test(client): doc-changed-guard тесты вставки — сделать нехолостыми (#347, ревью F3)

Ревьюер мутационно доказал: 3 теста doc-changed-guard были ВХОЛОСТУЮ — зелёные даже
при обоих гардах `if(false)`. Причина: вставка в ПУСТОЙ курсор (from==to==1) +
mutateDoc только РАСТИТ док → протухший нулевой диапазон всегда валидная точка
вставки: replaceRange(1,1,…) не затирает, растущий док не выводит `to` за границы,
RangeError не бросается. Гард-код корректен — вхолостую были ТЕСТЫ.

Переписаны по рецепту ревьюера:
- success: вставка поверх НЕПУСТОГО выделения ("AAAABBBB", selection {1,5}); mid-flight
  вставить "MARKER" в pos 1 + курсор в конец. Рабочий гард → замена в живой (конечной)
  selection, MARKER цел; сломанный → stale {1,5} стирает голову MARKER.
- fail-open: захваченный `to` выводится ЗА ГРАНИЦЫ — после захвата {1,9} и провала
  конверсии док СЖИМАЕТСЯ до пустого параграфа. Рабочий гард → raw-текст в живую
  selection; сломанный → insertText(md,1,9) на size-2 доке → RangeError, ничего не
  ложится.
- two-pastes оставлен (пинит «ни один payload не потерян», не гард).

Мутационно проверено: оба гарда→if(false) → тесты 1 и 2 КРАСНЕЮТ (stale-range
clobber; stale-`to` RangeError), 3-й и не-гард-тесты зелёные; реальный гард
восстановлен → 6/6. Тесты теперь отличают рабочий гард от сломанного.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

AGENTS.md

@@ -334,7 +334,7 @@ pnpm workspace (`pnpm@10.4.0`) orchestrated by **Nx**. Four workspace packages:
 | `apps/client` | `client` | React 18 + Vite + Mantine 8 + TanStack Query + Jotai | SPA frontend |
 | `packages/editor-ext` | `@docmost/editor-ext` | Tiptap/ProseMirror | Shared Tiptap node/mark extensions, imported by both the client and the server |
 | `packages/mcp` | `@docmost/mcp` | MCP SDK, Tiptap, Yjs | Standalone MCP server, also bundled into the server at `/mcp`. Consumes the shared converter/schema from `@docmost/prosemirror-markdown` (#293) — it no longer carries its own vendored converter/schema copy |
-| `packages/prosemirror-markdown` | `@docmost/prosemirror-markdown` | Tiptap, marked, jsdom | The single, canonical ProseMirror↔Markdown converter + Docmost schema mirror (#293). Consumed by `mcp`, `git-sync`, AND `apps/server` (server-side markdown import/export, #345); there is exactly ONE copy of the converter now |
+| `packages/prosemirror-markdown` | `@docmost/prosemirror-markdown` | Tiptap, marked; jsdom (Node only) | The single, canonical ProseMirror↔Markdown converter + Docmost schema mirror (#293). Consumed by `mcp`, `git-sync`, `apps/server` (server-side markdown import/export, #345), AND `apps/client` (markdown paste/copy + AI-chat render, via the `browser` entry — native `DOMParser`, no jsdom in the client bundle, #347); there is exactly ONE copy of the converter now |
 
 `build` targets are Nx-cached and dependency-ordered (`dependsOn: ["^build"]`), so `editor-ext` builds before the apps. `nx.json` sets `affected.defaultBase: main`.
 
