Hello, world

This post exists to prove the pipeline works, and to show what is available when writing a real one. Delete the whole src/content/blog/hello-world/ directory once it has served its purpose.

Writing a post

Create a directory under src/content/blog/. The directory name is the slug, so src/content/blog/why-i-built-todo/ is served at /blog/why-i-built-todo. Inside it, write DOCS-index.md — the DOCS- prefix is not decoration, it is what keeps the file legal under the repo-wide markdown-naming gate.

Frontmatter is strict: an unknown key is a build failure that names the file, rather than a post that quietly ships without a description.

Key	Required	Notes
title	yes	Used as the heading and the <title>
date	yes	YYYY-MM-DD
summary	yes	Shown on the index and in link previews
tags	no	[a, b]
cover	no	Post-relative image for link previews — use PNG or JPG
draft	no	true hides the post everywhere

Images

Images live next to the post and are referenced relatively:

Vite rewrites that path to a content-hashed URL. A path that does not resolve fails pnpm test immediately, so a broken image can never reach production.

One caveat worth knowing: link-preview scrapers do not render SVG. An SVG is fine inline, but a cover: should be a PNG or a JPG.

Code

Fenced blocks are highlighted:

export function slugOf(path: string): string {
  const directory = path.slice(0, path.lastIndexOf("/"));
  return directory.slice(directory.lastIndexOf("/") + 1);
}

pnpm dev      # write, with hot reload
pnpm test     # includes the broken-image check
pnpm build    # typecheck, bundle, then prerender every route

The rest of GFM

Tables, strikethrough, task lists and autolinks all work, courtesy of remark-gfm:

• Markdown posts with colocated images
• Real URLs, prerendered for link previews
• Something worth reading

Links go to the repo, and bare URLs like https://kuhy.duckdns.org autolink on their own.