What is flawless?

Flawless is a system for watching over Kubernetes clusters and cloud infrastructure and helping fix problems when they happen, using an AI model as part of the process rather than as a chatbot that only gives advice. It connects alerts, supporting evidence, a map of how services depend on each other, human approval, controlled fixes, and a check that the fix actually worked, into one loop that can be reviewed afterward. The loop the project follows is discover, diagnose, preview, approve, execute, verify, and learn. When something breaks, Flawless gathers logs, metrics, events, and recent changes as evidence, proposes a plan, and waits for a person to approve it before making any change. After the change runs, it checks the original symptom again instead of just assuming the command succeeded, and it remembers which strategies failed so future attempts do not repeat them. Core parts include a chat style console for asking about the state of a cluster, a queue that scans systems on a schedule or on demand, a topology view showing how workloads and services connect to each other, and a skills library for reusable operational knowledge. It also supports uploading documentation such as PDFs, Word files, and runbooks so the system can reference them, and it connects to Prometheus, Loki, Tempo, and Grafana for further observability data. To run it, someone needs Git plus either Docker with Compose, or Python 3.11 and Node.js 20 as a fallback. There are one line installer scripts for macOS, Linux, and Windows, or it can be cloned manually and started with a quickstart script that also includes status, log, and update commands. By default it starts without a connected AI model, a local model such as one served through Ollama, or an OAuth protected model gateway, has to be configured afterward before chat features work. The license is PolyForm Noncommercial, meaning commercial use is restricted.