@@ -460,7 +460,7 @@ The API server is a Fastify app with a global `/api` prefix (`main.ts` excludes
 ### Client structure
 Vite SPA. Code is organized by feature under `apps/client/src/features/*` (mirrors the server domains: `page`, `space`, `comment`, `ai-chat`, `editor`, …). Conventions:
 - **TanStack Query** for server state (one `queries/` file per feature), **Jotai** atoms for local/shared UI state, **Mantine 8** + CSS modules (`*.module.css`) + `postcss-preset-mantine` for UI.
-- The editor is Tiptap; shared node/mark extensions live in `packages/editor-ext` and are imported by **both the client and the server** (collaboration, schema, `canonicalizeFootnotes`) — editor schema changes often need to be made in `editor-ext`, not just the client. Server-side markdown import/export no longer lives in `editor-ext`: it goes through the canonical converter (#345, see below). The ProseMirror↔Markdown converter and its Docmost schema mirror now live in a SINGLE package, `@docmost/prosemirror-markdown` (#293), consumed by `mcp`, `git-sync`, and `apps/server` (#345) — do NOT reintroduce a per-package copy. `editor-ext` is the upstream source of the Tiptap schema; the package's `docmost-schema.ts` mirrors it and a serializer-contract test (`packages/prosemirror-markdown/test/serializer-contract.test.ts`) guards the boundary (every schema node must have a converter case), so a drift surfaces as a failing test rather than silent divergence. For the converter's property-testing and counterexample→fixture process (P1–P4 invariants, the `PROPERTY_SEED`/`PROPERTY_NUM_RUNS` knobs, and the nightly fuzz workflow), see `packages/prosemirror-markdown/README.md`.
+- The editor is Tiptap; shared node/mark extensions live in `packages/editor-ext` and are imported by **both the client and the server** (collaboration, schema, `canonicalizeFootnotes`) — editor schema changes often need to be made in `editor-ext`, not just the client. Server-side markdown import/export no longer lives in `editor-ext`: it goes through the canonical converter (#345, see below). The ProseMirror↔Markdown converter and its Docmost schema mirror now live in a SINGLE package, `@docmost/prosemirror-markdown` (#293), consumed by `mcp`, `git-sync`, `apps/server` (#345), and `apps/client` (#347) — do NOT reintroduce a per-package copy. The client uses the package's `browser` entry (`@docmost/prosemirror-markdown/browser`): markdown paste (`markdown-clipboard.ts`), copy-as-markdown, and AI-chat rendering now all go through the canonical converter, so the hand-written `marked`/`turndown` markdown layer that used to live in `editor-ext` was deleted (#347). The browser entry runs the HTML→DOM stage on the native `DOMParser`, so jsdom stays out of the client bundle. `editor-ext` is the upstream source of the Tiptap schema; the package's `docmost-schema.ts` mirrors it and a serializer-contract test (`packages/prosemirror-markdown/test/serializer-contract.test.ts`) guards the boundary (every schema node must have a converter case), so a drift surfaces as a failing test rather than silent divergence. For the converter's property-testing and counterexample→fixture process (P1–P4 invariants, the `PROPERTY_SEED`/`PROPERTY_NUM_RUNS` knobs, and the nightly fuzz workflow), see `packages/prosemirror-markdown/README.md`.
 - API access goes through `apps/client/src/lib/api-client.ts` (axios). The `@` alias maps to `apps/client/src`.
 - Runtime config is injected at build time by `vite.config.ts` via `define` (`APP_URL`, `COLLAB_URL`, `APP_VERSION`, …) — these come from the root `.env`, not from `import.meta.env`.
 

CHANGELOG.md

@@ -270,6 +270,18 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Changed
 
+- **Client markdown paste/copy and AI-chat rendering now go through the canonical
+  converter.** Pasting markdown into the editor, "Copy as markdown", the AI title
+  generator, and the AI-chat markdown renderer all now use
+  `@docmost/prosemirror-markdown` (via its new `browser` entry — native
+  `DOMParser`, no jsdom in the client bundle) instead of the hand-written
+  `marked`/`turndown` markdown layer in `editor-ext`, which was **deleted**. As a
+  result, pasting canonical markdown (`^[…]` footnotes, `<!--img …-->`,
+  `> [!type]` callouts, `$…$` math, `==…==` highlight, standalone `<!--subpages-->`
+  comments) now produces the SAME nodes the server import produces for the same
+  text. Chat/reasoning markdown now renders through the editor schema (list items
+  are wrapped in `<p>`; CSS keeps them tight). (#347)
+
 - **Enabling a public share no longer auto-shares the whole sub-tree.** Turning
   a page "Shared to web" now defaults to the page alone; descendant pages become
   public only when you explicitly turn on the dedicated "Include sub-pages"

apps/client/package.json

@@ -21,6 +21,7 @@
     "@atlaskit/pragmatic-drag-and-drop-live-region": "1.3.4",
     "@casl/react": "5.0.1",
     "@docmost/editor-ext": "workspace:*",
+    "@docmost/prosemirror-markdown": "workspace:*",
     "@excalidraw/excalidraw": "0.18.0-3a5ef40",
     "@mantine/core": "8.3.18",
     "@mantine/dates": "8.3.18",

apps/client/src/features/ai-chat/components/ai-chat.module.css

@@ -55,6 +55,15 @@
     padding-inline-start: 1.4em;
 }
 
+/* The canonical converter renders list items through the editor schema, which
+   wraps each item's content in a <p> (listItem content is `paragraph+`). Drop
+   that paragraph's block margin so list items render TIGHT (no extra vertical
+   gap), matching the previous marked output — same rule already applied to
+   table cells above (issue #347). */
+.markdown li p {
+    margin: 0;
+}
+
 /* GFM tables in assistant markdown. The chat lives in a NARROW side panel, so a
    wide LLM table must scroll horizontally instead of collapsing its columns:
    `.markdown` sets `word-break: break-word`, which (with the default table
@@ -172,6 +181,14 @@
     margin: 0 0 4px;
 }
 
