Deploy-time notify

Your deploy finishes. Nobody notices v1.8.2 rolled back until a customer pings in #support two hours later.

The problem

Most CI pipelines log the rollback to stdout and move on. You can glue a curl-to-Slack step on, then copy-paste the payload for Telegram, then do it again for every service. Or you install snitchbot, run init() once, and call notify() like any other Python function.

The recipe

# deploy.py — run at the end of your CI pipeline
import snitchbot

snitchbot.init("ci")

snitchbot.notify(
    "Deploy rolled back to v1.8.2",
    severity="warning",
    extras={
        "previous": "v1.8.3",
        "duration_s": 47,
        "rolled_by": "ci-bot",
    },
)

# `init()` registered an `atexit` hook that drains the socket on normal exit.
# On a short-lived CI runner, give it a beat before the process ends.
import time
time.sleep(1.0)

What you see

🟠 notify · ci · a1b2c3
Deploy rolled back to v1.8.2
Details
  time     17:02:11 UTC
  pid      42
  caller   deploy.py:5 in <module>()
Extras
  previous     v1.8.3
  duration_s   47
  rolled_by    ci-bot

Notes

• init() spawns a sidecar subprocess and registers an atexit hook that drains the socket on normal exit. On a short-lived CI runner add a time.sleep(1.0) before the script ends — otherwise the runner reaps the process before the event leaves the buffer.
• Use severity="error" for failed deploys, "critical" for prod rollbacks that paged someone. Routing follows severity.
• See notify() for the full signature.