Running Tasks

Run tasks, order dependencies, and inspect failures

Use marvin.run when you only need the result. Create a Task when you need to inspect state or declare dependencies.

import marvin

task = marvin.Task("What is 23 + 19?", result_type=int)
answer = task.run()
assert task.is_successful()

Dependencies

import marvin

outline = marvin.Task("Outline a short guide to Python type hints")
draft = marvin.Task("Write the guide from the outline", depends_on=[outline])
marvin.run_tasks([outline, draft])
print(draft.result)

The orchestrator includes dependencies and waits until they are complete. Complete includes failed or skipped tasks, not just successful tasks. Decide how your application should handle those outcomes.

run_tasks returns the task objects. run_tasks_async and Task.run_async are available for async code.

Failures

By default, a failed task raises an error. Pass raise_on_failure=False to inspect the outcome yourself:

import marvin

task = marvin.Task("Try the requested task; fail if it cannot be completed", allow_fail=True)
task.run(raise_on_failure=False)
if task.is_failed():
    print(task.result)

The state field is task.state, not task.status. Use is_pending, is_running, is_successful, is_failed, is_skipped, and is_complete rather than comparing strings.

See task results for result types and custom validation.