Example deploying a Panel app to Heroku using Docker

The default build environment on Heroku uses pip to install Python packages. Some Python packages cannot be installed via pip, particularly if they need to be compiled (cartopy is one example).

This example uses Heroku and Docker containers to install dependencies using a conda environment instead of pip.

To get started working with Heroku signup for a free account and download and install the CLI. Once you are set up follow the instructions to log into the CLI.

1. Log in to heroku on the command line by running heroku login. This should prompt you to open a browser and you can log in with your credentials.

2. Create you app with heroku. From the main folder of your repository run:

heroku create APP_NAME

3. Configure the heroku build step to use Docker.

heroku stack:set container

4. Update the environment.yml file to include all of the dependencies that your app notebook needs.

5. Update the run command in heroku.yml to use the APP_NAME from the previous step:

--allow-websocket-origin=APP_NAME.herokuapp.com

Also, make sure the name of the

web: panel serve --address="0.0.0.0" --port=$PORT app.ipynb --allow-websocket-origin=app-name.herokuapp.com

In this repository, I have specified "APP_NAME" as "gentle-earth-18442" but you should replace your "app-name" with whatever you want yours to be.

6. Add the files to the repository:

git add .
git ci -a -m 'adding files to repository'

7. Push the app to heroku and wait until it is deployed. From the main folder of your repository, run the following command from the command line:

git push heroku main

Note! the process of creating the conda environment can take a long time (> 10 minutes), expecially if you have a lot of packages to install.

7. Visit the app at APP_NAME.herokuapp.com