AI Question Testing System

A web-based system for data annotation and quality testing of professional domain questions using Hunyuan AI API.

Features

• Question Input Interface: Web form for contributors to input questions with LaTeX support
• AI Testing System: Automated testing with 8 stateless API calls per question
• Answer Verification: AI-powered verification of correctness
• Qualification System: Questions with success rate < 50% are marked as qualified
• Excel Export: Export results in standardized 10-column format

Technology Stack

• Backend: Flask, SQLAlchemy, PostgreSQL/SQLite
• Frontend: Bootstrap 5, MathJax (LaTeX rendering)
• AI Integration: OpenAI Python SDK (Hunyuan API)
• Export: openpyxl

Installation

1. Clone the repository:

cd question-testing-system

2. Create virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

4. Configure environment variables:

cp .env.example .env
# Edit .env and add your Hunyuan API key

5. Initialize database:

flask db init
flask db migrate -m "Initial migration"
flask db upgrade

Configuration

Edit .env file with your settings:

FLASK_APP=run.py
FLASK_ENV=development
SECRET_KEY=your-secret-key-here

DATABASE_URL=sqlite:///questions.db

HUNYUAN_API_KEY=your-api-key-here
HUNYUAN_BASE_URL=https://api.hunyuan.cloud.tencent.com/v1
HUNYUAN_MODEL=hunyuan-turbos-latest

TEST_ATTEMPTS=8
QUALIFICATION_THRESHOLD=50
MAX_ANSWER_LENGTH_MATH=40
MAX_ANSWER_LENGTH_OTHER=50

Usage

1. Start the application:

python run.py

2. Open browser and navigate to http://localhost:5000

3. Add questions via the web interface

4. Run tests on questions (8 stateless API attempts per question)

5. View test results and export qualified questions to Excel

Project Structure

question-testing-system/
├── app/
│   ├── __init__.py              # Flask app factory
│   ├── config.py                # Configuration
│   ├── models.py                # Database models
│   ├── routes/
│   │   ├── question_routes.py  # Question CRUD
│   │   └── testing_routes.py   # Testing & export
│   ├── services/
│   │   ├── hunyuan_service.py  # API integration
│   │   ├── testing_service.py  # Testing logic
│   │   └── export_service.py   # Excel export
│   ├── templates/               # HTML templates
│   └── static/                  # CSS/JS files
├── exports/                     # Generated Excel files
├── requirements.txt
├── run.py                       # Application entry point
└── README.md

Database Schema

Questions Table

• Basic info: title, type, subject, difficulty
• Content: question_text (LaTeX), standard_answer, solution_approach
• Metadata: knowledge_points, timestamps

Test Results Table

• Test metrics: correct_count, success_rate, qualified status
• Difficulty status: "X/8" format

API Call Logs Table

• Individual attempt details
• AI answers and verification responses
• Error tracking

API Integration

The system uses stateless API calls to ensure varied responses:

• Each call is independent (no conversation history)
• Temperature > 0 for response variation
• 0.5s delay between calls for rate limiting
• Automatic retry with exponential backoff

Excel Export Format

10-column format:

1. 标题 (Title)
2. 题目类型 (Question Type)
3. 领域 (Subject)
4. 难度 (Difficulty)
5. 知识点 (Knowledge Points)
6. 问题 (Question Text)
7. 答案 (Standard Answer)
8. 解题思路 (Solution Approach)
9. (Empty Column)
10. 查难情况 (Difficulty Status: X/8)

License

MIT License