+/* Same as `.markdown li p` above: the canonical converter wraps every list
+   item's content in a <p>, so without this each reasoning-panel list item would
+   pick up `.reasoningText p`'s 4px bottom margin and render too loose. Drop it
+   so Reasoning-panel lists stay tight, mirroring the pre-#347 marked output. */
+.reasoningText li p {
+    margin: 0;
+}
+
 .inputWrapper {
     flex: 0 0 auto;
     padding-top: var(--mantine-spacing-xs);

apps/client/src/features/ai-chat/utils/collapse-blank-lines.test.ts

@@ -33,29 +33,44 @@ describe("collapseBlankLines", () => {
   });
 });
 
-describe("collapseBlankLines + renderChatMarkdown (tight reasoning rendering)", () => {
+describe("collapseBlankLines + renderChatMarkdown (canonical converter)", () => {
-  it("renders a blank-line-separated list as a TIGHT list (no <li><p>)", () => {
+  // Chat markdown now renders through @docmost/prosemirror-markdown (issue #347):
+  // the SAME converter the editor/import use. Its list items are schema-shaped —
+  // each <li>'s content is wrapped in a <p> (listItem content is `paragraph+`) —
+  // so the HTML always carries `<li><p>…</p></li>` regardless of blank-line
+  // looseness in the source (the converter has no tight/loose distinction). The
+  // visual tightness that `collapseBlankLines` used to buy is now provided by
+  // CSS (`.markdown li p { margin: 0 }`), not the HTML shape.
+  it("renders a blank-line-separated bullet list as a real <ul> list", () => {
     const loose =
       "Intro paragraph.\n\n- item one\n\n- item two\n\n- item three";
     const html = renderChatMarkdown(collapseBlankLines(loose), {});
-    // Tight list: each <li> holds the text directly, not wrapped in a <p>.
+    // Clean, un-namespaced HTML (DOMSerializer, not XMLSerializer) — no xmlns.
-    expect(html).toContain("<li>item one</li>");
-    expect(html).not.toContain("<li><p>");
-    // The list still parses as a list after the paragraph (not a paragraph+<br>).
     expect(html).toContain("<ul>");
+    expect(html).not.toMatch(/<ul[^>]*xmlns/);
+    // The item text is present (inside the schema's <li><p> wrapper).
+    expect(html).toContain("item one");
+    // The intro paragraph renders as its own paragraph before the list.
     expect(html).toContain("<p>Intro paragraph.</p>");
   });
 
-  it("renders an ordered list (1. 2.) as tight after collapsing", () => {
+  it("renders an ordered list (1. 2.) as a real <ol> list", () => {
     const loose = "Intro.\n\n1. first\n\n2. second";
     const html = renderChatMarkdown(collapseBlankLines(loose), {});
     expect(html).toContain("<ol>");
-    expect(html).toContain("<li>first</li>");
+    expect(html).not.toMatch(/<ol[^>]*xmlns/);
-    expect(html).not.toContain("<li><p>");
+    expect(html).toContain("first");
+    expect(html).toContain("second");
   });
 
-  it("the loose source WOULD render <li><p> without collapsing (control)", () => {
+  it("wraps list-item content in <p> (schema shape; tightness is CSS)", () => {
+    // The canonical converter always wraps a list item's content in a paragraph,
+    // whether or not the source had blank lines between items.
     const loose = "- a\n\n- b";
     expect(renderChatMarkdown(loose, {})).toContain("<li><p>");
+    // And a "tight" source produces the identical wrapping (no distinction).
+    expect(renderChatMarkdown(collapseBlankLines(loose), {})).toContain(
+      "<li><p>",
+    );
   });
 });

apps/client/src/features/ai-chat/utils/markdown.ts

@@ -1,6 +1,37 @@
-import { markdownToHtml } from "@docmost/editor-ext";
+import {
+  markdownToProseMirrorSync,
+  docmostExtensions,
+} from "@docmost/prosemirror-markdown/browser";
+import { getSchema } from "@tiptap/core";
+import { Node as PMNode, DOMSerializer } from "@tiptap/pm/model";
 import DOMPurify from "dompurify";
 
