Async programming | AI Pathshala

Definition

Asynchronous programming with async and await lets one thread handle many waiting tasks. An async function is a coroutine, and await pauses it so the event loop can run something else. asyncio is the standard library for this.

In simple words

One waiter serves ten tables. They take an order, hand it to the kitchen, and go to the next table instead of standing and waiting. await is that "I'll come back when it's ready".

Why it matters

Web servers such as FastAPI handle thousands of users this way. LLM apps benefit a lot, because calling a model is mostly waiting on the network.

How it works

Need	Tool
Heavy calculation	Processes
Many slow network calls	asyncio (or threads)
A few blocking libraries	Threads

In code

async_demo.py

import asyncio
import time

async def call_model(name, seconds):
    await asyncio.sleep(seconds)        # stands in for a network call
    return f"{name} done"

async def main():
    start = time.perf_counter()
    results = await asyncio.gather(
        call_model("A", 0.2),
        call_model("B", 0.2),
        call_model("C", 0.2),
    )
    print(results)
    print("Together under 0.5s:", time.perf_counter() - start < 0.5)

asyncio.run(main())

Output

['A done', 'B done', 'C done']
Together under 0.5s: True

Common mistakes

Forgetting await, so you get a coroutine object, and no result.
Blocking the event loop with slow synchronous code.
Using async for CPU-heavy work. It won't make it faster.

Check yourself

1. What does await do?

2. Calling an async function without await returns:

Summary

async and await handle many waiting tasks on one thread.
It's great for network calls, and not for heavy maths.
Never block the event loop.