Release Notes¶

0.4.6¶

Fixed¶

Due to some internals, Lilya was missing for the dashboard dependency and causing a regression.

Added Lilya as dependency for AsyncMQ and remove several redundant others.

0.4.5¶

Fixed¶

Lazy loading of the settings via Monkay on stores.

0.4.4¶

Fixed¶

Lazy loading of the settings via Monkay.

0.4.3¶

Added¶

Python 3.14 support.

0.4.2¶

Added¶

Add config option for custom JSON loads/dumps #73 by chbndrhnns.
No negative numbers on workers list pagination #69 by chbndrhnns.

Changed¶

Accept a Redis client instance for RedisBackend #67 by chbndrhnns.

Fixed¶

Lazy loading for the global settings was pointing to the wrong location.
Error when parsing Boolean values from environment variables.
Do not reload dashboard test files #64 by chbndrhnns.
No negative numbers on workers list pagination #68 by chbndrhnns.

0.4.1¶

Changed¶

In the past, AsyncMQ was using dataclass to manage all the settings but we found out that can be a bit cumbersome for a lot of people that are more used to slightly cleaner interfaces and therefore, the internal API was updated to stop using @dataclass and use directly a typed Settings object.

Replace Settings to stop using @dataclass and start using direct objects instead.

Example before

from dataclasses import dataclass, field
from asyncmq.conf.global_settings import Settings


@dataclass
class MyCustomSettings(Settings):
    hosts: list[str] = field(default_factory=lambda: ["example.com"])

Example after

from asyncmq.conf.global_settings import Settings


class MyCustomSettings(Settings):
    hosts: list[str] = ["example.com"]

This makes the code cleaner and readable.

0.4.0¶

Added¶

Address session middleware from the dashboard config and can be changed.
Make cli commands more consistent across the client.
start_worker now uses the Worker object. This was previously added but never 100% plugged into the client.

Fixed¶

Register/De-register workers in run_worker.
CLI: Catch RuntimeError when running commands via anyio
Log exceptions during job execution.
Out-of-process workers do not pick up jobs from the queue
index.html had a typo for the <script />

0.3.1¶

Added¶

RabbitMQBackend
- Full implementation of the BaseBackend interface over AMQP via aio-pika:
  - enqueue / dequeue / ack
  - Dead-letter queue support (move_to_dlq)
  - Delayed jobs (enqueue_delayed, get_due_delayed, list_delayed, remove_delayed)
  - Repeatable jobs (enqueue_repeatable, list_repeatables, pause_repeatable, resume_repeatable)
  - Atomic flows & dependency resolution (atomic_add_flow, add_dependencies, resolve_dependency)
  - Queue pause/resume, cancellation/retry, job-state/progress/heartbeat tracking
  - Broker-side stats (queue_stats) and queue draining (drain_queue)
RabbitMQJobStore
- Implements BaseJobStore, delegating persistence to any other store (e.g. RedisJobStore), so metadata stays flexible.

Changed¶

Unified on the default exchange for enqueueing and DLQ publishing

Fixed¶

Proper wrapping of store results into DelayedInfo and RepeatableInfo dataclasses
Correct scheduling semantics using epoch timestamps (time.time())

0.3.0¶

Fixed¶

Enqueue/delay was not returning a job id.
Backend was not returning the id on enqueue.

Breaking change¶

This barely affects you but the order of parameters when the .enqueue(backend, ...) happens, it is now .enqueue(..., backend=backend)

Example¶

Before:

import anyio
from asyncmq.queues import Queue

async def main():
    q = Queue("emails")
    await send_welcome.enqueue(q.backend, "alice@example.com", delay=10)
anyio.run(main)

Now

import anyio
from asyncmq.queues import Queue

async def main():
    q = Queue("emails")
    await send_welcome.enqueue("alice@example.com", backend=q.backend, delay=10)
anyio.run(main)

0.2.3¶

Fixed¶

Add alternative check to get the concurrency field from the Redis backend.
Ensure pagination data is checked when listing workers.
Added check for workers.html pagination to ensure it is not empty.

0.2.2¶

Changed¶

Updated minimum version of Lilya.
Workers controller now reflects pagination and sizes.

Fixed¶

When starting a worker from the command-line, it was not automatically registering that same worker.
The workers.html was not reflecting the correct information about the active workers.

0.2.1¶

Fixed¶

StaticFiles were not enabling the html to True for the dashboard.

0.2.0¶

Release Notes¶

Added¶

Dashboard
- Brand new Dashboard where it shows how to integrate with your AsyncMQ. This dahboard is still in beta mode so any feedback is welcomed.
Top-Level CLI Commands Added four new top-level commands to the asyncmq CLI (powered by Click & Rich):
- list-queues — list all queues known to the backend
- list-workers — show all registered workers with queue, concurrency, and last heartbeat
- register-worker <worker_id> <queue> [--concurrency N] — register or bump a worker’s heartbeat and concurrency
- deregister-worker <worker_id> — remove a worker from the registry.
- AsyncMQGroup to cli ensuring the ASYNCMQ_SETTINGS_MODULE is always evaluated beforehand.
Worker Configuration
- Worker now accepts a heartbeat_interval: float parameter (default HEARTBEAT_TTL/3) to control how often it re-registers itself in the backend.

