README.md

---
title:  AI Content Classifier Server
emoji: 🐢
colorFrom: pink
colorTo: blue
sdk: gradio
sdk_version: 5.13.1
app_file: app.py
pinned: false
short_description: Classify text as human-written, AI-generated, or paraphrased
---


# AI Content Classifier API
A FastAPI backend service that classifies text as human-written, AI-generated, or paraphrased using a fine-tuned transformer model.

## Features

- **详细文本分类**: 将输入文本分类为三个类别：
  - Human-Written (人类写作)
  - AI-Generated (AI生成)
  - Paraphrased (改写文本)
- **完整概率分布**: 返回所有类别的概率百分比
- **深度文本分析**: 
  - 文本统计信息（长度、词数、句数）
  - AI生成指标识别
  - 风险等级评估
- **智能改进建议**: 基于分析结果提供具体的改进建议
- **置信度分析**: 返回预测的置信度评分
- **RESTful API**: 清洁的REST端点和标准HTTP状态码
- **输入验证**: 验证输入文本并优雅处理错误
- **健康检查**: 服务健康状态监控端点

## Installation

1. Install dependencies:
```bash
pip install -r requirements.txt
```

2. Run the FastAPI server:
```bash
python app.py
```

The server will start on `http://localhost:7860`

## API Endpoints

### POST `/detect`
Classify text content.

**Request Body:**
```json
{
    "text": "Your text to classify here"
}
```

**Response:**
```json
{
    "classification": "Human-Written",
    "confidence": 0.9234,
    "probabilities": {
        "Human-Written": 0.9234,
        "AI-Generated": 0.0566,
        "Paraphrased": 0.0200
    },
    "analysis": {
        "text_length": 245,
        "word_count": 42,
        "sentence_count": 3,
        "ai_indicators": [],
        "human_indicators": ["自然语言特征", "人类写作模式"],
        "risk_level": "low"
    },
    "suggestions": [
        "保持当前的自然写作风格",
        "继续保持语言的自然流畅性"
    ]
}
```

### GET `/health`
Health check endpoint.

**Response:**
```json
{
    "status": "healthy"
}
```

### GET `/`
Root endpoint with service information.

## Usage Example

```python
import requests

response = requests.post(
    "http://localhost:7860/detect",
    json={"text": "This is sample text to classify"}
)

result = response.json()

# 基本结果
print(f"分类结果: {result['classification']}")
print(f"置信度: {result['confidence']:.4f}")

# 详细概率分布
print("\n概率分布:")
for category, prob in result['probabilities'].items():
    print(f"  {category}: {prob*100:.2f}%")

# 文本分析
analysis = result['analysis']
print(f"\n文本分析:")
print(f"  风险等级: {analysis['risk_level']}")
print(f"  词数: {analysis['word_count']}")

# 改进建议
print(f"\n改进建议:")
for i, suggestion in enumerate(result['suggestions'], 1):
    print(f"  {i}. {suggestion}")
```

## Testing

Run the detailed test script to verify the API is working:

```bash
python test_detailed_api.py
```

This will show you:
- 完整的分类结果和概率分布
- 详细的文本分析
- 个性化的改进建议
- 可视化的概率条形图

## Interactive Documentation

Once the server is running, visit:
- Swagger UI: `http://localhost:7860/docs`
- ReDoc: `http://localhost:7860/redoc`

## Model Information

This service uses the `vai0511/ai-content-classifier` model from Hugging Face, which is fine-tuned for content source identification.