Byeongjun Moon, 2024, bj.moon@usc.edu

A machine learning-powered tool that analyzes GitHub repository documentation quality and suggests improvements using state-of-the-art language models.

• Analyzes README quality using DistilBERT for semantic understanding
• Evaluates code-documentation alignment using CodeBERT
• Generates enhanced documentation using OPT-125m (optimized for Apple Silicon)
• Provides detailed section-by-section analysis with quality scores
• Suggests actionable improvements based on best practices
• Web interface for easy interaction

src/
├── models/                 # Core ML models and analyzers
│   ├── code_documentation_analyzer.py  # CodeBERT-based code analysis
│   ├── quality_enhancer.py            # OPT-125m-based enhancement
│   ├── unified_scorer.py              # Combined scoring system
├── trainers/              # Model training scripts
└── main.py               # Application entry point

1. Clone the repository:

git clone https://github.com/yourusername/docugen.git
cd docugen

2. Create and activate a virtual environment:

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

3. Install dependencies:

pip install -r requirements.txt

4. Set up environment variables:

cp .env.example .env
# Edit .env with your GitHub token

1. Start the web interface:

python src/main.py

2. Open your browser and navigate to:

http://127.0.0.1:7862

3. Enter a GitHub repository URL to analyze

• Used for semantic analysis of documentation content
• Evaluates code-documentation alignment
• Measures documentation completeness and quality
• Analyzes docstrings and code comments

• Lightweight model optimized for Apple Silicon
• Generates contextual documentation improvements
• Memory-efficient operation
• Enhanced performance on M1/M2 chips

• Custom model for quality scoring
• Fine-tuned on high-quality documentation examples
• Evaluates clarity, completeness, and structure
• Provides section-specific quality metrics

1. Add your GitHub token to .env

2. Run the data collection script:

python src/models/getting_data.py

3. Train the README quality model:

python src/trainers/train_readme_model.py

4. Run the main script to start the web interface:

python src/main.py