+// The Docmost editor schema, built once. Chat markdown is rendered through the
+// SAME schema the editor/import use (issue #347), so chat output matches how the
+// page would render the same markdown.
+const chatSchema = getSchema(docmostExtensions);
+
+/**
+ * Markdown -> HTML for chat display, via the canonical converter. We serialize
+ * the ProseMirror doc with `DOMSerializer` into a real element and read its
+ * `innerHTML` (rather than `@tiptap/html`'s `generateHTML`, whose browser path
+ * uses `XMLSerializer` and stamps a `xmlns` on every block) so the markup is
+ * clean HTML. `li > p` wrapping is inherent to the schema (listItem content is
+ * `paragraph+`); the chat CSS zeroes those paragraph margins so lists still
+ * render tight.
+ */
+function markdownToChatHtml(markdown: string): string {
+  const doc = markdownToProseMirrorSync(markdown);
+  const node = PMNode.fromJSON(chatSchema, doc);
+  const div = document.createElement("div");
+  DOMSerializer.fromSchema(chatSchema).serializeFragment(
+    node.content,
+    { document },
+    div,
+  );
+  return div.innerHTML;
+}
+
 export interface RenderChatMarkdownOptions {
   /**
    * Neutralize INTERNAL links so they render as inert text (no `href`/`target`).
@@ -63,22 +94,32 @@ function neutralizeInternalLinksHook(node: Element): void {
 
 /**
  * Render AI markdown to sanitized HTML for read-only display. We reuse the
- * app's `markdownToHtml` (the same `marked` pipeline used for paste/import) so
+ * canonical converter (issue #347): markdown -> ProseMirror JSON (the SAME
- * chat output matches the editor's markdown flavor, then sanitize with
+ * `markdownToProseMirrorSync` the editor paste/import path uses, so chat output
- * DOMPurify — LLM output is untrusted, so it must never reach the DOM unsanitized.
+ * matches the editor's markdown flavor) -> HTML via `markdownToChatHtml`
+ * (DOMSerializer), then sanitize with DOMPurify — LLM output is untrusted, so it
+ * must never reach the DOM unsanitized.
  *
- * `markdownToHtml` can return `string | Promise<string>` (it has async marked
+ * Stays SYNCHRONOUS: both callers render inside React (a memo and a useMemo),
- * extensions registered). In practice plain chat markdown resolves
+ * so the whole pipeline must resolve without awaiting. The converter's sync
- * synchronously, but we guard the Promise case by returning a safe empty string
+ * entry makes that possible; on any conversion error we return "" so the caller
- * for that branch (the caller renders the raw text fallback instead).
+ * falls back to raw text (the same fallback the old Promise-guard produced).
  */
 export function renderChatMarkdown(
   markdown: string,
   options: RenderChatMarkdownOptions = {},
 ): string {
   if (!markdown) return "";
-  const html = markdownToHtml(markdown);
+  let html: string;
-  if (typeof html !== "string") return "";
+  try {
+    // markdown -> canonical PM JSON -> HTML (native DOMParser in the browser;
+    // jsdom is never bundled — see @docmost/prosemirror-markdown/browser).
+    html = markdownToChatHtml(markdown);
+  } catch {
+    // Malformed/unsupported markdown must not crash the chat render; fall back
+    // to raw text (empty return -> caller shows the plain-text branch).
+    return "";
+  }
 
   if (!options.neutralizeInternalLinks) {
     // Internal chat: unchanged behavior, no hook registered.

