UniteIO

UniteIO runs several asyncio applications in one Python process, each on its own thread with its own event loop. The applications share the process, its state, and one thread-pool executor for synchronous work.

It works with and without the GIL. On a free-threaded build, Python code on different application threads can run in parallel; with the GIL, the loops are still independent, but their Python code takes turns.

Status: early alpha. Requires Python 3.14.5 or newer. The API may still change between minor versions.

Installation

pip install uniteio

Example

import asyncio
from uniteio import App, Runtime


class Clock(App):
    """Ticks on its own thread and loop."""

    def __init__(self, interval):
        self.interval = interval   # ordinary constructor, any signature
        self.ticks = 0

    async def __call__(self):
        while True:
            await asyncio.sleep(self.interval)
            self.ticks += 1


async def double(value):
    return value * 2


with Runtime() as rt:
    clock = rt.start(Clock(0.1))       # starts the thread and loop, returns the same object
    assert rt["Clock"] is clock and rt[Clock] is clock

    # From any thread: coroutines run on the app's loop, callables on the shared executor.
    assert clock.submit(double, 21).result(timeout=2) == 42
    assert clock.submit(sum, [1, 2, 3]).result(timeout=2) == 6
# leaving the block stops every application, then shuts the executor down

• Runtime.start runs __call__ on the application's loop and returns once the loop is running. Several instances of one class need different names: rt.start(Clock(1), name="slow").
• App.submit is how other threads hand work to an application. Inside the loop, use asyncio directly (await, create_task, TaskGroup).
• Application loops reject non-thread-safe calls (call_soon, create_task, …) coming from other threads, and blocking on loop work from the loop's own thread raises instead of deadlocking.

The documentation covers errors, shutdown, and the rules for crossing threads.

Development

uv pip install -e ".[dev]"
python -m pytest
python -m sphinx -W --keep-going -b html docs docs/_build/html

Build and validate the release artifacts:

uv build
uvx twine check dist/*

License

UniteIO is distributed under the MIT License.