What is incident-ai?

Incident AI, also called CascadeFlow AI in the README, is a hackathon project built for HackwithChennai 2026 that aims to speed up how engineering teams respond to system outages. The idea is that when a major outage happens, engineers currently have to manually dig through monitoring tools like Datadog and Prometheus and coordinate over Slack, which takes time. This project tries to automate that triage process using AI. The README describes a dashboard showing live system health across monitoring services, an automated root cause analysis feature that explains what broke and suggests fixes, and a visual map showing how a failure in one part of a system could cascade into others. It also describes a memory feature built on a vector database that would let the system recall how past incidents were resolved, and a feature that routes different tasks to different sized AI models depending on how much reasoning they need. Under the hood, the project uses Next.js 15 for the frontend, Supabase with PostgreSQL and pgvector for data storage, and the Groq API running Llama 3.3 70B for the AI reasoning. Styling is done with Tailwind CSS and framer-motion for animations. To run it locally, you clone the repository, install dependencies with npm or pnpm, add Supabase and Groq API keys to an environment file, and start the development server. It can also be deployed to Vercel with the same environment variables. The README makes large claims about cost savings and accuracy numbers, but these read as illustrative rather than measured results, since this is presented as a hackathon build rather than a project with published benchmarks or a track record in production. There is no license file mentioned beyond a note that it targets SOC2 Type II compliance standards, which is a goal stated in the text rather than something verified.