AI / LLM Ops
Hermes Agent — Personal AI Infrastructure on a Raspberry Pi
I run Hermes Agent on a Raspberry Pi as my always-on personal AI assistant. It connects to my messaging platforms (WhatsApp, Discord) and operates as a single, persistent agent that knows my context across sessions — no app switching, no repeating myself.
What it handles for me daily
- Language learning
- Adaptive Spanish tutoring with SM-2 spaced repetition, sentence translation drills, and grammar mastery tracking. A cron job fires 3×/day with 5 custom sentences.
- Task management
- Natural-language Todoist task creation, rescheduling, and completion via chat. "Task P1: Kraxe und Badehose einpacken Samstag 9:00" just works.
- Health tracking
- Fitbit data via Google Health API. Steps, HR, HRV, sleep stages, calories — one message away. OAuth runs headlessly.
- Content work
- Editing my vacation rental website's React frontend straight from a chat message, with git commit and push.
- Scheduled automation
- Cron jobs for morning briefings, daily health reports, and recurring reminders, all routed to the right platform.
Stack
- Raspberry Pi 5
- Python
- systemd services
- OAuth2 device flow
- git-based deployment
Why it beats a chat app
Persistent memory. Tools that actually execute. Cross-session context. No "As an AI I can't…" — it just runs the script, refreshes the token, and gets it done.