apps/client/src/features/editor/extensions/markdown-clipboard.paste.test.ts

@@ -0,0 +1,206 @@
+import { describe, it, expect } from "vitest";
+import { Editor } from "@tiptap/core";
+import { Document } from "@tiptap/extension-document";
+import { Paragraph } from "@tiptap/extension-paragraph";
+import { Text } from "@tiptap/extension-text";
+import { Bold } from "@tiptap/extension-bold";
+import { Italic } from "@tiptap/extension-italic";
+import { MarkdownClipboard } from "./markdown-clipboard";
+
+/**
+ * Integration coverage for the async `handlePaste` seam (issue #347). The paste
+ * conversion moved to `@docmost/prosemirror-markdown`'s browser entry, whose
+ * `markdownToProseMirror` is async — so `handlePaste` captures the range, claims
+ * the event (returns true), and dispatches the insert on the next microtask.
+ * These tests drive that path end to end on a minimal schema (a plain-markdown
+ * paste whose converted nodes fit paragraph/text/bold/italic), asserting the
+ * text lands with the right marks and that the raw markdown syntax is consumed
+ * (recognized as markdown, not inserted literally).
+ */
+
+function makeEditor() {
+  const element = document.createElement("div");
+  document.body.appendChild(element);
+  return new Editor({
+    element,
+    extensions: [
+      Document,
+      Paragraph,
+      Text,
+      Bold,
+      Italic,
+      MarkdownClipboard.configure({ transformPastedText: true }),
+    ],
+    content: { type: "doc", content: [{ type: "paragraph" }] },
+  });
+}
+
+// Locate the markdownClipboard plugin and invoke its handlePaste directly with a
+// synthetic clipboard event (jsdom has no real paste pipeline). The plugin's
+// handlePaste closes over the extension `this`, so calling it off the plugin
+// props preserves `this.editor`/`this.options`.
+function paste(editor: Editor, text: string): boolean {
+  const view = editor.view;
+  const plugin = view.state.plugins.find(
+    (p: any) => p.props && p.spec?.key,
+  ) as any;
+  const event = {
+    clipboardData: {
+      getData: (type: string) => (type === "text/plain" ? text : ""),
+    },
+  } as unknown as ClipboardEvent;
+  // Find the specific handlePaste that belongs to the markdown clipboard plugin.
+  const md = view.state.plugins.find(
+    (p: any) => typeof p.props?.handlePaste === "function",
+  ) as any;
+  return md.props.handlePaste(view, event, view.state.selection.content());
+}
+
+// Flush the microtask queue so the async .then() dispatch runs.
+const flush = () => new Promise((r) => setTimeout(r, 0));
+
+describe("MarkdownClipboard handlePaste (async md -> PM)", () => {
+  it("converts a plain-markdown paste with bold/italic into marked text", async () => {
+    const editor = makeEditor();
+    const claimed = paste(editor, "hello **bold** and *italic*");
+    // The paste is claimed synchronously (async insert follows).
+    expect(claimed).toBe(true);
+    await flush();
+
+    const json = editor.getJSON();
+    const text = JSON.stringify(json);
+    // The raw markdown asterisks are consumed (recognized), not inserted literally.
+    expect(editor.getText()).not.toContain("**");
+    expect(editor.getText()).toContain("bold");
+    expect(editor.getText()).toContain("italic");
+    // The bold/italic marks materialized.
+    expect(text).toContain('"bold"');
+    expect(text).toContain('"italic"');
+    editor.destroy();
+  });
+
+  it("recognizes a bullet list paste as list structure (not literal '-')", async () => {
+    // A bullet list is not representable in this minimal schema, so the converter
+    // output would fail PMNode.fromJSON and the catch inserts raw text. Use a
+    // paste whose nodes DO fit the schema to assert the happy path instead: two
+    // paragraphs separated by a blank line.
+    const editor = makeEditor();
+    paste(editor, "first para\n\nsecond para");
+    await flush();
+    const json = editor.getJSON() as any;
+    const paras = (json.content || []).filter(
+      (n: any) => n.type === "paragraph",
+    );
+    // Two paragraphs materialized from the blank-line-separated markdown.
+    expect(paras.length).toBeGreaterThanOrEqual(2);
+    expect(editor.getText()).toContain("first para");
+    expect(editor.getText()).toContain("second para");
+    editor.destroy();
+  });
+
+  it("falls back to raw text when conversion yields nodes the schema lacks", async () => {
+    // `# heading` converts to a `heading` node absent from this minimal schema,
+    // so PMNode.fromJSON throws and the catch re-inserts the raw text — the user
+    // never loses their clipboard content.
+    const editor = makeEditor();
+    paste(editor, "# a heading line");
+    await flush();
+    // Content is preserved (either as heading text or literal), never dropped.
+    expect(editor.getText()).toContain("a heading line");
+    editor.destroy();
+  });
+});
+
