Unfurl card in link previews

src/deps.ts

@@ -29,3 +29,4 @@ export { default as linkifyStr } from 'npm:linkify-string@^4.1.0';
 import 'npm:linkify-plugin-hashtag@^4.1.0';
 // @deno-types="npm:@types/mime@3.0.0"
 export { default as mime } from 'npm:mime@^3.0.0';
+export { unfurl } from 'npm:unfurl.js@^6.3.1';

src/note.ts

@@ -6,16 +6,6 @@ linkify.registerCustomProtocol('wss');
 
 const url = (path: string) => new URL(path, LOCAL_DOMAIN).toString();
 
-/** Get pubkey from decoded bech32 entity, or undefined if not applicable. */
-function getDecodedPubkey(decoded: nip19.DecodeResult): string | undefined {
-  switch (decoded.type) {
-    case 'npub':
-      return decoded.data;
-    case 'nprofile':
-      return decoded.data.pubkey;
-  }
-}
-
 const linkifyOpts: linkify.Opts = {
   render: {
     hashtag: ({ content }) => {
@@ -46,16 +36,8 @@ type Link = ReturnType<typeof linkify.find>[0];
 interface ParsedNoteContent {
   html: string;
   links: Link[];
-}
-
-/** Ensures the URL can be parsed. Why linkifyjs doesn't already guarantee this, idk... */
-function isValidLink(link: Link): boolean {
-  try {
-    new URL(link.href);
-    return true;
-  } catch (_e) {
-    return false;
-  }
+  /** First non-media URL - eligible for a preview card. */
+  firstUrl: string | undefined;
 }
 
 /** Convert Nostr content to Mastodon API HTML. Also return parsed data. */
@@ -63,10 +45,12 @@ function parseNoteContent(content: string): ParsedNoteContent {
   // Parsing twice is ineffecient, but I don't know how to do only once.
   const html = linkifyStr(content, linkifyOpts);
   const links = linkify.find(content).filter(isValidLink);
+  const firstUrl = links.find(isNonMediaLink)?.href;
 
   return {
     html,
     links,
+    firstUrl,
   };
 }
 
@@ -77,9 +61,7 @@ interface MediaLink {
 
 function getMediaLinks(links: Link[]): MediaLink[] {
   return links.reduce<MediaLink[]>((acc, link) => {
-    const { pathname } = new URL(link.href);
-    const mimeType = mime.getType(pathname);
-
+    const mimeType = getUrlMimeType(link.href);
     if (!mimeType) return acc;
 
     const [baseType, _subType] = mimeType.split('/');
@@ -95,4 +77,39 @@ function getMediaLinks(links: Link[]): MediaLink[] {
   }, []);
 }
 
+function isNonMediaLink({ href }: Link): boolean {
+  return /^https?:\/\//.test(href) && !getUrlMimeType(href);
+}
+
+/** Ensures the URL can be parsed. Why linkifyjs doesn't already guarantee this, idk... */
+function isValidLink(link: Link): boolean {
+  try {
+    new URL(link.href);
+    return true;
+  } catch (_e) {
+    console.error(`Invalid link: ${link.href}`);
+    return false;
+  }
+}
+
+/** `npm:mime` treats `.com` as a file extension, so parse the full URL to get its path first. */
+function getUrlMimeType(url: string): string | undefined {
+  try {
+    const { pathname } = new URL(url);
+    return mime.getType(pathname) || undefined;
+  } catch (_e) {
+    return undefined;
+  }
+}
+
+/** Get pubkey from decoded bech32 entity, or undefined if not applicable. */
+function getDecodedPubkey(decoded: nip19.DecodeResult): string | undefined {
+  switch (decoded.type) {
+    case 'npub':
+      return decoded.data;
+    case 'nprofile':
+      return decoded.data.pubkey;
+  }
+}
+
 export { getMediaLinks, type MediaLink, parseNoteContent };

src/transmute.ts

@@ -1,4 +1,4 @@
-import { findReplyTag, lodash, nip19, z } from '@/deps.ts';
+import { findReplyTag, lodash, nip19, unfurl, z } from '@/deps.ts';
 import { type Event } from '@/event.ts';
 import { type MetaContent, parseMetaContent } from '@/schema.ts';
 
@@ -101,12 +101,13 @@ async function toStatus(event: Event<1>) {
     ),
   ];
 
-  const { html, links } = parseNoteContent(event.content);
+  const { html, links, firstUrl } = parseNoteContent(event.content);
   const mediaLinks = getMediaLinks(links);
 
   return {
     id: event.id,
     account,
+    card: firstUrl ? await unfurlCard(firstUrl) : null,
     content: html,
     created_at: new Date(event.created_at * 1000).toISOString(),
     in_reply_to_id: replyTag ? replyTag[1] : null,
@@ -128,7 +129,6 @@ async function toStatus(event: Event<1>) {
     mentions: await Promise.all(mentionedPubkeys.map(toMention)),
     tags: [],
     emojis: [],
-    card: null,
     poll: null,
     uri: `${LOCAL_DOMAIN}/posts/${event.id}`,
     url: `${LOCAL_DOMAIN}/posts/${event.id}`,
@@ -153,4 +153,47 @@ function renderAttachment({ url, mimeType }: MediaLink) {
   };
 }
 
+interface PreviewCard {
+  url: string;
+  title: string;
+  description: string;
+  type: 'link' | 'photo' | 'video' | 'rich';
+  author_name: string;
+  author_url: string;
+  provider_name: string;
+  provider_url: string;
+  html: string;
+  width: number;
+  height: number;
+  image: string | null;
+  embed_url: string;
+  blurhash: string | null;
+}
+
+async function unfurlCard(url: string): Promise<PreviewCard | null> {
+  console.log(`Unfurling ${url}...`);
+  try {
+    const result = await unfurl(url, { fetch });
+    return {
+      type: result.oEmbed?.type || 'link',
+      url: result.canonical_url || url,
+      title: result.oEmbed?.title || result.title || '',
+      description: result.open_graph.description || result.description || '',
+      author_name: result.oEmbed?.author_name || '',
+      author_url: result.oEmbed?.author_url || '',
+      provider_name: result.oEmbed?.provider_name || '',
+      provider_url: result.oEmbed?.provider_url || '',
+      // @ts-expect-error `html` does in fact exist on oEmbed.
+      html: result.oEmbed?.html || '',
+      width: result.oEmbed?.width || 0,
+      height: result.oEmbed?.height || 0,
+      image: result.oEmbed?.thumbnails?.[0].url || null,
+      embed_url: '',
+      blurhash: null,
+    };
+  } catch (_e) {
+    return null;
+  }
+}
+
 export { toAccount, toStatus };