README.md

plotly web app

Setup

• Create an environment and install the required dependencies:

  conda create -n plotly-web-app python=3.8 -y
  conda activate plotly-web-app
  pip install -r requirements.txt
  pip install -e .

• Local test of the app:
  Dash uses Flask as the web framework. The underlying Flask app is available at app.server

  export FLASK_APP=app:server
  flask run

• Speed up the app
  You have probably noticed that figures do not update instantly when you change some parameters. Instead of splitting the data, calculating the roc-auc scores, generating and plotting the new figures for every change in the data splitting parameters, we can generate, dump and load all possible figures and roc-auc scores.

  Generate the content by executing:

  python create_content.py

  This content is used in app_2.py:

  export FLASK_APP=app_2:server
  flask run

Deployment

Azure Web app deployment notes (do not use this service)

• In web-app -> Configuration -> General settings set the startup command to gunicorn -b 0.0.0.0 app_2:server or to gunicorn -b 0.0.0.0 app:server depending on what you want to run.

Use google app engine (do not use this service)

• Create a app.yaml that contains info about the runtime and about the entrypoint. Then use the gcloud cli:

  gcloud init
  # go into the right project
  gcloud app deploy ./app.yaml -Y

Use a docker image running on ubuntu VM

• Build the image:
  In case you want to deploy the app somewhere else:

  docker build -t plotly_app:1.0 . --build-arg runtime_path="."

  Launch a container to test the app:

  docker run --rm --name dash_app -d -p 80:80 plotly_app:1.0

  If everything is fine, add the image to the dockerhub container registry:

    docker login -u "$DOCKER_HUB_USR" \
                 -p "$DOCKER_HUB_PWD"
    
    docker tag plotly_app:1.0 "${DOCKER_HUB_USR}"/plotly_app:1.0
    
    docker push "${DOCKER_HUB_USR}"/plotly_app:1.0

• Setup a ubuntu machine to run the docker container:

  chmod +x vm_docker_setup.sh
  ./vm_docker_setup.sh