axiom-annotations 0.1.0

Zero-dependency workflow/task/Step/AutoStep annotation markers used by axiom-graph.

axiom-annotations

Zero-dependency Python package providing the shared annotation vocabulary used by axiom-graph.

Contains:

• @workflow and @task decorators
• Step() and AutoStep() inline markers
• StepValidationError exception
• Per-marker argument validation helpers

All markers are pure Python with no third-party runtime dependencies. This lets downstream tools annotate their code with workflow/task/Step/AutoStep markers without pulling in any heavier optional runtime stack.

Install

pip install axiom-annotations

Usage

from axiom_annotations import workflow, task, Step, AutoStep

@task(purpose="Load data from file")
def load_data(path: str):
    ...

@workflow(purpose="Run the analysis pipeline")
def run_pipeline(path: str):
    ax = AutoStep(step_num=1)
    data = load_data(path)

    ax = Step(step_num=2, name="Filter", purpose="Remove bad rows")
    data = [r for r in data if r.quality > 0.5]

Registry hook

By default @workflow and @task just validate and pass the function through. Tools that want to collect decorated functions into a registry can install a hook:

from axiom_annotations.decorators import set_register_hook

def my_register(func, *, purpose, kind, inputs, outputs, critical):
    ...  # persist to your own registry

set_register_hook(my_register)

axiom-graph installs such a hook during scanning to populate its index.