Telegram alerts for Python scripts that fail at night

Scripts that run on a timer fail quietly: a cron job that syncs orders, a worker that sends emails, a scraper that ran fine for months. This guide uses the justlog3 package to send their logs to JustLog3 Cloud and to get a Telegram message when a run starts misbehaving.

Install and set the token

pip install justlog3

Copy the API token from your dashboard into an environment variable, JUSTLOG3_TOKEN, so it never ends up in the code or in Git.

Log with levels

import os
import time

from justlog3 import BasicLogger, set_api_token

set_api_token(os.environ['JUSTLOG3_TOKEN'])
logger = BasicLogger('sync.log', console_level='INFO')


def sync_orders():
    for attempt in range(1, 6):
        try:
            count = fetch_and_store_orders()  # your code
            logger.success(f'synced {count} orders')
            return
        except ConnectionError as error:
            logger.warning(f'attempt {attempt}/5 failed: {error}')
            time.sleep(2 ** attempt)
    logger.critical('orders sync gave up after 5 attempts')


logger.info('sync started')
sync_orders()

Every line goes to sync.log and to the cloud; console_level only filters what the terminal shows. log() and the level methods never wait for the network: a background thread sends lines every 5 seconds or every 75 lines, and the rest on exit.

If the cloud can't be reached, the lines stay in memory (up to 10,000 per logger) and are resent with the same sequence number, so a retry never stores them twice. The local file is written either way.

One run, one session

Each process gets its own random session id, so every run of the script is a separate session in Logs. You see at a glance which run failed and read only its lines.

Alerts that mean something

Link Telegram from the dashboard: Telegram notifications → Create secure link, then press Start in the bot. From then on:

• When one session grows past another 1,000 lines, you get a message. A normal sync logs a handful of lines; a loop that hammers a dead API logs thousands. The alert comes while it's happening, not the next morning.
• At 80% and 100% of the day's log lines you get a heads-up, so a noisy script can't eat the quota unnoticed.

Ask why it failed

Open AI Assistant, switch on Logs and ask: "Why did the orders sync fail at 03:00?". The assistant reads your recent lines, newest first, with repeats collapsed, and points at the attempt that broke. With GitHub on as well, it can open the file the traceback mentions.

Shutdown matters

Lines are flushed on a normal exit and on SIGTERM (docker stop, systemd). kill -9 or a power cut loses at most the lines still in the buffer. If you end the process with os._exit(), call justlog3.shutdown() first.

Free plan

5,000 log lines a day kept for 2 days and 12 assistant replies a day, with no card. More in Plans and limits and the Python docs.