Why run a homelab
The two Home Assistant integrations on the homepage — ha-fgc-trains and ha-tmb-barcelona — don’t run in the cloud. They run here, on a box in Joan’s apartment, alongside the agent framework that writes code, drafts content, and maintains this very site. This page is where that infrastructure, usually left out of a “what I built” writeup, actually gets explained: what’s running and how it fits together.
Proxmox layout
Proxmox VE is the hypervisor underneath all of this — open-source, Debian-based, and the thing that turns one physical box into a set of isolated virtual machines (VMs, full guest OSes) and containers (CTs, lighter-weight Linux namespaces that share the host kernel). Everything below is a VM or CT on a single Proxmox node unless noted otherwise.
Home Assistant is the home-automation platform the two GitHub integrations plug into, running as one of the guests on this Proxmox node.
Alongside Home Assistant, the same node hosts the agent framework (Hermes) that writes code, drafts content, and keeps this site running, plus the handful of small services — a reverse proxy and a certificate issuer — needed to get traffic to the site safely without exposing anything directly to the internet.
Networking
The site isn’t reachable by a directly exposed port. Traffic goes through a tunnel to the origin server, so nothing on the home network accepts unsolicited inbound connections from the public internet.
Access to the boxes themselves doesn’t rely on long-lived SSH keys sitting on disk — it’s certificate-based, issued on demand and short-lived, so there’s nothing permanent to lose or leak.