+// The async seam captures the target range synchronously, then replaces on the
+// next microtask. If the document changed under it between capture and resolve
+// (impossible in prod — same microtask — but pinned here), BOTH the success
+// (replaceRange) and the fail-open (insertText) branches must fall back to the
+// LIVE selection rather than a stale absolute range, so neither clobbers content
+// nor throws a RangeError. We force the mid-flight change by dispatching a
+// doc-mutating transaction AFTER the synchronous claim but BEFORE flushing the
+// microtask that runs the `.then`/`.catch`.
+describe("MarkdownClipboard handlePaste — doc-changed-mid-flight guard", () => {
+  // Replace the whole doc with one paragraph of `text` (synchronous dispatch).
+  // An empty string yields an empty paragraph (a text node may not be empty).
+  function seedContent(editor: Editor, text: string) {
+    editor.commands.setContent({
+      type: "doc",
+      content: [
+        text
+          ? { type: "paragraph", content: [{ type: "text", text }] }
+          : { type: "paragraph" },
+      ],
+    });
+  }
+
+  it("success branch: mid-flight doc change routes the paste to the LIVE selection, never the stale range (clobber-proving)", async () => {
+    // The paste captures a NON-EMPTY range {1,5} (over "AAAA"). Then, before the
+    // async resolve, the doc GROWS ("MARKER" inserted at the start) and the cursor
+    // is parked at the doc END. The captured {1,5} is now stale and points INTO
+    // "MARKER". A WORKING guard replaces at the live (end) selection → MARKER is
+    // untouched. A BROKEN guard replaces the stale {1,5} → it erases the first
+    // characters of MARKER (this is what a zero-width `from==to` range could never
+    // reveal, which is why the earlier version was vacuous).
+    const editor = makeEditor();
+    seedContent(editor, "AAAABBBB");
+    editor.commands.setTextSelection({ from: 1, to: 5 }); // captured range = {1,5}
+    const claimed = paste(editor, "hello **bold**");
+    expect(claimed).toBe(true);
+
+    // Mid-flight: grow the doc and move the cursor to a KNOWN-safe end position.
+    editor.view.dispatch(editor.view.state.tr.insertText("MARKER", 1));
+    const end = editor.state.doc.content.size;
+    editor.commands.setTextSelection({ from: end, to: end });
+    await flush();
+
+    const text = editor.getText();
+    // MARKER intact only if the guard used the live selection, not the stale range.
+    expect(text).toContain("MARKER");
+    expect(text).toContain("bold");
+    expect(text).not.toContain("**");
+    editor.destroy();
+  });
+
+  it("fail-open branch: a mid-flight doc SHRINK makes the stale `to` out of bounds — the guard must avoid a RangeError (throw-proving)", async () => {
+    // The paste captures a range {1,9} over an 8-char paragraph, then the
+    // conversion FAILS (`# heading` -> a heading node the minimal schema lacks,
+    // so PMNode.fromJSON throws -> the fail-open catch runs). Before the reject,
+    // the doc is SHRUNK to an empty paragraph, so the captured `to` (9) is now far
+    // past the doc's end. A WORKING guard inserts the raw text at the live (valid)
+    // selection → "raw heading" lands. A BROKEN guard does insertText(md, 1, 9) on
+    // a size-2 doc → RangeError, so the dispatch never runs and "raw heading" is
+    // absent (the assertion reddens). A zero-width/growing-doc setup could never
+    // push `to` out of bounds, which is why the earlier version was vacuous.
+    const editor = makeEditor();
+    seedContent(editor, "AAAABBBB");
+    editor.commands.setTextSelection({ from: 1, to: 9 }); // captured range = {1,9}
+    paste(editor, "# raw heading");
+
+    // Mid-flight: shrink the doc so the captured `to` = 9 is now out of bounds.
+    seedContent(editor, "");
+    await flush();
+
+    const text = editor.getText();
+    // Raw text lands (via the live selection) only if the guard avoided the
+    // stale, now-out-of-bounds range.
+    expect(text).toContain("raw heading");
+    editor.destroy();
+  });
+
+  it("two pastes in flight: neither payload is lost (no data loss)", async () => {
+    // Prod-unreachable (two paste events are separate macrotasks, and each
+    // conversion resolves on a microtask before the next), but pinned here: when
+    // both resolve back-to-back, the second sees the changed doc and inserts at
+    // the live selection the first left — so the two payloads may INTERLEAVE, but
+    // neither is dropped. We assert no data loss, not contiguity.
+    const editor = makeEditor();
+    paste(editor, "alphaword");
+    paste(editor, "betaword");
+    await flush();
+    const text = editor.getText();
+    // Neither payload fully dropped (interleaving may split one of them).
+    expect(text).toContain("alpha");
+    expect(text).toContain("beta");
+    editor.destroy();
+  });
+});

