Agentic DevOps Extravaganza
demoAI-assisted Kubernetes triage you can run in one command — then watch the findings actually go away
A runnable demo of AI-assisted Kubernetes triage. The real k8sgpt binary (v0.4.36, 14 analyzers) finds 8 problems in a broken cluster; real kubectl writes fix them; a re-scan reports 0 because the state genuinely changed. Two paths: a 2-second mock API server, or a kind cluster with real Robusta.
The mock-path cluster is a simulation and labelled as such everywhere; the kind path runs the real Robusta Helm chart and kube-prometheus-stack. GIFs are recorded from live terminal sessions. Last verified 2026-10-07.
- Language
- Python, Go (k8sgpt)
- License
- MIT
Install
git clone https://github.com/adventurewave-labs/agentic-devops-extravaganza && cd agentic-devops-extravaganza && docker compose run --rm scanHow it works
Two run paths, both real: mock (a Python Kubernetes API server serving a deliberately broken payment-prod — real k8sgpt and real kubectl talk to it, including writes, ~2 seconds, needs only Python 3.10+) and kind (a real cluster with broken workloads and real Robusta installed by Helm, ~10 minutes). docker compose run --rm remediate issues the kubectl writes that take findings from 8 to 0.