python-patterns

Choose FastAPI for API-first microservices needing async support, Django for full-stack applications requiring a built-in admin and ORM, and Flask for simple scripts or learning projects. Consider whether you need an admin interface, async support, and your team's familiarity with each framework.

Use async def when performing I/O-bound operations like async database queries with asyncpg or HTTP calls with httpx. Use def for blocking operations, sync database drivers, or CPU-bound work, since FastAPI runs sync endpoints in a threadpool automatically.

Use httpx for async HTTP clients, asyncpg for PostgreSQL, redis-py async or aioredis for Redis, aiofiles for file I/O, and SQLAlchemy 2.0 async or Tortoise for ORM needs. Never use sync libraries inside async code paths.

Django 5.0 and later supports async views, async middleware, and limited async ORM operations with ASGI deployment. Use async in Django for external API calls, WebSockets via Channels, and high-concurrency views.

Use FastAPI BackgroundTasks for quick fire-and-forget operations in the same process without persistence. Choose Celery or ARQ for long-running tasks, retry logic, distributed workers, persistent queues, and complex workflows.

Calling blocking sync libraries inside async functions blocks the event loop and destroys concurrency benefits. Keep I/O-bound work async with async-native libraries, and offload CPU-bound work to multiprocessing instead of forcing async.