linkpeek vs unfurl.js
Compare linkpeek with unfurl.js, which is archived upstream: speed, install size, oEmbed, edge runtimes, and migrating unfurl() to preview().
unfurl.js is a Node.js metadata scraper that returns nested Open Graph, Twitter Card and oEmbed data. Its GitHub repository is archived and its last npm release, 6.4.0, dates from February 2024. linkpeek does the same preview-card job with a flat typed result, runs on any runtime with fetch, and is still maintained.
Measured side by side
| Metric | linkpeek 3.0.0 | unfurl.js 6.4.0 |
|---|---|---|
| 500 kB page, median | 2.76 ms | 11.0 ms |
| Small pages, mean of medians | 1.25 ms | 0.80 ms |
| Clean install size | 1.3 MB | 3.2 MB |
| Packages installed | 7 | 16 |
unfurl.js was faster than linkpeek on every small page in this run, by a fraction of a millisecond. On the large page linkpeek was faster because it stops reading after </head>. Both were measured through their own fetch against the same local server. See the benchmarks for the method and every package.
Choose unfurl.js when
- You want oEmbed payloads fetched for you. unfurl.js requests the discovered oEmbed endpoint by default and returns the parsed result. linkpeek returns the endpoint as
oEmbedUrland does not fetch it. - You want the full nested structure: every Open Graph image and video, Twitter Card players and app links, article times and tags.
- You preview small pages on Node.js and the last fraction of a millisecond matters.
- Your code already works and you accept a dependency that will not receive upstream fixes.
Choose linkpeek when
- You want a maintained dependency. unfurl.js is archived; the still-open PR #117 documents that URLs pass to
node-fetchwithout destination validation. linkpeek blocks private and special-use IP literals and checks every redirect before following it. On Node.js, passresolveHostto check each hop's DNS answers;fetchresolves the name again, so keep egress rules too, as the security model explains. - You run on Cloudflare Workers, Vercel Edge, Deno or Bun. unfurl.js is built on
node-fetchand documents Node.js only; linkpeek uses platformfetch, Web Streams andTextDecoder. - Your users paste links to large pages, where reading only the head pays off.
- You want one value per field: linkpeek falls back from Open Graph to Twitter Card, Dublin Core, the document title and JSON-LD on its own, and types every field as a value or
null.
Migrate unfurl() to preview()
npm uninstall unfurl.js
npm install linkpeek// unfurl.js
import { unfurl } from "unfurl.js";
const data = await unfurl(url);
const card = {
title: data.open_graph?.title ?? data.title,
description: data.open_graph?.description ?? data.description,
image: data.open_graph?.images?.[0]?.url,
favicon: data.favicon,
canonical: data.canonical_url,
};
// linkpeek
import { preview } from "linkpeek";
const result = await preview(url);
const card = {
title: result.title,
description: result.description,
image: result.image,
favicon: result.favicon,
canonical: result.canonicalUrl,
};The options map closely: timeout stays timeout, follow becomes maxRedirects, size becomes maxBytes and headers stays headers. unfurl.js sends facebookexternalhit as its user agent; if some sites only served tags to that, set userAgent explicitly.
HTTP error pages resolve with their metadata and a statusCode, so check result.statusCode < 400 before caching. Blocked targets, timeouts and network failures throw a LinkpeekError with a code; the API reference lists them.
If you used the oEmbed data
Fetch oEmbedUrl yourself, and only for providers you trust. The URL comes from the page, so treat it like any other untrusted destination:
const OEMBED_HOSTS = new Set(["www.youtube.com", "vimeo.com"]);
let oEmbed: unknown = null;
if (result.oEmbedUrl && OEMBED_HOSTS.has(new URL(result.oEmbedUrl).hostname)) {
const response = await fetch(result.oEmbedUrl, {
redirect: "error",
signal: AbortSignal.timeout(5000),
});
if (response.ok) oEmbed = await response.json();
}New to linkpeek? The quick start covers install, presets and where it runs.
vs open-graph-scraper
Compare linkpeek with open-graph-scraper: speed on large pages, install size, edge runtime support, SSRF defaults, and how to migrate ogs() to preview().
vs url-metadata
Compare linkpeek with url-metadata: large-page speed, install size, byte limits, edge runtimes, SSRF defaults, and migrating urlMetadata().