6.3 KiB

Raw Blame History

fastapi-420

Production rate limiting for FastAPI. Uses HTTP code 420 "Enhance Your Calm" because 429 is boring.

Installation

pip install fastapi-420

For Redis support:

pip install fastapi-420[redis]

Quick Start

Three ways to add rate limiting. Pick what fits your app.

Middleware (global)

Limits all routes automatically.

from fastapi import FastAPI
from fastapi_420 import RateLimiter, RateLimitMiddleware

app = FastAPI()
limiter = RateLimiter()

app.add_middleware(
    RateLimitMiddleware,
    limiter=limiter,
    default_limit="100/minute",
)

@app.get("/")
async def root():
    return {"message": "hello"}

Decorator (per route)

Fine grained control on specific endpoints.

from fastapi import FastAPI, Request
from fastapi_420 import RateLimiter

app = FastAPI()
limiter = RateLimiter()

@app.get("/search")
@limiter.limit("30/minute")
async def search(request: Request, q: str):
    return {"results": []}

@app.post("/upload")
@limiter.limit("5/minute", "20/hour")
async def upload(request: Request):
    return {"status": "ok"}

Dependency (FastAPI style)

Works with FastAPI's dependency injection.

from fastapi import FastAPI, Depends
from fastapi_420 import RateLimiter, RateLimitDep, set_global_limiter

app = FastAPI()
limiter = RateLimiter()
set_global_limiter(limiter)

@app.get("/api/data", dependencies=[Depends(RateLimitDep("50/minute"))])
async def get_data():
    return {"data": []}

Common Patterns

Different limits for different endpoints

Auth endpoints get strict limits. Public endpoints stay relaxed.

from fastapi_420 import ScopedRateLimiter

auth_limiter = ScopedRateLimiter(
    prefix="/auth",
    default_rules=["5/minute"],
    endpoint_rules={
        "POST:/auth/login": ["3/minute", "10/hour"],
        "POST:/auth/register": ["2/minute"],
    },
)

@app.post("/auth/login", dependencies=[Depends(auth_limiter)])
async def login():
    ...

Using Redis

Memory storage works fine for single instances. Redis for distributed apps.

from fastapi_420 import RateLimiter, RateLimiterSettings, StorageSettings

settings = RateLimiterSettings(
    storage=StorageSettings(
        REDIS_URL="redis://localhost:6379/0",
    ),
)

limiter = RateLimiter(settings=settings)

If Redis goes down, the limiter falls back to memory automatically.

Trusting proxy headers

Behind nginx or a load balancer? Trust the forwarded headers.

from fastapi_420 import FingerprintSettings, RateLimiterSettings

settings = RateLimiterSettings(
    fingerprint=FingerprintSettings(
        TRUST_X_FORWARDED_FOR=True,
    ),
)

Custom identification

Rate limit by user ID instead of IP.

def get_user_id(request):
    return request.state.user_id or request.client.host

@app.get("/api/resource")
@limiter.limit("100/minute", key_func=get_user_id)
async def resource(request: Request):
    ...

Configuration Reference

All settings with their defaults.

RateLimiterSettings

from fastapi_420 import RateLimiterSettings
from fastapi_420.types import Algorithm

RateLimiterSettings(
    # Algorithm
    ALGORITHM=Algorithm.SLIDING_WINDOW,  # SLIDING_WINDOW | TOKEN_BUCKET | FIXED_WINDOW

    # Defaults applied when no rules specified
    DEFAULT_LIMIT="100/minute",
    DEFAULT_LIMITS=["100/minute"],       # list form, multiple rules

    # Storage key configuration
    KEY_PREFIX="rl",
    KEY_VERSION="v1",

    # Response behavior
    INCLUDE_HEADERS=True,                # add RateLimit-* headers
    HTTP_420_MESSAGE="Enhance Your Calm",
    HTTP_420_DETAIL={"error": "rate_limit_exceeded", "message": "Enhance Your Calm"},

    # Failure handling
    FAIL_OPEN=True,                      # allow requests if storage fails
    LOG_VIOLATIONS=True,                 # log when limits exceeded

    # Nested settings (see below)
    storage=StorageSettings(...),
    fingerprint=FingerprintSettings(...),
)

StorageSettings

from fastapi_420 import StorageSettings

StorageSettings(
    # Redis (optional, falls back to memory if not set or unavailable)
    REDIS_URL=None,                      # "redis://localhost:6379/0"
    REDIS_KEY_PREFIX="rl",
    REDIS_SOCKET_TIMEOUT=5.0,
    REDIS_SOCKET_CONNECT_TIMEOUT=5.0,
    REDIS_MAX_CONNECTIONS=50,
    REDIS_RETRY_ON_TIMEOUT=True,
    REDIS_HEALTH_CHECK_INTERVAL=30,

    # Memory storage
    MEMORY_MAX_KEYS=100_000,             # max keys before LRU eviction
    MEMORY_CLEANUP_INTERVAL=60,          # seconds between expired key cleanup
)

FingerprintSettings

Controls how clients are identified. Higher levels are stricter but may cause issues with legitimate users behind proxies.

from fastapi_420 import FingerprintSettings
from fastapi_420.types import FingerprintLevel

FingerprintSettings(
    LEVEL=FingerprintLevel.NORMAL,       # RELAXED | NORMAL | STRICT

    # What to trust
    TRUST_X_FORWARDED_FOR=False,         # trust X-Forwarded-For header
    TRUSTED_PROXIES=[],                  # IPs that can set forwarded headers

    # IPv6 handling
    IPV6_PREFIX_LENGTH=64,               # normalize IPv6 to /64 prefix
)

Fingerprint Levels:

Level	What it uses
RELAXED	IP only
NORMAL	IP + User-Agent
STRICT	IP + User-Agent + Accept headers + Auth token hash

Algorithms

Algorithm	Behavior	Best for
SLIDING_WINDOW	Smooth, accurate limits	Most cases (default)
TOKEN_BUCKET	Allows short bursts	APIs with bursty traffic
FIXED_WINDOW	Simple, less accurate at window edges	High performance needs

Rate Limit Format

Rules follow the pattern {requests}/{period}:

100/minute
50/hour
1000/day
10/second

Multiple rules stack. The most restrictive one applies:

@limiter.limit("10/second", "100/minute", "1000/hour")
async def endpoint(request: Request):
    ...

Running the Example

cd examples
docker compose up -d
pip install fastapi uvicorn
python app.py

Then hit http://localhost:8000/docs to see the API.

Why 420?

Twitter used HTTP 420 "Enhance Your Calm" for rate limiting before switching to 429. It is more fun.

The exception is called EnhanceYourCalm and the response tells clients to chill out.

License

MIT

6.3 KiB Raw Blame History