Owl Blog

Taking an Italian learning app from hardcoded admin credentials to proper Passport.js auth with email/password + Google OAuth, role-based access, and MongoDB session persistence. Phase 1 of a 4-phase upgrade adding user accounts, progress tracking, gamification, and adaptive AI-driven learning.

There are trips you plan because you want a vacation, and there are trips you plan because something in you is pulling you toward a place. This September, I’m flying into Milan for ten days, and it feels like the latter. Not tourism. Not a checklist. Something closer to a return—despite the fact that I’ve never lived there. For years I’ve been studying the language, tracing my family lines, and piecing together the stories of the people who left Italy generations ago. Somewhere along the way, Italy stopped being a place on a map and became a place that explains me. This trip is the first time I get to walk through that feeling instead of imagining it.

This post documents a reproducible DevOps workflow to run a local LLM stack on an Acer Predator laptop (64 GB RAM, RTX 5070 Ti — 12 GB VRAM, 2 TB storage). It covers host preparation, optional GPU passthrough, running Ollama for local inference, creating a lightweight Kubernetes cluster with k3d for the UI, deploying Open WebUI, and automating the stack with Ansible. The automation and manifests referenced live in https://github.com/binarobb/local-llm-stack.

Learn how I automated a major infrastructure upgrade using Ansible, GitHub Actions, and Apache Traffic Control. A technical deep-dive into DevOps automation.