feat(scripts): strip translator credits and promo lines from fetched and existing EPUBs

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

scripts/clean-chapter.js

@@ -0,0 +1,38 @@
+// Translator credits and promo lines that sites paste into every chapter
+// ("Translator: …", "Discord: https://…", "Join the Discord…", ko-fi/Patreon).
+// A paragraph is dropped only when it is short and the whole thing is promo, so
+// story text that happens to mention a word like "discord" stays.
+// The illustrator's reader.js carries the same rule for in-app reading.
+const MAX_PROMO_LENGTH = 200;
+const URL = /https?:\/\/\S+|\b(dsc|discord)\.gg\/\S*/gi;
+const MAX_LINK_LABEL = 60; // "Discord: <url>", "Ko-Fi: <url>", but not prose with a link
+const PROMO = [
+  /^(discord|patreon|ko-?fi)\b\s*:?/i,
+  /\b(join|support)\b.{0,40}\b(discord|patreon|ko-?fi)\b/i,
+  /\bko-?fi\b|\bpatreon\b/i,
+  /^(translators?|translated by|translation|editors?|edited by|editing|proofreaders?|proofread by|tl|tln|pr|ed)(\s*\/\s*\w+)*\s*[:-]/i,
+  /\bread (the )?(latest|advanced?|more) chapters?\b/i,
+];
+
+export const isPromo = text => {
+  const t = text.replace(/\s+/g, ' ').trim();
+  if (!t || t.length > MAX_PROMO_LENGTH) return false;
+  const label = t.replace(URL, '').trim();
+  if (label.length < t.length && label.length <= MAX_LINK_LABEL) return true;
+  return PROMO.some(r => r.test(t));
+};
+
+/** Remove promo paragraphs from a cheerio document; returns how many. */
+export const stripPromo = $ => {
+  let removed = 0;
+  $('p, div, h3, h4, h5, h6, li').each((_, el) => {
+    const node = $(el);
+    // Innermost blocks only, so a wrapper holding the whole chapter is never dropped.
+    if (node.find('p, div, li').length) return;
+    if (isPromo(node.text())) {
+      node.remove();
+      removed++;
+    }
+  });
+  return removed;
+};

scripts/clean-epub.js

@@ -0,0 +1,41 @@
+// Strip translator credits and promo lines from an existing EPUB, in place.
+// Usage: node scripts/clean-epub.js <file.epub>...
+import JSZip from 'jszip';
+import { load } from 'cheerio';
+import fs from 'fs';
+import { stripPromo } from './clean-chapter.js';
+
+const files = process.argv.slice(2);
+if (!files.length) {
+  console.error('Usage: node scripts/clean-epub.js <file.epub>...');
+  process.exit(2);
+}
+
+for (const file of files) {
+  const zip = await JSZip.loadAsync(fs.readFileSync(file));
+  let removed = 0;
+  for (const name of Object.keys(zip.files)) {
+    if (!/\.x?html?$/i.test(name) || /(^|\/)nav\.xhtml$/i.test(name)) continue;
+    const $ = load(await zip.file(name).async('string'), { xml: true });
+    const n = stripPromo($);
+    if (n) {
+      zip.file(name, $.xml());
+      removed += n;
+    }
+  }
+  if (removed) {
+    const tmp = `${file}.tmp`;
+    // EPUB requires mimetype stored uncompressed; loaded entries lose that option.
+    zip.file('mimetype', 'application/epub+zip', { compression: 'STORE' });
+    fs.writeFileSync(
+      tmp,
+      await zip.generateAsync({
+        type: 'nodebuffer',
+        compression: 'DEFLATE',
+        mimeType: 'application/epub+zip',
+      }),
+    );
+    fs.renameSync(tmp, file);
+  }
+  process.stdout.write(`${file}: removed ${removed} promo paragraphs\n`);
+}

scripts/fetch-novel.js

@@ -6,6 +6,7 @@ import { load } from 'cheerio';
 import fs from 'fs';
 import path from 'path';
 import { loadPlugins, print, ROOT } from './load-plugins.js';
+import { stripPromo } from './clean-chapter.js';
 
 // Be gentle with the source site: few parallel requests, retries with backoff.
 const CONCURRENCY = 3;
@@ -51,10 +52,11 @@ const esc = s =>
   );
 
 // Chapter HTML is tag soup; re-serialise it as well-formed XHTML without
-// scripts or remote resources (an EPUB must be self-contained).
+// scripts, remote resources (an EPUB must be self-contained) or promo lines.
 const toXhtml = html => {
   const $ = load(html, null, false);
   $('script, style, iframe, link, meta, img, video, audio, source').remove();
+  stripPromo($);
   return $.xml();
 };