azaynul10/kube-autofix

Kube-AutoFix is a Python tool that automatically detects and repairs failing Kubernetes deployments without human intervention. Kubernetes is a system for running and managing containerized applications at scale, when a deployment breaks, for example, a container image fails to download or a pod crashes on startup, someone normally has to investigate the logs, diagnose the root cause, and fix the configuration manually. Kube-AutoFix automates that entire process. When a broken configuration file (a YAML manifest) is applied, the agent enters a loop: it deploys the configuration, monitors whether the pods (the running units of the application) come up successfully, and if they fail, it collects diagnostic information including pod descriptions, namespace events, and container logs. It sends that debug bundle with the original configuration to GPT-4o, which acts as an automated site reliability engineer. GPT-4o returns a structured response with a root cause analysis and a corrected YAML file. The agent applies the fix and tries again, up to five iterations before stopping. The tool includes safety guardrails. It is locked to a single namespace so it cannot accidentally modify other parts of the infrastructure. The AI is instructed to make the smallest possible change and is prohibited from adding new resources unless strictly necessary. A dry-run mode lets you inspect what the AI would fix before any changes are applied. The tech stack is Python, the official Kubernetes Python client, and the OpenAI SDK for GPT-4o.