How the scan works

This portfolio is self-updating. There is no hand-maintained list of projects — a script scans GitHub and regenerates the data every time the site builds.

The pipeline

1. prebuild hook. package.json runs bun run scripts/scan-repos.ts immediately before astro build (and before astro dev), so the data is always fresh.

2. List repositories. The script calls the GitHub REST API for every public repo, falling back to the authenticated gh CLI when direct fetch is unavailable (e.g. behind a strict proxy). Forks and the profile repo itself are filtered out by default.

3. Collect social images. For each repo it fetches the GitHub OpenGraph card from opengraph.githubassets.com and saves it under public/og/. If the host is unreachable at build time, it records the remote URL instead so the browser can load it directly.

4. Write the dataset. Normalized metadata (name, description, language, stars, topics, timestamps, image path) is written to src/data/repos.json.

5. Render the mosaic. Astro imports that JSON at build time and the RepoMosaic component renders it with WebCoreUI components, styled with UnoCSS.

Configuration

The scan is controlled by environment variables:

Variable	Default	Purpose
GITHUB_USER	archont561	Which GitHub login to scan.
GITHUB_TOKEN	(unset)	Raises API rate limits; provided by CI.
INCLUDE_FORKS	false	Set to true to include forked repositories.
INCLUDE_SELF	false	Set to true to include the profile repo.

Tech stack

• Runtime / package manager: Bun
• Framework: Astro + Starlight
• Components: WebCoreUI (Badge, AspectRatio, Ribbon)
• Styling: UnoCSS (WebCoreUI’s --w-* tokens are pre-compiled to plain CSS)
• Deploy: GitHub Actions → GitHub Pages