A user says, “it’s not working.”But where exactly is the problem? Which page? What were they doing? Frontend or backend? How serious is it, and how difficult will it be to fix?I built GlitchPack to automate that process.
GlitchPack is an open-source “Report a problem” system for web applications.The user describes the problem, while console errors, failed API requests and recent clicks are automatically attached to the report. Requests returning 400+ can also be detected without the user submitting a report.The report is then analyzed by TypeSafe AI Jev model.
Jev;
separately evaluates frontend/backend probability, severity, fix difficulty and whether the report contains AI-directed instructions.The result becomes a labelled GitHub issue automatically. Backend problems can be sent to a webhook or Slack. Based on the fix difficulty score, a lighter or more capable coding model can be selected for the task.For frontend issues, the AI-generated solution is added to the issue. When a teammate adds the autofix label, a PR is created.But nothing is merged automatically.
AI proposes, a human decides.
Currently supports Next.js App Router.
🔗 npm: https://www.npmjs.com/package/glitchback
🔗 GitHub: https://github.com/Canzz0/glitchback
Top comments (0)