Post

Your SRE buddy that speeds up investigations and learns over time? 🕵️‍♂️ I know: "Oh great, yet another AI agent!" 😛 But this one earns its spot through its simplicity and its core design: keeping the human in the loop. Blog post & architecture: blog.ogenki.io/post/series/... 🧵👇

To learn from incidents, it uses the OKF standard (inspired by @karpathy's llmwiki). The mechanics: • Single binary • Context compounds over time in a local Git repo • Total control: absolutely nothing enters memory without a human-reviewed PR.

The project is still very young and needs real-world usage to capitalize on gathered knowledge. Give it a spin (you can easily limit LLM calls to start), feedback and PRs are more than welcome! 🤝 github.com/Smana/runlore #SRE #OpenSource #Kubernetes #Observability

Start with testing in smaller, stable environments first before full-scale deployment.