1,11 +1,103 @@
 ---
-title: ContentGuardian
-emoji: 🏃
-colorFrom: purple
-colorTo: gray
-sdk: docker
+title: ContentGuardian - Content Audit Agent
+emoji: 🛡️
+colorFrom: blue
+colorTo: purple
+sdk: gradio
+sdk_version: 4.44.0
+app_file: app.py
 pinned: false
-license: apache-2.0
+license: mit
+short_description: China · Simplified Chinese · Hanyu Xinxie Style Content Audit Agent
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# 🛡️ ContentGuardian - Content Audit Agent
+
+**China · Simplified Chinese · Hanyu Xinxie Style · Text-based Version**
+
+An intelligent content audit system that integrates Chinese culture with modern design concepts, embodying the restrained elegance of Chinese aesthetics.
+
+## ✨ Key Features
+
+- 🚫 **Inappropriate Advertising Language Detection**: Detects 32 common violation words
+- 🔍 **Precise Keyword Marking**: Marks user-specified keyword positions
+- 📊 **Structured Audit Reports**: Standardized text output format
+- 🎨 **Hanyu Xinxie Aesthetic Style**: Embodies the restrained elegance of Chinese culture
+- ⚡ **Stable and Reliable**: Avoids segmentation faults, ensures system stability
+
+## 🔍 Detection Scope
+
+### Inappropriate Advertising Language (32 types)
+
+- **Absolute terms**: 绝对, 完全, 100%, 永远, 从不, 必须, 一定
+- **Superlative terms**: 第一, 最好, 最佳, 最优, 最强, 最大, 最小
+- **Exaggerated terms**: 极度, 超级, 顶级, 王牌, 冠军, 霸主
+- **Time-sensitive terms**: 立即, 马上, 瞬间, 秒杀, 急速, 飞速
+- **Mystical terms**: 神奇, 奇迹, 万能, 无敌, 完美, 终极
+
+### Keyword Marking
+
+- Precisely locates keywords in text
+- Counts keyword occurrences
+- Supports multiple keyword marking simultaneously
+
+## 📊 Report Format
+
+The system generates audit reports containing:
+
+- **Audit Overview**: Total issues, severity level, text length, audit time
+- **Inappropriate Advertising Language Detection**: Detailed words, occurrence counts, position information
+- **Keyword Marking**: Keyword occurrence counts and positions
+- **Detailed Analysis**: Compliance recommendations, risk level, handling suggestions
+- **Design Philosophy**: Reflects the cultural connotations of Hanyu Xinxie
+
+## 🎨 Design Philosophy
+
+This system adopts the Hanyu Xinxie style, integrating Chinese culture with modern design:
+
+- **Cultural Connotation**: Embodies the restrained elegance of Chinese culture
+- **Visual Aesthetics**: Uses ASCII art borders and dividers
+- **Text Expression**: Focuses on accuracy and standardization of expression
+- **Structured Output**: Clear information organization and presentation
+
+## 🌟 Example
+
+**Input Text:**
+```
+This product is absolutely effective, completely side-effect free, the first brand! Buy immediately, instant results!
+```
+
+**Keywords:**
+```
+product,effect
+```
+
+**Output Report:**
+```
+╔══════════════════════════════════════════════════════════════╗
+║                    📋 内容审核报告                           ║
+║                  中国·简体中文·汉语新解风格                   ║
+╚══════════════════════════════════════════════════════════════╝
+
+📊 审核概览
+├─ 总问题数：5 处
+├─ 严重程度：高
+├─ 文本长度：95 字符
+└─ 审核时间：2024-01-01 12:00:00
+...
+```
+
+## 🚀 Usage
+
+1. **Input Text**: Enter the text content to be audited
+2. **Keyword Marking**: Enter keywords to mark (separated by commas)
+3. **Start Audit**: Click the "Submit" button
+4. **View Report**: The system generates a detailed structured audit report
+
+## 📄 License
+
+MIT License
+
+---
+
+**ContentGuardian** - Guarding content quality, embodying cultural beauty

