Write workflows in Python (PythonWorkflow)

Rbbt workflows are implemented using Scout’s Workflow engine. In modern Scout-based installs you can:

define the workflow in Ruby (inputs/dependencies/persistence/provenance)
implement the task logic as plain Python functions (with type hints)

This is provided by the PythonWorkflow integration (from scout-rig).

Mental model

A Python task is a normal Python function.
You register it with scout.task(...).
The Ruby workflow imports it with python_task.
At runtime the Ruby task executes the Python script in a subprocess (so you can use your normal Python stack).

Minimal directory layout

A typical workflow ships:

workflow.rb
python/task/hello.py

1) Define a Python task

Create python/task/hello.py:

import scout

def hello(name: str, excited: bool = False) -> str:
    """Greet a user."""
    return f"Hello, {name}{'!' if excited else ''}"

scout.task(hello)

Notes:

Type hints and defaults are used to generate task metadata.
The docstring becomes the task description.

2) Import the Python task into a Ruby workflow

Create workflow.rb:

require 'scout'

module TestPythonWF
  extend Workflow
  extend PythonWorkflow

  self.name = 'TestPythonWF'

  python_task :hello
end

This defines a Scout/Rbbt task named hello whose implementation is in python/task/hello.py.

3) Run the task

Once the workflow is discoverable (depends on your installation/layout), you can run it like any other workflow task:

rbbt workflow task TestPythonWF hello --name Alice --excited

Advanced notes (what PythonWorkflow supports)

Metadata mode

A Python task file can be executed directly to print machine-readable metadata:

python python/task/hello.py --scout-metadata

Return value decoding

The Python runner prints the return value to stdout.

If it’s JSON, Ruby parses it.
Otherwise it is treated as a string (or split into lines for array-like declared outputs).

List inputs

List inputs can usually be passed as:

a Ruby Array
a comma-separated string
a file path whose lines are taken as items