paperboy

A web frontend for scheduling Jupyter Notebooks as reports

Overview

Paperboy is a production-grade application for scheduling reports. It has a flexible architecture and extensible APIs, and can integrate into a wide variety of deployments. It is composed of various industrial-strength technologies from the open source world.

• Jupyter Notebooks for the reports themselves

- Jupyter notebooks are an ideal report template, and with NBConvert support a wide variety of output types, including PDFs, HTML, Emails, etc

• Papermill to parameterize notebooks
• SQLAlchemy for Storage (default)
• Apache Airflow for Scheduling (default)

- Dask for Airflow Workers - Luigi as a scheduling alternative for Airflow (relies on cron)

• PhosphorJS for the frontend
• Support for Python Virtualenvs via requirements.txt or custom Docker images via Dockerfiles on a per-notebook level
• Traitlets parameterization of storage and scheduler classes for easy integration with custom storage backends and custom schedulers
• Single click notebook deployment with Voila and Dokku

Process Flow

• Upload notebook
• Configure job

- start time - interval - papermill parameters to autoconfigure reports - if autoconfiguring reports from papermill: - run or publish - output - notebook - pdf - html - email - script - strip or keep code

• To edit or create additional reports on a job, configure reports

- run or publish - output - notebook - pdf - html - email - script - strip or keep code

Installation from source

Paperboy requires Python and Node.js, which can be installed from conda-forge if conda is available.

Clone the repository and run following commands to install and launch the application:

• npm install
• npm run build
• pip install -e .
• python -m paperboy

Visit http://0.0.0.0:8080 in a browser to view the application.

The default authentication backend requires the registration of a username that can be used on subsequent launches.