app.py

@@ -0,0 +1,194 @@
+"""
+ContentGuardian - Content Audit Agent
+China · Simplified Chinese · Hanyu Xinxie Style · Text-based Version
+Optimized for Hugging Face Spaces
+"""
+import gradio as gr
+import datetime
+
+def comprehensive_text_audit(text, keywords=""):
+    """
+    Comprehensive text audit - structured text output
+    """
+    if not text.strip():
+        return "❌ Please enter text content"
+    
+    # Parse keywords
+    keyword_list = [k.strip() for k in keywords.split(",") if k.strip()] if keywords else []
+    
+    # 1. Inappropriate advertising language detection
+    inappropriate_words = [
+        "绝对", "完全", "100%", "永远", "从不", "必须", "一定",
+        "第一", "最好", "最佳", "最优", "最强", "最大", "最小",
+        "极度", "超级", "顶级", "王牌", "冠军", "霸主",
+        "立即", "马上", "瞬间", "秒杀", "急速", "飞速",
+        "神奇", "奇迹", "万能", "无敌", "完美", "终极",
+        "absolutely", "completely", "perfect", "ultimate", "best", "first",
+        "immediately", "instantly", "magical", "miracle", "supreme"
+    ]
+    
+    found_inappropriate = []
+    for word in inappropriate_words:
+        if word.lower() in text.lower():
+            positions = []
+            start = 0
+            text_lower = text.lower()
+            word_lower = word.lower()
+            while True:
+                pos = text_lower.find(word_lower, start)
+                if pos == -1:
+                    break
+                positions.append(pos)
+                start = pos + 1
+            if positions:
+                found_inappropriate.append({
+                    "word": word,
+                    "count": len(positions),
+                    "positions": positions
+                })
+    
+    # 2. Keyword marking
+    found_keywords = []
+    for keyword in keyword_list:
+        if keyword.lower() in text.lower():
+            positions = []
+            start = 0
+            text_lower = text.lower()
+            keyword_lower = keyword.lower()
+            while True:
+                pos = text_lower.find(keyword_lower, start)
+                if pos == -1:
+                    break
+                positions.append(pos)
+                start = pos + 1
+            if positions:
+                found_keywords.append({
+                    "keyword": keyword,
+                    "count": len(positions),
+                    "positions": positions
+                })
+    
+    # 3. Generate structured report
+    total_issues = len(found_inappropriate)
+    severity = "低" if total_issues < 3 else "中" if total_issues < 6 else "高"
+    
+    # Build standardized text report
+    report = f"""
+╔══════════════════════════════════════════════════════════════╗
+║                    📋 内容审核报告                           ║
+║                  中国·简体中文·汉语新解风格                   ║
+╚══════════════════════════════════════════════════════════════╝
+
+📊 审核概览
+├─ 总问题数：{total_issues} 处
+├─ 严重程度：{severity}
+├─ 文本长度：{len(text)} 字符
+└─ 审核时间：{datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S')}
+
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+
+🚫 不当广告用语检测 ({len(found_inappropriate)} 处)
+"""
+    
+    if found_inappropriate:
+        for i, item in enumerate(found_inappropriate, 1):
+            report += f"""
+   {i}. 词汇："{item['word']}"
+      └─ 出现次数：{item['count']} 次
+      └─ 位置：{', '.join(map(str, item['positions']))}
+"""
+    else:
+        report += "\n   ✅ 未发现不当广告用语\n"
+    
+    report += f"""
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+
+🔍 关键词标记 ({len(found_keywords)} 处)
+"""
+    
+    if found_keywords:
+        for i, item in enumerate(found_keywords, 1):
+            report += f"""
+   {i}. 关键词："{item['keyword']}"
+      └─ 出现次数：{item['count']} 次
+      └─ 位置：{', '.join(map(str, item['positions']))}
+"""
+    else:
+        report += "\n   ℹ️ 未指定关键词或关键词未出现\n"
+    
+    report += f"""
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+
+📋 详细分析
+├─ 合规建议：
+"""
+    
+    if found_inappropriate:
+        report += f"""│  • 建议修���或删除 {len(found_inappropriate)} 个不当用词
+│  • 使用相对性表述替代绝对化用词
+│  • 避免夸大宣传和误导性表述
+"""
+    else:
+        report += "│  • 内容表述规范，符合广告法要求\n"
+    
+    report += f"""├─ 风险等级：{severity}
+└─ 处理建议：{"需要重点关注和修改" if severity == "高" else "建议适当调整" if severity == "中" else "可以正常使用"}
+
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+
+🎨 设计理念
+本审核系统融合中华文化与现代设计，体现内敛典雅之美。
+采用汉语新解风格，注重文字的准确性和表达的规范性。
+
+╔══════════════════════════════════════════════════════════════╗
+║  审核完成 - 感谢使用 ContentGuardian 内容审核智能体          ║
+╚══════════════════════════════════════════════════════════════╝
+"""
+    
+    return report
+
+# Create Gradio interface optimized for HF Spaces
+demo = gr.Interface(
+    fn=comprehensive_text_audit,
+    inputs=[
+        gr.Textbox(
+            label="Text to Audit",
+            placeholder="Please enter the text content to be audited...",
+            lines=8
+        ),
+        gr.Textbox(
+            label="Keywords (Optional)",
+            placeholder="Please enter keywords to mark, separated by commas",
+            lines=2
+        )
+    ],
+    outputs=gr.Textbox(
+        label="Audit Report",
+        lines=25,
+        max_lines=30
+    ),
+    title="🛡️ ContentGuardian - Content Audit Agent",
+    description="""
+    **China · Simplified Chinese · Hanyu Xinxie Style · Text-based Version**
+    
+    This system uses standardized structured text output, integrating Chinese culture with modern design concepts.
+    Detects inappropriate advertising language, marks specified keywords, and generates detailed audit reports.
+    
+    **Key Features:**
+    - 🚫 Inappropriate Advertising Language Detection (32+ common violation words)
+    - 🔍 Precise Keyword Marking
+    - 📊 Structured Audit Reports
+    - 🎨 Hanyu Xinxie Aesthetic Style
+    """,
+    examples=[
+        ["This product is absolutely effective, completely side-effect free, the first brand! Buy immediately, instant results!", "product,effect"],
+        ["Our product quality is very good, trustworthy, welcome to purchase.", "product,quality"],
+        ["Buy now, instant effect, 100% effective, absolutely satisfying!", "buy,effect"],
+        ["这款产品效果绝对好，完全无副作用，第一品牌！立即购买，马上见效！", "产品,效果"]
+    ],
+    theme=gr.themes.Soft(),
+    flagging_mode="never"
+)
+
+if __name__ == "__main__":
+    demo.launch()

requirements.txt

@@ -0,0 +1 @@
+gradio>=4.0.0