vercel-dramatiq

Dramatiq broker backed by Vercel Queue Service.

# tasks.py
import dramatiq

from vercel.integrations.dramatiq import VercelQueueBroker

broker = VercelQueueBroker()
dramatiq.set_broker(broker)


@dramatiq.actor
def send_email(user_id: str) -> None: ...

Constructing the broker explicitly is the recommended pattern: it always takes effect and the broker options are visible at the construction site. To store actor results in Vercel Runtime Cache, add the results middleware:

from dramatiq.results import Results

from vercel.integrations.dramatiq import VercelRuntimeCacheBackend

broker.add_middleware(Results(backend=VercelRuntimeCacheBackend()))

Alternatively, install_vercel_dramatiq_integration() sets Dramatiq's global broker to a default VercelQueueBroker with the results middleware when no broker has been configured yet. Note that this is a no-op once a global broker exists, so broker options passed to the installer do not apply in environments that install the integration first (Vercel functions do).

The broker uses push delivery when VERCEL is truthy in the environment and poll delivery otherwise. Pass poll=False or poll=True to VercelQueueBroker(...) to force a mode.

For Vercel push delivery, declare a module that exposes the broker as a queue subscriber in pyproject.toml:

# worker.py
# Importing tasks declares the actors' queues on the broker.
from tasks import broker

__all__ = ["broker"]

[[tool.vercel.subscribers]]
entrypoint = "worker:broker"

The Vercel build introspects every queue declared on the broker and compiles the subscriber into a queue-triggered function. That includes each queue's Dramatiq delay queue: the default queue maps to both the default topic and the sanitized default_DDQ delay topic, and retries and delayed messages are delivered through the delay queue topic.

Set VERCEL_DRAMATIQ_DEBUG=1 to enable debug logging for the integration and Dramatiq worker loggers.

The package is standalone and depends on vercel-queue and Dramatiq.