throttle.py

import threading
import time

TOKENS_PER_MINUTE = 5500
_tokens_per_second = TOKENS_PER_MINUTE / 60.0
_capacity = TOKENS_PER_MINUTE

_lock = threading.Lock()
_tokens = _capacity
_last_timestamp = time.monotonic()

def consume(n: int) -> None:
    """Consume *n* tokens, waiting if necessary.

    This function implements a simple thread-safe token bucket to keep
    requests under the configured tokens-per-minute rate.
    """
    global _tokens, _last_timestamp
    if n <= 0:
        return
    while True:
        with _lock:
            now = time.monotonic()
            elapsed = now - _last_timestamp
            _tokens = min(_capacity, _tokens + elapsed * _tokens_per_second)
            _last_timestamp = now
            if n <= _tokens:
                _tokens -= n
                return
            needed = n - _tokens
            wait_time = needed / _tokens_per_second
            _tokens = 0
        time.sleep(wait_time)