retryback

A tiny, dependency-free retry decorator for Python with exponential backoff and jitter. Designed to be obvious, testable, and easy to drop into any project.

from retryback import retry, RetryError

@retry(attempts=5, base_delay=0.5, multiplier=2.0, max_delay=30, jitter=0.1)
def fetch():
    return some_flaky_call()

try:
    fetch()
except RetryError as exc:
    print("gave up:", exc.__cause__)

Install

pip install retryback

From source:

git clone https://github.com/nripankadas07/retryback.git
cd retryback
pip install -e .[dev]

API

retry(attempts=3, base_delay=0.1, max_delay=30.0, multiplier=2.0, jitter=0.1, exceptions=(Exception,), sleep=time.sleep, rng=None)

Returns a decorator. Parameters:

• attempts — total number of attempts (>= 1).
• base_delay — initial delay in seconds before the first retry.
• max_delay — cap on the delay between attempts. Use 0 to disable.
• multiplier — exponential growth factor (>= 1).
• jitter — random jitter as a fraction of the current delay, in [0, 1].
• exceptions — tuple of exception types that trigger a retry. Other exceptions propagate immediately.
• sleep — sleep function (injected for testing).
• rng — random.Random instance (injected for deterministic testing).

When all attempts fail, retry raises RetryError. The original exception is chained via __cause__.

RetryError

Subclass of RuntimeError raised after the final attempt fails.

Running tests

pip install -e .[dev]
pytest

License

MIT