arvel-workflow

Durable execution for the arvel framework — backed by Temporal, with arvel's ergonomics instead of Temporal's setup tax.

Some work outlives a single request: an order that moves through authorize → fulfil → ship over hours, a multi-step job that has to survive a restart halfway through. arvel's queue is fire-and-forget; this is for work that must remember where it got to. You write plain workflows and activities, run one command, and drive them through a facade — no Temporal boilerplate.

uv add 'arvel-workflow[temporal]'

Installing auto-registers the provider through arvel's entry-point mechanism — no wiring.

The idea in one screen

An activity does the I/O (and can use any arvel facade); a workflow orchestrates activities and stays pure. Both autoload from their folders.

# app/activities/greet.py
from arvel_workflow import activity
from arvel import Cache

@activity
async def greet(name: str) -> str:
    count = await Cache.increment("greet:count")   # activities do I/O — arvel DI works here
    return f"hello {name} (#{count})"

# app/workflows/greet.py
from arvel_workflow import workflow, run_activity
from app.activities.greet import greet          # a plain import — no ceremony

@workflow
class Greet:
    async def run(self, name: str) -> str:
        return await run_activity(greet, name)

Run the worker (--dev boots a throwaway local Temporal server — nothing to install first):

arvel workflow:work --dev

Start work and read the result from anywhere you have the app:

handle = await app.make("workflow").start(Greet, "world")
print(await app.make("workflow").result(handle))   # hello world (#1)

No with workflow.unsafe.imports_passed_through() anywhere — the framework configures Temporal's sandbox so plain imports work.

Documentation

Scope & roadmap

This release ships the durable-execution core: define workflows and activities, run a worker, start work, and read its result. Documented capabilities are the ones that actually run today.

Coming next (not shipped yet — so the docs never promise what the code can't do):

• signals / queries / updates — send events into a running workflow, read its live state, drive human-in-the-loop approvals
• child workflows & fan-out, sagas / compensation, durable timers, scheduled workflows
• per-activity retry & timeout sugar and a time-skipping test harness

Requirements

• Python 3.14+
• arvel (installed with the package)
• The engine extra: arvel-workflow[temporal]. --dev downloads a local Temporal dev-server binary on first run.

License

MIT — see LICENSE.