create-pykit

Interactive CLI to scaffold Python data science projects — powered by uv.

Usage

npx create-pykit
npx create-pykit my-project
npx create-pykit my-project --data /path/to/data.csv
npx create-pykit my-project --url https://example.com/data.csv

What it does

1. Asks a few questions (project name, template, database)
2. Scaffolds the project
3. Runs uv sync automatically
4. If --data is provided — executes eda.ipynb immediately so outputs are already saved when you open it

Features

Base dependencies (always included)

Every project ships with pandas, numpy, matplotlib, seaborn, jupyterlab, and python-dotenv out of the box.

Templates

| Template | Additional deps | |---|---| | Data Analysis | (base only) | | Machine Learning | scikit-learn, xgboost | | Deep Learning | torch, torchvision | | NLP | transformers, spaCy | | Computer Vision | opencv-python, torchvision, Pillow | | Custom | pick AI/ML deps yourself |

CSV / Data analysis

| Flag | Description | |---|---| | --data <path> | Analyze a local CSV and pre-fill the notebook with real column names, types, and auto-generated EDA | | --url <url> | Download a CSV from a URL then analyze it the same way |

When a CSV is detected, the notebook is generated with:

• Missing value heatmap
• Histogram + boxplot per numeric column (with IQR outlier table)
• Bar chart + pie chart per categorical column
• Monthly trend + row count for date columns
• Correlation matrix with high-correlation alerts (|r| > 0.7)
• Summary export → data/eda_summary.txt

Database support

Selected database generates db/connection.py + .env boilerplate, and injects an Export to CSV cell at the top of the notebook that clones every table/collection into data/db_export/*.csv.

Local | | Driver | |---|---| | SQLite | sqlalchemy | | DuckDB | duckdb |

Self-hosted | | Driver | |---|---| | PostgreSQL | sqlalchemy + psycopg2 | | MySQL | sqlalchemy + pymysql | | MongoDB | pymongo |

Cloud | | Driver | |---|---| | MongoDB Atlas | pymongo | | Supabase | supabase-py | | Neon | sqlalchemy + psycopg2 | | Firebase Firestore | firebase-admin | | Redis | redis-py | | Prisma ORM | prisma Python client |

Project structure

my-project/
├── data/
│   ├── your-data.csv        # copied from --data / --url
│   └── db_export/           # tables exported from DB (if DB selected)
├── notebooks/
│   └── eda.ipynb            # auto-generated EDA, outputs pre-executed
├── db/
│   └── connection.py        # DB connection boilerplate (if DB selected)
├── pyproject.toml
├── Makefile
├── .env                     # credentials you entered (gitignored)
├── .env.example             # placeholder values (safe to commit)
└── .gitignore

Makefile targets

make install      # uv sync
make run          # uv run jupyter lab notebooks/
make db-init      # uv run python db/connection.py
make db-generate  # uv run prisma generate  (Prisma only)
make clean        # remove .venv, __pycache__, .ruff_cache

Requirements

• Node.js >= 18
• uv installed

Dev

pnpm install
pnpm dev my-project
pnpm dev my-project --data ./sample.csv
pnpm dev my-project --url https://example.com/data.csv

Publish

pnpm build
npm publish