How the terminal on this site works

2026-10-07 · astro, web, this-site

The homepage of this site looks like a Proxmox host booting into a shell. None of it runs on a server. Here is how it fits together.

Static first

Astro builds every page to plain HTML at build time. The pages you are reading now ship no JavaScript at all. Only the homepage loads a script: the terminal.

One config, one content folder

Everything the terminal knows comes from two places:

src/site.config.ts holds the facts about me, the links, the machine, the services and the man page.
src/content/ holds one Markdown file per project and per post.

At build time a small module turns both into one JSON document and puts it in the page. The terminal reads that JSON on load. Nothing is fetched later.

Projects are programs

Each project file becomes an executable in ~/projects, and that folder is on the shell’s PATH. So homelab, ./homelab and ~/projects/homelab all run the same thing: a short summary with a link to the full page.

Adding a project means adding one Markdown file:

---
title: Homelab
summary: Proxmox cluster that runs my self-hosted LLM stack.
date: 2026-01-02
stack: [Proxmox, vLLM]
---

The long version goes here.

The next build adds the page, the executable, tab completion and, because it has a date, a line in git log.

git log without git

The git log --oneline in the intro is not real git history. It sorts projects and posts by their date, newest first, and gives each one a stable short hash made from its id. That way the log never goes stale and never needs editing by hand.

Easy to leave

A terminal is fun, but it should never trap anyone:

Every page also has a real link, in a navigation menu that screen readers can find.
Without JavaScript you get a plain page with the same links.
Any key or tap skips the intro, and returning visitors skip the boot log.
With reduced motion turned on, nothing types, scrolls or rains.

Try help in the terminal to see what else it can do. Some commands are not on that list.