ratelimit-lite

A lightweight, client-side rate limiter for outbound API calls — no database required. Works entirely in memory by default, with optional Redis support for coordinating limits across multiple processes or pods.

Install

pip install ratelimit-lite

For Redis support:

pip install ratelimit-lite[redis]

Quickstart

from ratelimit_lite.limiter import Limiter
from ratelimit_lite.decorator import rate_limited

limiter = Limiter(algorithm="sliding_window", backend="memory", limit=50, window_size=60)

@rate_limited(limiter, key="weather_api")
def call_weather_api():
    ...

Or use it as a context manager:

with limiter.acquire("weather_api"):
    call_weather_api()

By default, a blocked call raises RateLimitExceeded. Pass wait=True to block and retry automatically instead:

@rate_limited(limiter, key="weather_api", wait=True)
def call_weather_api():
    ...

Algorithms

• "fixed_window" — simple counter that resets every window_size seconds. Fast and predictable, but allows brief bursts near window boundaries.
• "sliding_window" — weighted counter that blends the previous and current window, avoiding boundary bursts. Recommended default for most use cases.

Backends

• "memory" — in-process, no dependencies. State resets on restart and does not coordinate across multiple processes.
• "redis" — coordinates limits across multiple processes or pods, using Lua scripts for atomicity. Requires a running Redis server.

from ratelimit_lite.limiter import Limiter

# uses localhost:6379 by default
limiter = Limiter(algorithm="sliding_window", backend="redis", limit=50, window_size=60)

To use a custom Redis connection, pass a configured client directly:

import redis
from ratelimit_lite.stores.redis_store import RedisStore

custom_client = redis.Redis(host="my-redis-host", port=6379)
store = RedisStore(redis_client=custom_client)