Engines overview

z4j supports six engines, each with its own adapter package (z4j-<engine>).

Engine	Package	Broker	Strengths	Default scheduler
Celery	z4j-celery	Redis / RabbitMQ / SQS	Massive ecosystem, chords/groups/chains	celery-beat
RQ	z4j-rq	Redis	Simple, small surface, Python-only	rq-scheduler
Dramatiq	z4j-dramatiq	Redis / RabbitMQ	Reliable, middleware-first	APScheduler
Huey	z4j-huey	Redis / Sqlite / in-memory	Minimal deps, great for Django	built-in periodic
arq	z4j-arq	Redis	Async-native, great for FastAPI	arq cron
taskiq	z4j-taskiq	Redis / RabbitMQ / NATS / in-memory	Modern async, pluggable brokers	taskiq-scheduler

Installing

Pick one or more. They can coexist in the same process:

pip install z4j-celery z4j-rq z4j-dramatiq

Capability matrix

Feature	Celery	RQ	Dramatiq	Huey	arq	taskiq
Retry (native)	✓	✓	✓	✓	polyfill	✓
Cancel (native)	partial	✓	-	partial	✓	✓
Bulk retry	polyfill	polyfill	polyfill	polyfill	polyfill	polyfill
Queue purge	✓	✓	✓	✓	✓	✓
Chord / group aware	✓	-	-	-	-	-
Result store visibility	✓	✓	partial	✓	✓	✓
Native scheduler	celery-beat	rq-scheduler	-	built-in	cron jobs	taskiq-scheduler

“polyfill” = brain implements via re-enqueue + cancel. Users see no difference in the UI.

Event capture strategy

Each adapter uses the engine’s native signal / middleware / hook system:

Engine	Capture mechanism
Celery	task_sent / task_prerun / task_postrun / task_retry / task_failure signals
RQ	Queue.enqueue_call monkey-patch + job lifecycle hooks
Dramatiq	custom Middleware that hooks before_process_message / after_process_message
Huey	signal handlers (SIGNAL_ENQUEUED / SIGNAL_EXECUTING / SIGNAL_EXECUTED)
arq	on_job_start / on_job_end hooks on the worker
taskiq	middleware: TaskiqMiddleware.pre_send / post_execute

Patches are additive - they don’t replace your existing signal handlers.

Per-engine pages

See the left sidebar - one page per engine with install, config, events, actions, and caveats.