jupyter-fly-deploy

An experimental Dockerfile based method to deploying multiple Jupyter notebooks to a Fly.io machine.

There's no OAuth and user management. Unexpected things could happen if multiple people use the server at the same time.

Usage

0. Use this repository as a template (button on GitHub) to create your own
   copy of it.

1. Setup an account on fly.io

2. Install flyctl

3. Run `fly auth login``

4. Run fly launch --generate-name --no-deploy.

   Answer y to use existing config.

5. Remove project_a/ directories and create separate directories for your own
   notebooks. The directories should contain a requirements.txt if it
   contains a notebook directory. It can be empty if you don't need any
   additional dependencies other than Jupyter notebook.

6. The Jupyter starts each time with a random token which could be a problem,
   if you'd like to share URLs with others. You could instead set the
   JUPYTER_TOKEN environment variable using the fly secrets set JUPYTER_TOKEN="some-long-secret" command.

7. Run fly deploy to deploy your changes.

8. You can share links to individual notebooks using the token you set above, as follows:

   https://<app-name>.fly.dev/notebooks/<dir>/<notebook_name>.ipynb?token=some-long-secret
   

9. (Optional) You can setup the app to be automatically
   deployed
   on future pushes to GitHub using GitHub actions by creating a Fly token and
   adding it to the repository's settings. The GitHub workflow file is already
   present.