link-preview-js

Link Preview JS

Before creating an issue

It’s more than likely there is nothing wrong with the library:

• It’s very simple; fetch HTML, parse HTML, and search for OpenGraph HTML tags.
• Unless HTML or the OpenGraph standard change, the library will not break
• If the target website you are trying to preview redirects you to a login page the preview will fail, because it will parse the login page
• If the target website does not have OpenGraph tags the preview will most likely fail, there are some fallbacks but in general, it will not work
• You cannot preview (fetch) another web page from YOUR web page. This is an intentional security feature of browsers called CORS

DO NOT FETCH CONTENT DIRECTLY FROM A USERS DEVICE. ONLY RUN THIS IN YOUR SERVER AND SANDBOX IT IF YOU CAN

Browsers block this via cors, but you might be clever like a fox and run this in React Native. This is a bad idea, you are exposing the device user to potentially malicious links

If you use this library and find it useful please consider sponsoring me, open source takes a lot of time and effort.

Link Preview

Allows you to extract information from an HTTP URL/link (or parse an HTML string) and retrieve meta information such as title, description, images, videos, etc. via OpenGraph tags.

Discord

Join the Discord

https://discord.gg/W9XmqCQCKP

GOTCHAs

• You cannot request a different domain from your web app (Browsers block cross-origin-requests). If you don’t know how same-origin-policy works, here is a good intro, therefore this library works on Node.js and certain mobile run-times (Cordova or React-Native).
• This library acts as if the user would visit the page, sites might re-direct you to sign-up pages, consent screens, etc. You can try to change the user-agent header (try with google-bot or with Twitterbot), but you need to work around these issues yourself.

Entrypoints

This package ships two variants:

• link-preview-js / link-preview-js/node - full-featured, for Node.js servers. Includes the resolveDNSHost SSRF/DNS-rebinding protection (see SSRF Concerns), which pins HTTPS connections at the TCP layer via undici.
• link-preview-js/mobile - for React Native, Cordova, or other non-Node/browser runtimes. Has no undici or node:net dependency, so it won’t break bundlers that can’t resolve Node-only modules. It has no resolveDNSHost option either: mobile apps don’t sit in front of a private network of internal services the way a server does, so there’s no loopback/localhost target worth protecting against on-device.

// Node.js server (default import, same as "link-preview-js/node")
import { getLinkPreview } from "link-preview-js";

// React Native / browser / other non-Node runtimes
import { getLinkPreview } from "link-preview-js/mobile";

API

getLinkPreview: you have to pass a string, doesn’t matter if it is just a URL or a piece of text that contains a URL, the library will take care of parsing it and returning the info o the first valid HTTP(S) URL info it finds.

getPreviewFromContent: useful for passing a pre-fetched Response object from an existing async/etc. call. Refer to the example below for required object values.

import { getLinkPreview, getPreviewFromContent } from "link-preview-js";

// pass the link directly
getLinkPreview("https://www.youtube.com/watch?v=MejbOFk7H6c").then((data) => console.debug(data));

////////////////////////// OR //////////////////////////

// pass a chunk of text
getLinkPreview(
  "This is a text supposed to be parsed and the first link displayed https://www.youtube.com/watch?v=MejbOFk7H6c",
).then((data) => console.debug(data));

////////////////////////// OR //////////////////////////

// pass a pre-fetched response object
// The passed response object should include, at minimum:
// {
//   data: '<!DOCTYPE...><html>...',     // response content
//   headers: {
//     ...
//     // should include content-type
//     content-type: "text/html; charset=ISO-8859-1",
//     ...
//   },
//   url: 'https://domain.com/'          // resolved url
// }
yourAjaxCall(url, (response) => {
  getPreviewFromContent(response).then((data) => console.debug(data));
});

Options

Additionally, you can pass an options object which should add more functionality to the parsing of the link

Property Name	Result
imagesPropertyType (optional) (ex: ‘og’)	Fetches images only with the specified property, meta[property='${imagesPropertyType}:image']
headers (optional) (ex: { ‘user-agent’: ‘googlebot’, ‘Accept-Language’: ‘en-US’ })	Add request headers to fetch call
timeout (optional) (ex: 1000)	Timeout for the request to fail
followRedirects (optional) (default ‘error’)	For security reasons, the library does not automatically follow redirects (‘error’ value), a malicious agent can exploit redirects to steal data, posible values: (‘error’, ‘follow’, ‘manual’)
handleRedirects (optional) (with followRedirects ‘manual’)	When followRedirects is set to ‘manual’ you need to pass a function that validates if the redirectinon is secure, below you can find an example
resolveDNSHost (optional)	Function that resolves the final address of the detected/parsed URL to prevent SSRF attacks
onResponse (optional)	Function that handles the response object to allow for managing special cases

getLinkPreview("https://www.youtube.com/watch?v=MejbOFk7H6c", {
  imagesPropertyType: "og", // fetches only open-graph images
  headers: {
    "user-agent": "googlebot", // fetches with googlebot crawler user agent
    "Accept-Language": "fr-CA", // fetches site for French language
    // ...other optional HTTP request headers
  },
  timeout: 1000,
}).then((data) => console.debug(data));

SSRF Concerns

Doing requests on behalf of your users or using user-provided URLs is dangerous. One of such attack is trying to fetch a domain that redirects to localhost so the users get the contents of your server (doesn’t affect mobile runtimes). To mitigate this attack you can use the resolveDNSHost option, available on the default/link-preview-js/node import only - see Entrypoints.

