Built-in Rate Limiter - Questions #4145

Q1 — No, it is not enabled by default. Installing the base package does not throttle anything, and installing python-telegram-bot[rate-limiter] on its own does not either: that extra only installs aiolimiter, which AIORateLimiter needs. Throttling starts only when you pass an instance explicitly, e.g.:

from telegram.ext import AIORateLimiter, ApplicationBuilder

app = (
    ApplicationBuilder()
    .token(TOKEN)
    .rate_limiter(AIORateLimiter())
    .build()
)

Q2 — Your throughput guess is roughly right, the mechanism is not batches. With the defaults, AIORateLimiter combines overall_max_rate=30 per overall_time_period=1 second with a group limit of group_max_rate=20 per group_time_period=60 seconds, applied per request via aiolimiter. That is a continuous limiter, not "build 4 batches of 30 and sleep 1s between batches": roughly the first 30 requests can go out as a burst, and later requests wait inside the limiter as capacity frees up at about 30/second.

For your 120 different private users, the 20/minute limit for a single group is not what determines the total time — the 30/second overall limit is — so expect the sends to be spread over roughly 3–4 seconds in total (plus API/network latency), assuming you issue them concurrently; a sequential await loop adds its own latency on top.

One caveat: max_retries defaults to 0, so if Telegram still answers with RetryAfter (429), the limiter does not retry that call for you by default. Set max_retries to a value above 0 on AIORateLimiter(...), or handle RetryAfter and wait for its retry_after value, if a failed broadcast message must eventually be delivered.