Redis Caching With FastAPI (Optional)

Introduction

Redis stores short-lived data in memory—cached API responses, rate-limit counters, OTP codes. This chapter connects redis-py to FastAPI via lifespan, caches expensive reads, and points to the dedicated Redis track. Skip until you need shared cache across workers.

Prerequisites

Settings and Configuration
APIRouter and Project Structure—lifespan hooks
Redis running locally or Docker: docker run -d -p 6379:6379 redis:7
pip install redis

Settings

app/core/config.py:

python

class Settings(BaseSettings):
    ...
    redis_url: str = "redis://localhost:6379/0"

.env:

env

REDIS_URL=redis://localhost:6379/0

Lifespan Redis Client

app/main.py:

python

Import get_settings from app.core.config.

Or sync client in lifespan—fine for moderate traffic:

python

settings = get_settings()
client = redis.Redis.from_url(settings.redis_url, decode_responses=True)
app.state.redis = client

Code explanation:

app.state.redis shared across requests
Close connection on shutdown

Compare Spring Boot Redis and Flask caching.

Cache-Aside Pattern

python

Code explanation:

setex sets TTL 300 seconds
Invalidate on update: r.delete(f"item:{item_id}")

Rate Limiting Counter

python

Use slowapi for richer limits—concept in security chapter.

Dependency Wrapper

python

def get_redis(request: Request):
    return request.app.state.redis
 
@router.get("/ping-redis")
def ping_redis(r=Depends(get_redis)):
    return {"pong": r.ping()}

Async Redis (Optional)

bash

pip install redis[hiredis]

python

import redis.asyncio as aioredis
 
@asynccontextmanager
async def lifespan(app: FastAPI):
    client = aioredis.from_url(settings.redis_url, decode_responses=True)
    app.state.redis = client
    yield
    await client.aclose()

Use await r.get in async def routes.

FAQ

Cache stale after DB update?

Delete keys on write or use short TTL.

Redis down?

Decide fail-open (skip cache) vs fail-closed (503)—document behavior.

Multiple Uvicorn workers?

Redis shares cache—unlike in-process lru_cache.

Session in Redis?

Possible with session middleware—JWT stateless is simpler for APIs.

Learn Redis deeply?

Continue Redis track.

MongoDB as primary database?

Use PyMongo or Motor in lifespan—not JDBC. See PyMongo with FastAPI and Mongo + Redis project.

Skip chapter?

Yes until traffic or rate limits require shared memory store.