Command line

Convert UTF-8 HTML from stdin to Markdown or structured JSON.

All four implementations provide the same commands and JSON field names.

Install

Choose one implementation:

With the Python package installed, python -m htomd also works. You can run the Node CLI without a global installation using npx @jamiedavenport/htomd.

Convert HTML

cat page.html | htomd convert > page.md
cat page.html | htomd extract > page.json
curl -fsSL https://example.com/article | htomd convert --url https://example.com/article

Both commands read UTF-8 HTML from stdin until EOF, stripping an initial BOM. They accept no file arguments; use pipes or redirection. --url URL supplies source context and resolves relative references. In the last example, curl fetches the page; htomd does not.

convert writes Markdown. extract writes indented JSON with this shape:

{
  "markdown": "",
  "metadata": {
    "title": null,
    "author": null,
    "description": null,
    "language": null,
    "published_time": null,
    "url": null,
    "canonical_url": null
  },
  "diagnostics": {
    "strategy": "none",
    "notes": []
  }
}

Absent metadata stays null. Strategy values are semantic, scored, fallback, and none; notes describe extraction and recovery. The JSON keys are the same across implementations, including published_time and canonical_url.

Help and errors

htomd --help
htomd help extract
htomd extract --help
htomd --version

Help and version commands do not read stdin. htomd with no command shows help. htomd version is an alias for htomd --version.

Errors go to stderr. Broken output pipes exit quietly. Malformed HTML alone is not an error; see shared behavior and limitations.