Oops.stack

oops.stack

Inspiration

Every engineer has had that moment: production breaks, logs explode, Slack lights up, and everyone is scrolling instead of fixing. Debugging is noisy, stressful, and slow — especially under on-call pressure.

We wanted to build something that turns that chaos into clarity. Instead of “what just happened?”, we wanted “here’s the fix.”

That’s how oops.stack was born.

What it does

oops.stack turns raw logs and stack traces into PR-ready fixes.

It pulls errors from Sentry, replays the failure inside a clean Daytona sandbox to ensure reproducibility, analyzes what actually went wrong, and then generates CodeRabbit-style PR review comments and patch diffs.

For on-call speed, it can also read out the diagnosis using ElevenLabs, so engineers don’t even have to read logs when things are on fire.

In short: logs in, pull request out.

How we built it

FastAPI backend to ingest logs and stack traces
Sentry as the incident and error source
Daytona to execute log analysis inside isolated, reproducible sandboxes
CodeRabbit-style output to generate PR review comments and unified diffs
ElevenLabs to convert root-cause summaries into voice for on-call scenarios

The system is designed to gracefully fall back if any integration isn’t available, keeping the workflow reliable during demos and real usage.

Challenges we ran into

Integrating Daytona reliably while keeping the system resilient to environmental issues
Designing CodeRabbit-style output that feels like a real PR review, not just AI text
Balancing a Gen-Z, playful tone with production-grade technical credibility
Making the demo fast, clear, and judge-friendly under hackathon time constraints