A CLI tool for verifying third-party LLM API authenticity through fingerprint testing.
一个通过指纹测试验证第三方 LLM API 真伪的命令行工具。
English:
- Fingerprint Testing - Uses carefully designed riddles to detect model characteristics
- Multi-Model Detection - Supports GPT-4, GPT-3.5, and Llama model detection
- Async Concurrent Testing - Fast parallel API testing for efficiency
- Confidence Scoring - Provides detailed confidence scores and analysis
- Extensible Architecture - Easy to add new models and detection strategies
中文:
- 指纹测试 - 使用精心设计的谜题检测模型特征
- 多模型检测 - 支持 GPT-4、GPT-3.5 和 Llama 模型检测
- 异步并发测试 - 快速并行 API 测试,提高效率
- 置信度评分 - 提供详细的置信度评分和分析
- 可扩展架构 - 轻松添加新模型和检测策略
# Clone the repository | 克隆仓库
git clone https://github.com/PerryLink/llm-detective.git
cd llm-detective
# Install dependencies | 安装依赖
pip install -e .# Basic usage | 基本使用
sherlock check --key "your-api-key" --url "https://api.openai.com/v1" --model "gpt-4"
# Example output | 示例输出
# ✅ VERIFIED (Confidence: 95.5%)
# All responses match GPT-4 patterns--key- Your API key | 你的 API 密钥--url- API endpoint URL | API 端点 URL--model- Model name to test | 要测试的模型名称
llm-detective/
├── src/llm_detective/
│ ├── models.py # Data models | 数据模型
│ ├── config.py # Configuration | 配置管理
│ ├── detectors/ # Model detectors | 模型检测器
│ │ ├── base.py # Detector protocol | 检测器协议
│ │ ├── registry.py # Detector registry | 检测器注册表
│ │ ├── gpt4.py # GPT-4 detector | GPT-4 检测器
│ │ ├── gpt35.py # GPT-3.5 detector | GPT-3.5 检测器
│ │ └── llama.py # Llama detector | Llama 检测器
│ ├── strategies/ # Analysis strategies | 分析策略
│ │ ├── base.py # Strategy protocol | 策略协议
│ │ └── confidence.py # Confidence strategy | 置信度策略
│ ├── fingerprints/ # Fingerprint database | 指纹数据库
│ │ └── riddles.py # Test riddles | 测试谜题
│ ├── analyzer.py # Response analyzer | 响应分析器
│ ├── examiner.py # API examiner | API 检查器
│ └── cli.py # CLI interface | 命令行界面
├── tests/ # Test suite | 测试套件
├── pyproject.toml # Project config | 项目配置
├── LICENSE # Apache 2.0 License | 许可证
├── CONTRIBUTING.md # Contribution guide | 贡献指南
└── README.md # This file | 本文件
- Python 3.9+ - Core language | 核心语言
- OpenAI SDK - API client | API 客户端
- Typer - CLI framework | 命令行框架
- Rich - Terminal formatting | 终端格式化
- Pytest - Testing framework | 测试框架
English: The project uses a Strategy Pattern + Registry Architecture for extensibility:
- Detectors - Pluggable model detection strategies
- Registry - Dynamic detector management
- Strategies - Configurable analysis algorithms
- Fingerprints - Universal test pattern format
中文: 项目采用策略模式 + 注册表架构实现高扩展性:
- 检测器 - 可插拔的模型检测策略
- 注册表 - 动态检测器管理
- 策略 - 可配置的分析算法
- 指纹 - 通用测试模式格式
# Run all tests | 运行所有测试
pytest -v
# Run with coverage | 运行覆盖率测试
pytest --cov=llm_detective --cov-report=htmlThis project is licensed under the Apache License 2.0 - see the LICENSE file for details.
本项目采用 Apache License 2.0 许可证 - 详见 LICENSE 文件。
Copyright 2026 Chance Dean (novelnexusai@outlook.com)
Contributions are welcome! Please see CONTRIBUTING.md for details.
欢迎贡献! 详见 CONTRIBUTING.md。
- GitHub: @PerryLink
- Email: novelnexusai@outlook.com
Made with ❤️ by Chance Dean