eventiq 0.3.0rc2

Publish/Subscribe asyncio framework for Python

Asyncio native pub/sub framework for Python

Installation

pip install eventiq

or

poetry add eventiq

Installing optional dependencies

pip install 'eventiq[broker]'

Available brokers

• nats
• rabbitmq
• kafka
• redis

Features

• Modern, asyncio based python 3.8+ syntax
• Fully type annotated
• Minimal external dependencies (anyio, pydantic, typer)
• Automatic message parsing based on type annotations using pydantic
• Code hot-reload
• Highly scalable: each service can process hundreds of tasks concurrently, all messages are load balanced between all instances by default
• Resilient - at least once delivery for all messages by default (except for Redis*)
• Customizable & pluggable message encoder/decoder (json as default)
• Multiple broker support
  • Memory (for testing)
  • Nats
  • Kafka
  • Rabbitmq
  • Redis
• Result Backend implementation for Nats & Redis
• Lifespan protocol support
• Lightweight (and completely optional) dependency injection system based on type annotations
• Easy and lightweight (~3k lines of code including types definitions and brokers implementations)
• Cloud Events standard as base message structure (no more python specific *args and **kwargs in messages)
• AsyncAPI documentation generation from code
• Twelve factor app approach - stdout logging, configuration through environment variables
• Easily extensible via Middlewares
• Multiple extensions and integrations including:
  • Prometheus - mertics exporter
  • OpenTelemetry - tracing and metrics
  • Message Pack - message pack encoder for messages
  • FastAPI - integrating eventiq Service with FastAPI applications (WIP)
  • Dataref - data reference resolver for messages (WIP)
  • Eventiq Workflows - orchestration engine built on top of eventiq (WIP)

Basic Usage

import asyncio
from eventiq import Service, Middleware, CloudEvent, GenericConsumer
from eventiq.backends.nats import JetStreamBroker

class SendMessageMiddleware(Middleware):
    async def after_broker_connect(self):
        print(f"After service start, running with {service.broker}")
        await asyncio.sleep(10)
        for i in range(100):
            message = CloudEvent(topic="test.topic", data={"counter": i})
            await service.publish(message)
        print("Published messages(s)")

broker = JetStreamBroker(url="nats://localhost:4222")

service = Service(
    name="example-service",
    broker=broker,
    middlewares=[SendMessageMiddleware()]
)


@service.subscribe(topic="test.topic")
async def example_run(message: CloudEvent):
    print(f"Received Message {message.id} with data: {message.data}")

@service.subscribe(topic="test.topic2")
class MyConsumer(GenericConsumer[CloudEvent]):
    async def process(self, message: CloudEvent):
        print(f"Received Message {message.id} with data: {message.data}")
        await self.publish(CloudEvent(topic="test.topic", data={"response": "ok"})

Run with

eventiq run app:service --log-level=info

Watching for changes

eventiq run app:service --log-level=info --reload=.

Testing

StubBroker class is provided as in memory replacement for running unit tests

import os


def get_broker(**kwargs):
    if os.getenv('ENV') == 'TEST':
        from eventiq.backends.stub import StubBroker
        return StubBroker()
    else:
        from eventiq.backends.rabbitmq import RabbitmqBroker
        return RabbitmqBroker(**kwargs)

broker = get_broker()

Furthermore, subscribers are just regular python coroutines, so it's possible to test them simply by invocation

# main.py
@service.subscribe(topic="test.topic")
async def my_subscriber(message: CloudEvent):
    return 42

# tests.py
from main import my_subscriber

async def test_my_subscriber():
    result = await my_subscriber(None)
    assert result == 42

CLI

Getting help:

eventiq --help

Installing shell autocompletion:

eventiq --install-completion [bash|zsh|fish|powershell|pwsh]

Basic commands

• run - run service
• docs - generate AsyncAPI docs
• send - send message to broker