Changed¶

Redis Backend: Store Concurrency
- Now register_worker stores both heartbeat and concurrency as a JSON blob in each hash field.
- list_workers parses that JSON so the returned WorkerInfo.concurrency reflects the actual setting.
RedisBackend.cancel_job
- Now walks the waiting and delayed sorted sets via ZRANGE/ZREM instead of using list operations, then marks the job in a Redis SET.
- Eliminates “WRONGTYPE” errors when cancelling jobs.
InMemoryBackend.remove_job
- Expanded to purge a job from waiting, delayed, DLQ, and to clean up its in-memory state (job_states, job_results, job_progress).
Postgres Backend: register_worker Fix
- The queues column (a text[] type) now correctly receives a one-element Python list ([queue]) instead of a bare string.
- This resolves DataError: invalid input for query argument $2.
General Pause–Check Safety
- process_job now guards against backends that don’t implement is_queue_paused by checking with hasattr, avoiding AttributeError on simple in-memory or dummy backends.

Fixed¶

Fixed Redis hash-scan code to handle both bytes and str keys/values, preventing .decode() errors.
Ensured Postgres connection pool is wired into both list_queues and all worker-heartbeat methods.
Cleaned up duplicate fixtures in test modules to prevent event-loop and fixture-resolution errors.
Worker registration heartbeat tests were previously timing out, now pass reliably thanks to the configurable heartbeat interval.

0.1.0¶

Welcome to the first official release of AsyncMQ!

🚀 Highlights¶

🎉 A 100% asyncio & AnyIO foundation—no more thread hacks or callback nightmares.
🔌 A pluggable backend system: Redis, Postgres, MongoDB, In-Memory, or your own.
⏱️ Robust delayed and repeatable job scheduling, including cron expressions.
🔄 Built-in retries, exponential backoff, and time-to-live (TTL) semantics.
💀 Dead Letter Queues (DLQ) for failed-job inspection and replay.
⚡ Rate limiting and concurrency control to protect downstream systems.
🐚 Sandboxed execution with subprocess isolation and fallback options.
📊 Event Pub/Sub hooks for job:started, completed, failed, progress, cancelled, and expired.
🔀 Flow/DAG orchestration via FlowProducer, with atomic and fallback dependency wiring.
🛠️ A powerful CLI for managing queues, jobs, and workers—JSON output for scripting.

✨ New Features¶

Core APIs¶

@task decorator
Define sync or async functions as tasks.
Attach .enqueue() (alias .delay()) for one-line scheduling.
Support for retries, ttl, progress flag, depends_on, and repeat_every.
Queue class
add(), add_bulk(), add_repeatable() for single, batch, and periodic jobs.
pause(), resume(), clean(), queue_stats(), and in-depth inspection.
Configurable concurrency, rate_limit, rate_interval, and scan_interval.
process_job & handle_job
End-to-end lifecycle: dequeue, pause detection, TTL, delayed re-enqueue, execution (sandbox or direct), retry logic, DLQ, and events.
run_worker orchestrator
Combines ConcurrencyLimiter, RateLimiter, delayed_job_scanner, and repeatable_scheduler into a single async entrypoint.
repeatable_scheduler
Dynamic cron and interval scheduling with smart sleep intervals and high accuracy.
Utility compute_next_run() for dashboards and testing.
FlowProducer
Enqueue entire job graphs/DAGs with dependencies, using atomic backend calls or safe fallback.
Job abstraction
Rich state machine (WAITING, ACTIVE, COMPLETED, FAILED, DELAYED, EXPIRED).
Serialization via to_dict()/from_dict(), TTL checks, custom backoff strategies, dependencies, and repeat metadata.

Observability & Configuration¶

Settings dataclass
Centralized configuration (debug, logging_level, backend, DB URLs, TTL, concurrency, rate limits, sandbox options, scan intervals).
Environment variable ASYNCMQ_SETTINGS_MODULE for overrides.
LoggingConfig protocol
Built-in StandardLoggingConfig with timestamped console logs.
Pluggable for JSON, file rotation, or third-party handlers via custom LoggingConfig implementations.
EventEmitter hooks for real-time job events, ideal for Prometheus metrics or Slack alerts.

Developer Experience¶

CLI
asyncmq queue, asyncmq job, asyncmq worker, asyncmq info groups with intuitive commands and flags.
JSON and human-readable outputs, piping-friendly for shell scripts.
Documentation
Comprehensive Learn section with deep-dive guides on every component.
Features reference for quick lookups.
Performance Tuning and Security & Compliance guides (coming soon).
Testing Utilities
InMemoryBackend for fast, isolated unit tests.
Helpers for simulating delays, retries, failures, and cancellations.

🔄 Breaking Changes¶

This is the initial 0.1.0 release. There are no breaking changes yet! 🎉

🎯 Roadmap & Next Steps¶

Dashboard UI: real-time job monitoring and management interface.
Plugin Ecosystem: community-driven extensions for metrics, retries, and custom stores.

Thank you for choosing AsyncMQ! We can’t wait to see what you build.

Happy tasking! 🚀