A blog about my home server: how, step by step, I work through Docker, monitoring, backups and local AI. Practical notes from real deployments — what worked, what broke and what I learned. Concrete details you can reuse in your own setup.
A workshop on writing rules that wake you only when it matters: verifying metrics before writing thresholds, predict_linear instead of static limits, log-alert pitfalls — and why, after measuring it, I removed language models from the default notification path and left them as an opt-in.
Alertmanager webhooks, a per-user notification dispatcher, a queue for quiet hours, and the pitfalls of n8n's internal data model. Patterns from twelve workflows doing quiet background work — and an honest line where n8n stopped being the right tool.
Six cores, 32 GB of RAM, zero GPU — and language models that genuinely earn their keep at home: summarizing alerts, running a home agent, searching by meaning. On model selection, temperatures, and the pitfalls no tutorial mentions.
I added image uploads, Polish voice dictation, and Tesseract OCR to my self-hosted family app. Everything runs locally, no external APIs.