# Ds Notebook Refactor

> Bridging the gap between exploratory Jupyter notebooks and production-ready Python modules.

- Skill: `jcorpac/ds-notebook-refactor` (Agent Skill)
- Install (CLI): `npx skillmds@latest add jcorpac/ds-notebook-refactor`
- Raw SKILL.md: https://api.skillmd.com/api/skills/jcorpac/ds-notebook-refactor/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: jcorpac (https://skillmd.com/u/jcorpac)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/jcorpac/ds-notebook-refactor

---


# Notebook Refactor

Jupyter notebooks are great for research but risky for production. This skill helps you safely migrate code out of cells and into testable modules.

## The Refactor Checklist
1. **Identify Logic**: Find the cells that actually perform data transformations (ignore the plotting cells for now).
2. **Extract Functions**: Move cell code into well-named functions with type hints.
3. **Handle Config**: Extract hard-coded paths and variables into a `config.yaml` or `.env` file.
4. **Parameterize**: Ensure your main script can take arguments (e.g., input path).
5. **Unit Test**: Write a small test for your new function using representative dummy data.

## Before vs. After
- **Before**: A notebook with 50 cells, global variables, and `print()` statements everywhere.
- **After**: A single `pipeline.py` that imports functions from `utils.py` and logs progress.

## Best Practices
- Use a `.ipynb` to `.py` exporter for a head start, but always manually clean the result.
- Keep your plotting logic separate from your transformation logic.