Have you tried using an XML parser instead?

package.json

@@ -22,6 +22,7 @@
     "@peertube/http-signature": "^1.7.0",
     "@projectdysnomia/dysnomia": "github:projectdysnomia/dysnomia#dev",
     "dumpy": "github:Cynosphere/dumpy.js",
+    "fast-xml-parser": "^5.2.0",
     "google-images": "^2.1.0",
     "jimp": "^0.22.7",
     "murmurhash": "^2.0.1",

pnpm-lock.yaml

@@ -20,6 +20,9 @@ importers:
       dumpy:
         specifier: github:Cynosphere/dumpy.js
         version: https://codeload.github.com/Cynosphere/dumpy.js/tar.gz/5fc22353cdcb97084bab572266390e780d9f7a7b(encoding@0.1.13)
+      fast-xml-parser:
+        specifier: ^5.2.0
+        version: 5.2.0
       google-images:
         specifier: ^2.1.0
         version: 2.1.0
@@ -814,6 +817,10 @@ packages:
   fast-levenshtein@2.0.6:
     resolution: {integrity: sha512-DCXu6Ifhqcks7TZKY3Hxp3y6qphY5SJZmrWMDrKcERSOXWQdMhU9Ig/PYrzyw/ul9jOIyh0N4M0tbC5hodg8dw==}
 
+  fast-xml-parser@5.2.0:
+    resolution: {integrity: sha512-Uw9+Mjt4SBRud1IcaYuW/O0lW8SKKdMl5g7g24HiIuyH5fQSD+AVLybSlJtqLYEbytVFjWQa5DMGcNgeksdRBg==}
+    hasBin: true
+
   fastq@1.15.0:
     resolution: {integrity: sha512-wBrocU2LCXXa+lWBt8RoIRD89Fi8OdABODa/kEnyeyjS5aZO5/GNvI5sEINADqP/h8M29UHTHUb53sUu5Ihqdw==}
 
@@ -1445,6 +1452,9 @@ packages:
     resolution: {integrity: sha512-6fPc+R4ihwqP6N/aIv2f1gMH8lOVtWQHoqC4yK6oSDVVocumAsfCqjkXnqiYMhmMwS/mEHLp7Vehlt3ql6lEig==}
     engines: {node: '>=8'}
 
+  strnum@2.0.5:
+    resolution: {integrity: sha512-YAT3K/sgpCUxhxNMrrdhtod3jckkpYwH6JAuwmUdXZsmzH1wUyzTMrrK2wYCEEqlKwrWDd35NeuUkbBy/1iK+Q==}
+
   strtok3@6.3.0:
     resolution: {integrity: sha512-fZtbhtvI9I48xDSywd/somNqgUHl2L2cstmXCCif0itOf96jeW18MBSyrLuNicYQVkvpOxkZtkzujiTJ9LW5Jw==}
     engines: {node: '>=10'}
@@ -2542,6 +2552,10 @@ snapshots:
 
   fast-levenshtein@2.0.6: {}
 
+  fast-xml-parser@5.2.0:
+    dependencies:
+      strnum: 2.0.5
+
   fastq@1.15.0:
     dependencies:
       reusify: 1.0.4
@@ -3244,6 +3258,8 @@ snapshots:
 
   strip-json-comments@3.1.1: {}
 
+  strnum@2.0.5: {}
+
   strtok3@6.3.0:
     dependencies:
       '@tokenizer/token': 0.3.0

src/util/html.js

@@ -1,3 +1,5 @@
+const {XMLParser} = require("fast-xml-parser");
+
 // https://stackoverflow.com/a/39243641
 const htmlEntities = {
   nbsp: " ",
@@ -30,7 +32,91 @@ function parseHtmlEntities(str) {
   });
 }
 
