fluxgate 0.7.0

A modern, composable circuit breaker library for Python

Fluxgate

A composable circuit breaker library for Python.

Installation

pip install fluxgate

Quick Start

from fluxgate import CircuitBreaker

cb = CircuitBreaker(name="my_api")

@cb
def call_api():
    return requests.get("https://api.example.com")

That's it. The circuit breaker will:

• Track the last 100 calls
• Open when failure rate exceeds 50% (after 100+ requests)
• Wait 60 seconds before testing recovery
• Gradually allow more calls as the service recovers

Async Support

For async applications, use AsyncCircuitBreaker:

from fluxgate import AsyncCircuitBreaker

cb = AsyncCircuitBreaker(name="my_api")

@cb
async def call_api():
    async with httpx.AsyncClient() as client:
        return await client.get("https://api.example.com")

Customization

Every component is customizable. You can apply different policies to different services based on their criticality:

import httpx
from fluxgate import AsyncCircuitBreaker
from fluxgate.windows import TimeWindow
from fluxgate.trackers import TypeOf, Custom
from fluxgate.trippers import MinRequests, FailureRate, FailureStreak
from fluxgate.retries import Backoff
from fluxgate.permits import RampUp

# Track only 5xx errors and network failures (not 4xx client errors)
def is_server_error(e: Exception) -> bool:
    if isinstance(e, httpx.HTTPStatusError):
        return e.response.status_code >= 500
    return isinstance(e, (httpx.ConnectError, httpx.TimeoutException))

# Conservative policy for critical payment service
payment_cb = AsyncCircuitBreaker(
    name="payment_service",
    window=TimeWindow(size=300),              # Track calls over 5 minutes
    tracker=Custom(is_server_error),          # Custom error filtering
    tripper=(
        FailureStreak(5) |                    # Trip on 5 consecutive failures OR
        (MinRequests(20) & FailureRate(0.4))  # 40% failure rate after 20 calls
    ),
    retry=Backoff(initial=30.0, max_duration=600.0),  # 30s, 60s, 120s... up to 10min
    permit=RampUp(0.1, 1.0, 60.0),            # Gradually allow 10% → 100% over 60s
)

# Aggressive policy for less critical inventory service
inventory_cb = AsyncCircuitBreaker(
    name="inventory_service",
    window=TimeWindow(size=60),               # Track calls over 1 minute
    tracker=TypeOf(httpx.HTTPError),          # Track all HTTP errors
    tripper=MinRequests(10) & FailureRate(0.6),  # 60% failure rate after 10 calls
    retry=Backoff(initial=10.0, max_duration=300.0),
)

@payment_cb
async def charge_payment(amount: float):
    async with httpx.AsyncClient() as client:
        response = await client.post("https://payment.example.com/charge", json={"amount": amount})
        response.raise_for_status()
        return response.json()

@inventory_cb
async def check_inventory(product_id: str):
    async with httpx.AsyncClient() as client:
        response = await client.get(f"https://inventory.example.com/products/{product_id}")
        response.raise_for_status()
        return response.json()

Components

Component	Purpose	Examples
Window	How to track call history	CountWindow(100), TimeWindow(60)
Tracker	Which exceptions to track	All(), TypeOf(HTTPError), Custom(func)
Tripper	When to open the circuit	MinRequests(n), FailureRate(0.5), FailureStreak(5), SlowRate(0.3, threshold=1.0), AvgLatency(1.0), Closed(), HalfOpened()
Retry	When to attempt recovery	Cooldown(60.0), Backoff(initial=10.0), Always(), Never()
Permit	How to allow calls during recovery	All(), Random(0.5), RampUp(0.1, 1.0, 60.0)

Monitoring

Attach listeners to get notified on state transitions:

from fluxgate import CircuitBreaker
from fluxgate.listeners.log import LogListener
from fluxgate.listeners.prometheus import PrometheusListener
from fluxgate.listeners.slack import SlackListener

cb = CircuitBreaker(
    name="my_api",
    listeners=[
        LogListener(),
        PrometheusListener(),
        SlackListener(channel="C123", token="xoxb-..."),
    ],
)

Documentation

Full documentation