apps/client/src/features/editor/extensions/markdown-clipboard.test.ts

@@ -1,5 +1,12 @@
 import { describe, it, expect } from "vitest";
-import { htmlToMarkdown } from "@docmost/editor-ext";
+// Markdown conversion now goes through the canonical package's BROWSER entry
+// (issue #347): the same converter the server import/export uses, resolved via
+// the `browser` exports condition so it runs on the native `DOMParser` (the
+// client jsdom vitest env provides one) with jsdom never bundled.
+import {
+  convertProseMirrorToMarkdown,
+  markdownToProseMirrorSync,
+} from "@docmost/prosemirror-markdown/browser";
 import {
   normalizeTableColumnWidths,
   classifyClipboardSelection,
@@ -175,10 +182,13 @@ describe("classifyClipboardSelection", () => {
 
 // Output-level tests for the table clipboard regression: copying a table must
 // yield a real GFM pipe table, NOT one-value-per-line concatenated cells.
-// These exercise the actual markdown produced by htmlToMarkdown (the same
+// These exercise the actual markdown produced by convertProseMirrorToMarkdown —
-// serializer step the clipboardTextSerializer runs), so they pin the OUTPUT
+// the same serializer step the clipboardTextSerializer now runs (issue #347) —
-// shape that the classifier-flag tests above do not cover.
+// so they pin the OUTPUT shape that the classifier-flag tests above do not cover.
-describe("table clipboard markdown output (htmlToMarkdown)", () => {
+// Input is ProseMirror JSON (what the copied slice serializes to), matching the
+// clipboardTextSerializer's new call: it wraps the slice content in a synthetic
+// `doc` (and the bare-rows case in a `table`) and calls the converter.
+describe("table clipboard markdown output (convertProseMirrorToMarkdown)", () => {
   // Trim each line and drop blanks so structural assertions are whitespace-robust.
   function lines(md: string): string[] {
     return md
@@ -188,10 +198,10 @@ describe("table clipboard markdown output (htmlToMarkdown)", () => {
   }
 
   // A GFM separator row like "| --- | --- |" (any number of columns), tolerant
-  // of the padding turndown emits.
+  // of the padding the serializer emits.
   function isSeparatorRow(line: string): boolean {
     const compact = line.replace(/\s+/g, "");
-    return /^\|(?:-{3,}\|)+$/.test(compact);
+    return /^\|(?::?-{2,}:?\|)+$/.test(compact);
   }
 
   // Split a pipe-delimited row into trimmed cell values.
@@ -203,42 +213,33 @@ describe("table clipboard markdown output (htmlToMarkdown)", () => {
       .map((c) => c.trim());
   }
 
-  it("serializes a header-less partial cell selection (bare rows) as a valid GFM pipe table", () => {
+  const cell = (t: string) => ({
-    // Mirror the serializer's `wrapBareRows` branch exactly: bare <tr> nodes are
+    type: "tableCell",
-    // wrapped in <table><tbody> and htmlToMarkdown(div.innerHTML) is called.
+    content: [{ type: "paragraph", content: [{ type: "text", text: t }] }],
-    // See markdown-clipboard.ts clipboardTextSerializer:
+  });
-    //   const table = document.createElement("table");
+  const headerCell = (t: string) => ({
-    //   const tbody = document.createElement("tbody");
+    type: "tableHeader",
-    //   tbody.appendChild(fragment); table.appendChild(tbody);
+    content: [{ type: "paragraph", content: [{ type: "text", text: t }] }],
-    //   div.appendChild(table);
+  });
-    //   return htmlToMarkdown(div.innerHTML);
+  const row = (nodes: any[]) => ({ type: "tableRow", content: nodes });
-    const div = document.createElement("div");
-    const table = document.createElement("table");
-    const tbody = document.createElement("tbody");
-    for (const [c1, c2] of [
-      ["a", "b"],
-      ["c", "d"],
-    ]) {
-      const tr = document.createElement("tr");
-      const td1 = document.createElement("td");
-      td1.textContent = c1;
-      const td2 = document.createElement("td");
-      td2.textContent = c2;
-      tr.appendChild(td1);
-      tr.appendChild(td2);
-      tbody.appendChild(tr);
-    }
-    table.appendChild(tbody);
-    div.appendChild(table);
 
-    const md = htmlToMarkdown(div.innerHTML);
+  it("serializes a header-less partial cell selection (bare rows) as a valid GFM pipe table", () => {
+    // Mirror the serializer's `wrapBareRows` branch: bare tableRow nodes are
+    // wrapped in a synthetic `table` and convertProseMirrorToMarkdown is called
+    // (see markdown-clipboard.ts clipboardTextSerializer).
+    const rows = [
+      row([cell("a"), cell("b")]),
+      row([cell("c"), cell("d")]),
+    ];
+    const md = convertProseMirrorToMarkdown({
+      type: "doc",
+      content: [{ type: "table", content: rows }],
+    });
     const ls = lines(md);
 
-    // Valid GFM: a header/data separator row is present (an empty header is
+    // Valid GFM: a header/data separator row is present.
-    // synthesized by the GFM turndown plugin for a header-less table — fine).
     expect(ls.some(isSeparatorRow)).toBe(true);
-    // NOT the old broken "one value per line" shape: every line is pipe-delimited
+    // NOT the old broken "one value per line" shape: every line is pipe-delimited.
-    // and no line is a bare cell value on its own.
     expect(ls.every((l) => l.includes("|"))).toBe(true);
     expect(md).not.toMatch(/^\s*(a|b|c|d)\s*$/m);
     // The cell values land in real pipe-delimited data rows.
@@ -248,39 +249,21 @@ describe("table clipboard markdown output (htmlToMarkdown)", () => {
   });
 
