The bundled script will look in an incoming folder in Google drive for PDF files, run them through the pool processing pipeline and write output to another folder. An output index PDF and summary CSV are written to the processed file folder.

Create a service account and download JSON-formatted credentials for it. Add the email address associated with that service account as owner on the Tem drives containing the incoming and output folders.

The JSON credentials should be present in a file called credentials.json.

Create a jobspec.yaml file in the current directory based on the example template. The incoming and processed folder ids may be cut-and-paste from URLs in the browser.

The script is best run from a Docker image since it requires a lot of additional software installed:

$ docker build -t rjw57/winterpool-experiment .

Run the script via:

$ docker run --rm -it \
    -v $PWD/jobspec.yaml:/jobspec.yaml:ro \
    -v $PWD/credentials.json:/credentials.json:ro \
    rjw57/winterpool-experiment --spec=/jobspec.yaml

An example stack for deployment to a Docker swarm is included. This assumes that one is running the traefik traffic manager and that one can serve traffic for the host pool.swarm.usvc.gcloud.automation.uis.cam.ac.uk.

A docker config must be present which holds the jobspec and a docker secret must be present with the client secrets. Authorisation tokens are stored in a persistent volume. See the stack for details.