Parse and Render
Problem
You have Markdown (a README fragment, an archive publication, an LLM reply) and need it on screen as real semantic HTML — without wiring parser + sanitizer + escaping yourself.
Solution
ts
import { parseProse, proseOutline, proseToText, renderProseHtml } from '@vielzeug/scribe';
const source = `# Release 2.0
Highlights:
- **Streaming** replies
- Faster cold starts
- ~~Legacy API~~ gone
See the [changelog](/changelog).
`;
const doc = parseProse(source);
// 1. HTML string (SSR / static sites)
console.log(renderProseHtml(doc, { headingIds: true }));
// 2. Plain text (search index, notifications)
console.log(proseToText(doc));
// 3. Table of contents (ids match the HTML above exactly)
console.log(proseOutline(doc));
// [{ level: 1, text: 'Release 2.0', id: 'release-20' }]
// 4. In the browser: build real nodes instead of strings
document.getElementById('article')?.replaceChildren(
// import { renderProseDom } from '@vielzeug/scribe'
// document.replaceChildren(renderProseDom(doc))
);In the browser the same doc renders with renderProseDom(doc): a DocumentFragment built exclusively with createTextNode/createElement, ready for replaceChildren.
Pitfalls
- Setext headings (
Title\n=====) are not in the profile: the underline renders literally. Use#headings. - A blank line is required before a table or a new block; single newlines inside a paragraph are soft breaks.
- The AST is frozen by design — derive new docs by reparsing, not by mutation.
renderProseDomneedsdocument; in Node userenderProseHtml.