Know the Instant Your Request Is Active — No Polling

The Polling Problem

The typical pattern developers end up writing without real-time feedback:

# What you end up building without real-time feedback
while True:
    status = check_queue_status(request_id)
    if status == "active":
        break
    await asyncio.sleep(0.2)  # poll every 200ms

await make_api_call()

Problems with this approach: wasted CPU on every check, up to 200ms of additional latency between a slot becoming available and your worker noticing, flooded logs, and catastrophic behavior under load when hundreds of requests are all polling simultaneously — creating a thundering herd of status checks.

Real-Time Activation with the SDK

The SDK handles this transparently. When you call acquire, your code blocks efficiently until the slot is granted — no polling loop on your side. Under the hood, RateQueue uses real-time feedback to wake up your coroutine the instant the slot opens.

import ratequeue.aio as rq

# Awaits efficiently — no busy polling on your side
async with rq.acquire("openai-gpt4", load=2000, api_key=RATEQUEUE_API_KEY):
    # Runs only once the slot is active
    response = await openai_client.chat.completions.create(
        model="gpt-4o",
        messages=messages
    )
# Slot released — next queued request is notified immediately

No while True: check_status() in your code. The async with suspends your coroutine and resumes it when RateQueue grants the activation.

Why This Matters Under Load

At scale, polling creates a thundering herd: 500 queued requests all polling every 200ms generates 2,500 status requests per second — none of which do useful work. Real-time feedback eliminates this. The coordination happens server-side — each worker is notified exactly once when it's time.

Polling Is Still Supported

If your environment doesn't support persistent connections, RateQueue's polling mode works fine. Switch between real-time and polling per resource in the dashboard. Both modes are handled transparently by the SDK — you don't change your code, just your plan.