# aiohttp — Async HTTP Client and Server for Python

> aiohttp provides both an asynchronous HTTP client and a full web server framework built on asyncio. It handles thousands of concurrent connections efficiently without threads.

## Install

Save in your project root:

# aiohttp — Async HTTP Client and Server for Python

## Quick Use
```bash
pip install aiohttp
python -c "
import aiohttp, asyncio
async def main():
    async with aiohttp.ClientSession() as s:
        async with s.get('https://httpbin.org/get') as r:
            print(await r.json())
asyncio.run(main())
"
```

## Introduction
aiohttp is a mature asynchronous HTTP framework for Python built on top of asyncio. It provides both a client for making concurrent HTTP requests and a server for building web applications and APIs that handle high concurrency without thread pools, making it a foundation for many async Python projects.

## What aiohttp Does
- Provides an async HTTP client with connection pooling and session management
- Includes a full web server framework with routing, middleware, and signals
- Supports WebSocket connections on both client and server sides
- Handles streaming request and response bodies for large payloads
- Offers multipart form data parsing and file upload handling

## Architecture Overview
aiohttp runs on Python's asyncio event loop. The client uses a `ClientSession` that maintains a connection pool with keep-alive and cookie persistence. The server maps URL patterns to async handler coroutines via a router, processes requests through a middleware chain, and supports both traditional handler functions and class-based views. Internally it uses a hand-written HTTP parser for performance.

## Self-Hosting & Configuration
- Install with `pip install aiohttp` and optionally `aiohttp[speedups]` for C extensions
- Create a server app with `web.Application()` and add routes via `app.router.add_get()`
- Run the server with `web.run_app(app, host='0.0.0.0', port=8080)`
- Deploy behind Nginx or with Gunicorn using the `aiohttp.worker.GunicornWebWorker` class
- Configure timeouts, connection limits, and SSL via `TCPConnector` and `ClientTimeout`

## Key Features
- Connection pooling with configurable limits per host and total
- Built-in WebSocket support with ping/pong and close frame handling
- Middleware system for auth, CORS, logging, and error handling
- Streaming responses for serving large files without loading into memory
- Signal hooks for application startup, shutdown, and cleanup lifecycle events

## Comparison with Similar Tools
- **httpx** — async client only, no server; aiohttp provides both client and server in one package
- **FastAPI** — higher-level with auto-generated docs; aiohttp is lower-level with more control over the HTTP layer
- **requests** — synchronous and blocking; aiohttp handles thousands of concurrent requests via asyncio
- **Tornado** — similar async approach but older; aiohttp is built on the standard asyncio event loop
- **Starlette** — ASGI-based and lighter; aiohttp has its own protocol implementation with more built-in features

## FAQ
**Q: Should I use aiohttp client or httpx?**
A: aiohttp client is mature and performant. httpx offers a requests-like API and HTTP/2 support. Choose based on your API preferences.

**Q: Can I use aiohttp with Django or Flask?**
A: Not directly. aiohttp has its own server framework. Use the client portion alongside any framework for async HTTP calls.

**Q: How do I handle graceful shutdown?**
A: Register cleanup callbacks with `app.on_shutdown.append(cleanup_func)`. The server runs them when it receives a shutdown signal.

**Q: Does aiohttp support HTTP/2?**
A: aiohttp currently supports HTTP/1.1. For HTTP/2, consider httpx or h2.

## Sources
- https://github.com/aio-libs/aiohttp
- https://docs.aiohttp.org/en/stable/

---
Source: https://tokrepo.com/en/workflows/a8014482-40a1-11f1-9bc6-00163e2b0d79
Author: AI Open Source