How jakeson99/rugai works
Main Components
A. Data Collection Layer
• Web scraping components
• Both selenium_scraper.py and simple_scraper.py as alternative scrapers
• Configuration management (.env file)
B. Model Layer
• ResNet model architecture
• Weights directory
• Training workflow (notebook)
• Integration with Weights & Biases for experiment tracking
C. Application Layer
• Streamlit UI component
• Inference pipeline
• Preprocessing utilities
• Configuration management
Component Relationships
[Data Collection] → [Model Training/Storage] → [Application Serving]
Specific Implementation Details
• Rectangles: Core components (Scraper, Model, UI)
• Hexagons: External services (Streamlit Cloud, Weights & Biases)
Technical Stack Highlights
Technology badges or labels for:
• Python
• Streamlit
• CNN/ResNet
• Selenium
• Poetry (dependency management)
• MPS (M1 GPU acceleration)
Key Flows
A. Training Flow:
Web Sources → Scrapers → Training Data → Model Training → Weights Storage
B. Inference Flow:
User Upload → Preprocessing → Model Inference → Classification Result → UI Display
Important Elements
• Configuration Management: Show how config.yaml and environment variables are used
• Model Artifacts: Indicate where weights are stored and loaded
• Deployment: Show the Streamlit Cloud hosting
• Development Environment: Indicate Poetry and devcontainer setup