Course outline

Every chapter has: the goal, the time it takes, the steps, tested template files, a "check that it works" section, "what can go wrong", and the security notes.

1. What you will build, and the 10 safety rules - free - read it
2. Plan it: hardware, memory, power, and what one small box can really do - free - read it
3. Linux + Docker, done safely - users, updates, firewall, automatic security updates, no open ports
4. The local brain - running an open language model on a mini PC, choosing a model, a queue for one-job-at-a-time, streaming answers
5. Hearing - a wake word that is always on and nearly free, then speech-to-text (server and gaming-PC variants)
6. Speaking - local text-to-speech that sounds natural (voices, licences, loudness)
7. The tool gateway - the assistant may only use allowlisted tools; the policy file; the audit log
8. Approvals on your phone - exact, single-use, expiring OK buttons with a chat bot, done safely
9. Secrets - encrypted, one secret per service, never in git, logs or chats
10. Backups that restore - automatic backups, the restore drill, safe updates, the reboot test
11. Remote access without open ports - a private VPN link, and what never to expose
12. The smart-home link (optional) - letting the assistant see and control your home safely

Every month after that: a new chapter (for example: comparing models on your own questions, a second voice device, notes and memory, measuring power and cost).