A deploy goes red. The log is 4,000 lines. Someone scrolls, sighs, and re-runs it.

What it does

  • A deploy fails → the AI pulls the log, reads it against your repo, and writes a triage issue
  • The issue includes the relevant log excerpt plus a likely cause, not the whole dump
  • The same commit fails again → it comments on the existing issue instead of opening a new one
  • Everything it read and concluded is saved as a CI artifact you can audit

What it does not do

It never re-runs, rolls back, or promotes a deployment. It explains; you act.

Honest note: beta — built and smoke-tested, still to be proven across real failure modes.

Setup 

clone → ./setup.sh → pick Vercel → answer a few questions

~5 minutes. No server, no database, free.

Get it

The recipe is open source: github.com/akshaypatel-ai/ai-automations

This is one of 30+ automations in the series — the full write-up lives on my site.