IncidentZero

Try it out

Live demo: https://incidentzero-web.onrender.com

IncidentZero is publicly deployed on Render for judging.

The hosted version uses an explicitly labeled simulation mode for Kubernetes infrastructure actions, which allows judges to evaluate the complete incident-resolution workflow without requiring access to a Kubernetes cluster or production credentials.

For the recommended demo:

1. Open the live application.
2. Start with the OOMKilled incident scenario.
3. Run the investigation.
4. Review the evidence, agent timeline, diagnosis, confidence score, remediation diff, validation results, and generated RCA.

The project was also tested against a controlled local Kind Kubernetes environment, including repeated end-to-end OOMKilled validation runs.

The hosted application intentionally does not receive unrestricted production-cluster write access. Production investigation is designed to remain read-only, and proposed remediation remains behind human review.

Source code: https://github.com/AryanSharma9917/incidentzero

Demo video: https://youtu.be/RaPNBI7qMDQ

Built With

• docker
• fastapi
• github
• kubernetes
• langchain
• langgraph
• nebius
• nemotron
• next.js
• nvidia
• python
• tailwind-css
• tavily
• typescript