kryndex / commuter

🚎 share, publish, and discover notebooks

com·mut·er

/kəˈmyo͞odər/ a person who travels some distance to work on a regular basis.

Opinionated nteract focused server. Commuter reads notebooks from S3, has a directory explorer, and provides a jupyter compatible version of the contents API.

Demo

Development

Requires Node.js 6+ and npm 3+.

1. git clone [email protected]:nteract/commuter.git
2. npm install
3. COMMUTER_BUCKET=<name> COMMUTER_S3_KEY=<key> COMMUTER_S3_SECRET=<secret> npm start

Available env options

COMMUTER_BUCKET (required, without s3://)
COMMUTER_S3_KEY (required)
COMMUTER_S3_SECRET (required)
COMMUTER_ES_HOST (required only for discovery api)
COMMUTER_BASEPATH (optional, prefix for s3 bucket)
COMMUTER_PATH_DELIMITER (optional, defaults to "/")
COMMUTER_PORT (optional, defaults to 4000)

Notes

The API server (express) runs on port 4000 while the client (UI) runs on port 3000. The client uses the webpack dev server and proxies 3000 -> 4000 to avoid CORS issues. On production the express App serves index.html and static assets.

View the dashboard at http://localhost:3000/<S3_PATH> and the API at http://localhost:4000/api/contents/<S3_PATH>

Project uses prettier for code formatting (npm run format:code and package.json has more options).

Test

1. npm test

Deployment

coming soon...

ROADMAP

This roadmap is organized into stages of development, leading towards a backend for (mostly) real-time collaboration.

Stage I

• List and Load notebooks from S3
  • Bucket, etc. loaded from configuration (e.g. COMMUTER_BUCKET=xyz)
  • Roles or Amazon environment variables automatically picked up (via aws-sdk)
• Tree view of notebook content
• Render page using notebook-preview

Stage II

• Start outlining an authentication and permissions strategy
• Provide/use kernels from configured source (e.g. tmpnb.org, jupyterhub, or your private setup)
• Render page using nteract/nteract components

Stage III

• Save notebooks back to S3
• Delete notebooks

Stage IV

• Create server side in-memory model of notebook and transient models, push to clients