+const parser = new XMLParser({
+  ignoreAttributes: false,
+  preserveOrder: true,
+  unpairedTags: ["hr", "br", "link", "meta"],
+  stopNodes: ["*.pre", "*.script", "*.code"],
+  processEntities: true,
+  htmlEntities: true,
+  attributeNamePrefix: "$",
+  alwaysCreateTextNode: true,
+  trimValues: false,
+});
+
+const tagBlacklist = ["script", "style"];
+function whenYouWalking(tree, images = true, embed = true) {
+  let str = "";
+
+  for (const elem of tree) {
+    if (elem["#text"]) {
+      str += elem["#text"].replaceAll(/[\\*\-_~#<[]/g, "\\$&").replaceAll("||", "|\u200b|");
+    } else if (elem.a) {
+      const content = whenYouWalking(elem.a, images, embed);
+      const link = elem.$href ?? elem[":@"]?.$href;
+      const alt = elem.$alt ?? elem.$title ?? elem[":@"]?.$alt ?? elem[":@"]?.$title;
+
+      if (link == null || !link.startsWith("http")) {
+        str += content;
+      } else if (content == link || link.replace(/^https?:\/\//, "") == content) {
+        str += embed ? link : `<${link}>`;
+      } else {
+        str += `[${content}](${embed ? "" : "<"}${link}${embed ? "" : ">"}${alt != null ? ` "${alt}"` : ""})`;
+      }
+    } else if (elem.b) {
+      str += `**${whenYouWalking(elem.b, images, embed)}**`;
+    } else if (elem.blockquote) {
+      str += `> ${whenYouWalking(elem.blockquote, images, embed).replaceAll(/\n/g, "\n> ")}`;
+    } else if (elem.br) {
+      str += `\n${whenYouWalking(elem.br, images, embed)}`;
+    } else if (elem.code) {
+      str += `\`${whenYouWalking(elem.code, images, embed)}\``;
+    } else if (elem.dd) {
+      str += `\u3000\u3000${whenYouWalking(elem.dd, images, embed).replaceAll(/\n/g, "\n\u3000\u3000")}`;
+    } else if (elem.em) {
+      str += `_${whenYouWalking(elem.em, images, embed)}_`;
+    } else if (elem.i) {
+      str += `_${whenYouWalking(elem.i, images, embed)}_`;
+    } else if (elem.img && images) {
+      const link = elem.$src ?? elem[":@"]?.$src;
+      const alt = elem.$alt ?? elem.$title ?? elem[":@"]?.$alt ?? elem[":@"]?.$title;
+
+      str += `[${alt ?? "<Image>"}](${embed ? "" : "<"}${link}${embed ? "" : ">"})`;
+    } else if (elem.ol) {
+      str += `1. ${whenYouWalking(elem.ol, images, embed)}`;
+    } else if (elem.p) {
+      str += `\n${whenYouWalking(elem.p, images, embed)}\n`;
+    } else if (elem.pre) {
+      str += `\`\`\`\n${whenYouWalking(elem.pre, images, embed)}\`\`\``;
+    } else if (elem.s) {
+      str += `~~${whenYouWalking(elem.s, images, embed)}~~`;
+    } else if (elem.small) {
+      str += `-# ${whenYouWalking(elem.s, images, embed)}`;
+    } else if (elem.u) {
+      str += `__${whenYouWalking(elem.u, images, embed)}__`;
+    } else if (elem.ul) {
+      str += `- ${whenYouWalking(elem.ul, images, embed)}`;
+    } else {
+      for (const [key, val] of Object.entries(elem)) {
+        if (key != ":@" && !key.startsWith("$") && !tagBlacklist.includes(key)) {
+          str += whenYouWalking(val, images, embed);
+          break;
+        }
+      }
+    }
+  }
+
+  return str;
+}
+
 function htmlToMarkdown(str, images = true, embed = true) {
+  if (!str.startsWith("<body>")) str = `<body>${str}</body>`;
+  const tree = parser.parse(str)[0].body;
+
+  return whenYouWalking(tree, images, embed);
+}
+
+function htmlToMarkdownOld(str, images = true, embed = true) {
   str = str.replaceAll("\\", "\\\\");
   str = str.replace(/<style(\s*[^>]+)?>(.|\n)*?<\/style>/gi, "");
   str = str.replace(/<a (\s*[^>]+)?href="([^"]+?)"(\s*[^>]+)?>(.+?)<\/a>/gi, (_, __, url, ___, text) => {
@@ -89,4 +175,4 @@ function htmlToMarkdown(str, images = true, embed = true) {
   return str;
 }
 
-module.exports = {parseHtmlEntities, htmlToMarkdown};
+module.exports = {parseHtmlEntities, htmlToMarkdown, htmlToMarkdownOld};