Cloudflare Markdown Proxy

A Hono-based Cloudflare Worker that converts web pages to clean Markdown or styled, print-friendly HTML.

Features

• /text — Converts HTML pages to Markdown via Cloudflare's Workers AI toMarkdown API
• /print — Cleans original HTML with HTMLRewriter, injects a print stylesheet, preserves semantic structure
• Content filtering: scope Markdown conversion to a CSS selector and strip unwanted elements by ID or class
• Domain whitelisting to restrict which sites can be proxied
• CORS-enabled responses
• Zero external runtime dependencies — Hono and all libraries are bundled at build time

Usage

GET /text?page=https://news.ucsc.edu/2026/03/some-article/

Returns the page content as Markdown (text/markdown).

GET /print?page=https://news.ucsc.edu/2026/03/some-article/

Returns cleaned HTML (text/html) with screen and print stylesheets. Original HTML structure is preserved — tables, figures, semantic elements all survive intact. Suitable for reading in a browser or printing to PDF.

Configuration

Environment variables are set in wrangler.toml under [vars]:

Variable	Description	Default
WHITELISTED_DOMAIN	Only this hostname can be proxied	news.ucsc.edu
CSS_SELECTOR	Scope AI conversion to a content area, /text only (e.g. .entry-content)	.entry-content
REMOVE_IDS	Comma-separated element IDs to strip	""
REMOVE_CLASSES	Comma-separated class names to strip	(see wrangler.toml)

The AI binding is configured under [ai] in wrangler.toml.

Development

npm install        # Install dependencies
npm run dev        # Start local dev server
npm run test       # Run tests (46 tests across unit and integration suites)
npm run deploy     # Deploy to Cloudflare

Project Structure

src/
  index.js                  # Hono app — mounts routes, CORS
  routes/
    home.js                 # GET / — front page
    print.js                # GET /print — cleaned HTML with print stylesheet
    text.js                 # GET /text — Markdown via AI.toMarkdown
  middleware/
    validate-page.js        # Validates ?page param, fetches upstream HTML
  lib/
    tidy-html.js            # HTMLRewriter-based HTML cleaner
    preprocess-html.js      # HTML prep for Markdown conversion
    utils.js                # parseList(), stripFrontMatter()
  pretty.css                # Screen and print stylesheet
tests/
  unit/                     # Pure function tests
  integration/              # Route tests (workerd runtime)
  vitest.workspace.js       # Vitest workspace config
wrangler.toml               # Worker config, env vars, AI binding

License

MIT