aretry 0.1.0

Lightweight retry decorator with async support and exponential backoff — zero dependencies

aretry

Lightweight retry decorator for sync and async Python functions. Zero dependencies.

invl/retry — the most popular Python retry library — gets 8M+ weekly downloads but hasn't shipped a release since 2018 and has no async support. This is the modern replacement.

Install

pip install aretry

Usage

from aretry import retry

# Works on sync functions
@retry(times=3, on=requests.Timeout, backoff=2.0)
def fetch():
    return requests.get(url, timeout=5)

# Works on async functions — same decorator, no changes needed
@retry(times=3, on=aiohttp.ClientError, backoff=2.0, jitter=True)
async def fetch_async():
    async with session.get(url) as r:
        return await r.json()

# @retry with no parens uses all defaults (3 attempts, 1s delay, 2x backoff)
@retry
def simple():
    ...

Parameters

Parameter	Default	Description
times	3	Total attempts (1 = try once, no retry)
on	Exception	Exception type or tuple of types to catch
delay	1.0	Initial wait in seconds before first retry
backoff	2.0	Multiply delay by this after each attempt
jitter	False	Add ±20% random noise to avoid thundering herd
before_retry	None	Callback (exc, attempt, delay) called before each retry

Backoff example

With times=4, delay=1.0, backoff=2.0:

Attempt	Wait before next
1 (fail)	1s
2 (fail)	2s
3 (fail)	4s
4 (fail)	raises RetryExhausted

Error handling

from aretry import RetryExhausted

try:
    fetch()
except RetryExhausted as e:
    print(f"Gave up after {e.attempts} attempts. Last error: {e.last_exc}")

RetryExhausted is a subclass of Exception. The original exception is also accessible via standard __cause__ chaining.

License

MIT