pipeline/           # Data processing
├── transcript.py   # Video transcription
├── llm.py          # LLM integration
├── slides.py       # Slide extraction
└── main.py         # Pipeline orchestration

frontend/src/       # React frontend
├── components/     # UI components
└── App.js          # Main app

• Pipeline modules are independent
• Frontend uses React patterns
• LLM calls isolated in llm.py
• No direct API calls in components

• Pipeline: *.py
• Components: *.js in components/
• Services: *_service.js

• Max 300 lines per file (warn at 300)
• Max 30 lines per function (warn at 30)
• Max nesting depth ≤ 4
• Max 3 parameters per function

If code exceeds these limits, REFACTOR immediately using Extract Function or Extract Class.

1. Identify frontend vs pipeline
2. Keep pipeline functions pure
3. Frontend follows React patterns
4. Keep files under 300 lines, functions under 30 lines
5. Run: npm run lint / python -m pytest