Running services locally

Services can be operated locally (e.g. for testing or ad-hoc data processing). You can:

  • Run your service once (i.e. run one analysis):

    • Via the CLI

    • By using the octue library in a python script

  • Start your service as a child, allowing it to answer any number of questions from any other Octue service:

    • Via the CLI

Running a service once

Via the CLI

  1. Ensure you’ve created a valid octue.yaml file for your service

  2. If your service requires inputs, create an input directory with the following structure

    |---  values.json    (if input values are required)
    |---  manifest.json  (if an input manifest is required)
  3. Run:

    octue run --input-dir=my_input_directory

Any output values will be printed to stdout and any output datasets will be referenced in an output manifest file named output_manifest_<analysis_id>.json.

Via a python script

Imagine we have a simple app that calculates the area of a square. It could be run locally on a given height and width like this:

from octue import Runner

runner = Runner(app_src="path/to/", twine="path/to/twine.json")
analysis ={"height": 5, "width": 10})

>>> {"area": 50}

>>> None

See the Runner API documentation for more advanced usage including providing configuration, children, and an input manifest.

Starting a service as a child

Via the CLI

  1. Ensure you’ve created a valid octue.yaml file for your service

  2. Run:

    octue start

This will run the service as a child waiting for questions until you press Ctrl + C or an error is encountered. The service will be available to be questioned by other services at the service ID organisation/name as specified in the octue.yaml file.


You can use the --timeout option to stop the service after a given number of seconds.