Incident Root

Inspiration

As a software engineer, I often saw how production failures can cost teams 20–30 minutes or more per incident just to identify the root cause. Logs scattered across services, combined with error screenshots, made troubleshooting slow and stressful. I wanted to build a tool that could reduce this time dramatically and give engineers clarity in seconds instead of minutes.

What it does

IncidentRoot is an AI-powered incident analysis assistant that helps engineers respond faster during production outages. By analyzing application logs and error screenshots together, it automatically:

Identifies the root cause of the incident
Determines incident severity (P0–P3) with clear labels
Highlights affected services or components
Generates a step-by-step remediation plan
Produces a ready-to-use postmortem report

What normally takes 20–30 minutes of manual investigation is reduced to a few seconds.

How we built it

Frontend: Next.js 14 with a clean, dark terminal-inspired developer UI
Backend: Next.js API route using Google Gemini Pro (multimodal) to analyze logs and screenshots
AI Integration: Gemini receives text logs + inline screenshots, returning structured JSON including root cause, severity, remediation steps, and postmortem
File Handling: Multipart/form-data to handle both text logs and image uploads
UI/UX: Monospaced fonts, neon accents, two-column layout, and smooth animations for real-time feedback

Challenges we ran into

Ensuring Gemini output strictly matched our JSON schema while handling complex incident analysis
Preventing incomplete or truncated AI responses during long log investigations
Designing a responsive UI without sacrificing the terminal-inspired hacker aesthetic
Iterating rapidly under tight timelines, debugging AI edge cases late at night to keep the demo stable
Rebuilding prompt logic multiple times after failed outputs until the AI responses were reliable and actionable
Balancing product polish with engineering depth to deliver something that actually solves a real production pain

Accomplishments that we're proud of

Built a fully functional AI dashboard that processes both logs and screenshots simultaneously
Automated postmortem generation with professional formatting
Created a developer-friendly UI that clearly separates input, AI analysis, and actionable output