Stream parsing tweets - json

Question

The bottom line is to listen to a specific user and transfer his tweets with minimal delay to the telegram bot.
To implement this task, I use the Twiipi library in which, as I understand it, there are 2 most important types of authentication for me:
On behalf of the user - 900 requests in 15 minutes (i.e. 1 request in 1 second)
On behalf of the application - 300 requests in 15 minutes (i.e. 1 request in 3 seconds)
I authenticate my script on behalf of the user using the OAuth1UserHandler function (listed below).
But despite this, the delay in speed appears after 7.5 minutes of work, given that the script for interacting with twitter resources runs once every 1.5 seconds. That is one way or another Authentication of my script takes place on behalf of the application. But despite this, I just made a second bot that starts after 7 minutes of the previous one, thereby updating the time for the previous one to work. My main problem is that the performance of parsing tweet data drops (acceptable speed is a maximum of 1.5 seconds, delays sometimes last 13 seconds).
Please tell me what I'm doing wrong or how can I solve my problem better?
code of one of 2 bots
import tweepy
import datetime
import time
from notifiers import get_notifier
from re import sub
TOKEN = 'telegram token here'
USER_ID = telegram user id here
ADMIN_ID = teelgram my id here for check work bots
auth = tweepy.OAuth1UserHandler(
consumer_key="consumer key",
consumer_secret="consumer secret",
access_token="access token",
access_token_secret="access token secret",
)
api = tweepy.API(auth)
# auth.set_access_token(access_token, access_token_secret)
print("############### Tokens connected ###############")
user = 'whose username we will listen to'
username_object = api.get_user(screen_name=user)
def listening_to_the_user():
print(' We start listening to the user...')
print(' When a user posts a tweet, you will hear an audio notification...')
seconds_left = 60*10
while seconds_left >= 0:
for i in api.user_timeline(user_id=username_object.id, screen_name=user, count=1):
tweet_post = i.created_at
tweet_text = sub(r"https?://t.co[^,\s]+,?", "", i.text)
tweet_time_information = [tweet_post.day, tweet_post.month, tweet_post.year, tweet_post.hour, tweet_post.minute]
now = datetime.datetime.now()
current_time = [now.day, now.month, now.year, now.hour, now.minute]
if tweet_time_information == current_time:
telegram = get_notifier('telegram')
notification_about_tweet = f'️{user}⬇️'
notification_about_tweet_time = f'{tweet_post.day}.{tweet_post.month}.{tweet_post.year}, {tweet_post.hour}:{tweet_post.minute}.{tweet_post.second}'
notification_about_current_time = f'{now.day}.{now.month}.{now.year}, {now.hour}:{now.minute}.{now.second}'
telegram.notify(token=TOKEN, chat_id=USER_ID, message=notification_about_tweet)
telegram.notify(token=TOKEN, chat_id=USER_ID, message=tweet_text)
try:
entities = i.extended_entities
itr = entities['media']
for img_dict in range(len(itr)):
telegram.notify(token=TOKEN, chat_id=ADMIN_ID, message=(entities['media'][img_dict]['media_url_https']))
except:
entities = 0
telegram.notify(token=TOKEN, chat_id=ADMIN_ID, message=notification_about_tweet)
telegram.notify(token=TOKEN, chat_id=ADMIN_ID, message=tweet_text)
telegram.notify(token=TOKEN, chat_id=ADMIN_ID, message=notification_about_tweet_time)
telegram.notify(token=TOKEN, chat_id=ADMIN_ID, message=notification_about_current_time)
try:
entities = i.extended_entities
itr = entities['media']
for img_dict in range(len(itr)):
telegram.notify(token=TOKEN, chat_id=USER_ID,
message=(entities['media'][img_dict]['media_url_https']))
except:
entities = 0
seconds_left -= 60
time.sleep(60)
seconds_left -= 1.5
time.sleep(1.5)
listening_to_the_user()
Initially, I tried to use bearer_token for authentication, but this literally did not affect the operation of my program in any way and I simply reduced it to those tokens that are now in it.
I rummaged through the documentation in search of an answer to my question, but in search I only thought of the script calling the second bot after 7 minutes of the previous one, and so they worked in turn