The default options offer no SSRF protection. Without resolveDNSHost the library only rejects URLs that literally contain a private or loopback address, or a known internal hostname. It does not check where a hostname resolves, so any public domain pointing at an internal address will still be fetched. If you fetch user-provided URLs on a server, always pass resolveDNSHost.

// example how to use node's dns resolver
const dns = require("node:dns");
getLinkPreview("http://maliciousLocalHostRedirection.com", {
  resolveDNSHost: async (url: string) => {
    return new Promise((resolve, reject) => {
      const hostname = new URL(url).hostname;
      dns.lookup(hostname, (err, address, family) => {
        if (err) {
          reject(err);
          return;
        }

        resolve(address); // if address resolves to localhost or '127.0.0.1' library will throw an error
      });
    });
  },
}).catch((e) => {
  // will throw a detected redirection to localhost
});

This might add some latency to your request but prevents loopback attacks.

Note: for https:// URLs the request keeps the original hostname (rewriting it to the resolved IP would break TLS: certificates and SNI are validated against the hostname, not an IP). Instead, the connection itself is pinned to the validated address (via a Node undici dispatcher), so a DNS resolver that returns a different address on the real request than it did during the resolveDNSHost check (DNS rebinding) can’t route the request elsewhere. This pinning is only available under Node.js; other runtimes fall back to the resolveDNSHost check alone.

Redirections

Same to SSRF, following redirections is dangerous, the library errors by default when the response tries to redirect the user. There are however some simple redirections that are valid (e.g. HTTP to HTTPS) and you might want to allow them, you can do it via:

await getLinkPreview(`http://google.com/`, {
  followRedirects: `manual`,
  handleRedirects: (baseURL: string, forwardedURL: string) => {
    const urlObj = new URL(baseURL);
    const forwardedURLObj = new URL(forwardedURL);
    if (
      forwardedURLObj.hostname === urlObj.hostname ||
      forwardedURLObj.hostname === "www." + urlObj.hostname ||
      "www." + forwardedURLObj.hostname === urlObj.hostname
    ) {
      return true;
    } else {
      return false;
    }
  },
});

handleRedirects is called for every hop of a redirect chain (baseURL is the URL that redirected, forwardedURL where it points to), up to 20 redirects. Whether redirects are followed with 'follow' or 'manual', every hop goes through the same SSRF checks as the original URL (including resolveDNSHost) before it’s requested.

onResponse

In some cases the website might be missing OpenGraph tags and you might want to provide your own custom logic to try to parse data. For example, if the library is unable to detect a description, you might want to use the text value of the first paragraph instead. This callback gives you access to the Cheerio (the library internally used to parse the HTML) instance, as well as the URL object so you could handle cases on a site-by-site basis, if you need to. This callback must return the modified response object

await getLinkPreview(`https://example.com/`, {
  onResponse: (response, doc, URL) => {
    if (URL.hostname == "example.com") {
      response.siteName = "Example Website";
    }

    if (!response.description) {
      response.description = doc("p").first().text();
    }

    return response;
  },
});

Response

Returns a Promise that resolves with an object describing the provided link. The info object returned varies depending on the content type (MIME type) returned in the HTTP response (see below for variations of response). Rejects with an error if the response can not be parsed or if there was no URL in the text provided.

Text/HTML URL

{
  url: "https://www.youtube.com/watch?v=MejbOFk7H6c",
  title: "OK Go - Needing/Getting - Official Video - YouTube",
  siteName: "YouTube",
  description: "Buy the video on iTunes: https://itunes.apple.com/us/album/needing-getting-bundle-ep/id508124847 See more about the guitars at: http://www.gretschguitars.com...",
  images: ["https://i.ytimg.com/vi/MejbOFk7H6c/maxresdefault.jpg"],
  mediaType: "video.other",
  contentType: "text/html",
  charset: "utf-8"
  videos: [],
  favicons:["https://www.youtube.com/yts/img/favicon_32-vflOogEID.png","https://www.youtube.com/yts/img/favicon_48-vflVjB_Qk.png","https://www.youtube.com/yts/img/favicon_96-vflW9Ec0w.png","https://www.youtube.com/yts/img/favicon_144-vfliLAfaB.png","https://s.ytimg.com/yts/img/favicon-vfl8qSV2F.ico"]
}

Image URL

{
  url: "https://media.npr.org/assets/img/2018/04/27/gettyimages-656523922nunes-4bb9a194ab2986834622983bb2f8fe57728a9e5f-s1100-c15.jpg",
  mediaType: "image",
  contentType: "image/jpeg",
  favicons: [ "https://media.npr.org/favicon.ico" ]
}

Audio URL

{
  url: "https://ondemand.npr.org/anon.npr-mp3/npr/atc/2007/12/20071231_atc_13.mp3",
  mediaType: "audio",
  contentType: "audio/mpeg",
  favicons: [ "https://ondemand.npr.org/favicon.ico" ]
}

Video URL

{
  url: "https://www.w3schools.com/html/mov_bbb.mp4",
  mediaType: "video",
  contentType: "video/mp4",
  favicons: [ "https://www.w3schools.com/favicon.ico" ]
}

Application URL

{
  url: "https://assets.curtmfg.com/masterlibrary/56282/installsheet/CME_56282_INS.pdf",
  mediaType: "application",
  contentType: "application/pdf",
  favicons: [ "https://assets.curtmfg.com/favicon.ico" ]
}

License

MIT license