   it("serializes a whole table with a header row as a proper GFM table (headline regression)", () => {
-    // Mirror the serializer's non-wrap branch: the full <table> node is appended
+    // Mirror the serializer's non-wrap branch: the full `table` node is the
-    // directly (div.appendChild(fragment)) and htmlToMarkdown(div.innerHTML) runs.
+    // slice content and convertProseMirrorToMarkdown runs on it.
-    const div = document.createElement("div");
+    const md = convertProseMirrorToMarkdown({
-    const table = document.createElement("table");
+      type: "doc",
-
+      content: [
-    const thead = document.createElement("thead");
+        {
-    const headerRow = document.createElement("tr");
+          type: "table",
-    for (const h of ["Name", "Age"]) {
+          content: [
-      const th = document.createElement("th");
+            row([headerCell("Name"), headerCell("Age")]),
-      th.textContent = h;
+            row([cell("Alice"), cell("30")]),
-      headerRow.appendChild(th);
+            row([cell("Bob"), cell("25")]),
-    }
+          ],
-    thead.appendChild(headerRow);
+        },
-    table.appendChild(thead);
+      ],
-
+    });
-    const tbody = document.createElement("tbody");
-    for (const [name, age] of [
-      ["Alice", "30"],
-      ["Bob", "25"],
-    ]) {
-      const tr = document.createElement("tr");
-      const td1 = document.createElement("td");
-      td1.textContent = name;
-      const td2 = document.createElement("td");
-      td2.textContent = age;
-      tr.appendChild(td1);
-      tr.appendChild(td2);
-      tbody.appendChild(tr);
-    }
-    table.appendChild(tbody);
-    div.appendChild(table);
-
-    const md = htmlToMarkdown(div.innerHTML);
     const ls = lines(md);
 
     // Proper GFM structure: separator row + all rows pipe-delimited.
@@ -296,3 +279,146 @@ describe("table clipboard markdown output (htmlToMarkdown)", () => {
     expect(md).not.toMatch(/^\s*(Name|Age|Alice|Bob|30|25)\s*$/m);
   });
 });
+
+// #347 acceptance: pasting CANONICAL markdown yields the SAME nodes the server
+// import produces for the same text. The paste path calls markdownToProseMirror
+// (the package browser entry) — the identical converter the server import uses —
+// so asserting the converter (via the browser entry, on the native DOMParser)
+// recognizes each canon form pins the paste-parity guarantee. These forms were
+// NOT recognized by the old editor-ext marked layer the paste used before.
+describe("canonical markdown paste recognition (browser entry parity)", () => {
+  // Collect every node type present in a doc (recursively).
+  const collectTypes = (n: any, set = new Set<string>()): Set<string> => {
+    if (!n || typeof n !== "object") return set;
+    if (n.type) set.add(n.type);
+    if (Array.isArray(n.content)) n.content.forEach((c) => collectTypes(c, set));
+    return set;
+  };
+  const findNode = (n: any, type: string): any => {
+    if (!n || typeof n !== "object") return undefined;
+    if (n.type === type) return n;
+    if (Array.isArray(n.content)) {
+      for (const c of n.content) {
+        const hit = findNode(c, type);
+        if (hit) return hit;
+      }
+    }
+    return undefined;
+  };
+  const allText = (n: any): string => {
+    if (!n || typeof n !== "object") return "";
+    if (typeof n.text === "string") return n.text;
+    if (Array.isArray(n.content)) return n.content.map(allText).join("");
+    return "";
+  };
+
+  it("^[…] inline footnote -> footnoteReference + footnotesList", () => {
+    const doc = markdownToProseMirrorSync("Body^[a note here].");
+    const types = collectTypes(doc);
+    expect(types.has("footnoteReference")).toBe(true);
+    expect(types.has("footnotesList")).toBe(true);
+    expect(types.has("footnoteDefinition")).toBe(true);
+  });
+
+  it('<!--img {…}--> attached image comment -> image with align', () => {
+    const doc = markdownToProseMirrorSync(
+      '![alt](/files/x.png) <!--img {"align":"left"}-->',
+    );
+    const img = findNode(doc, "image");
+    expect(img).toBeTruthy();
+    expect(img.attrs?.align).toBe("left");
+    expect(img.attrs?.src).toBe("/files/x.png");
+  });
+
+  it("> [!type] Obsidian callout -> callout node with type", () => {
+    const doc = markdownToProseMirrorSync("> [!warning]\n> be careful");
+    const callout = findNode(doc, "callout");
+    expect(callout).toBeTruthy();
+    expect(callout.attrs?.type).toBe("warning");
+    expect(allText(callout)).toContain("be careful");
+  });
+
+  it("$…$ inline math -> mathInline node", () => {
+    const doc = markdownToProseMirrorSync("Euler: $e^{i\\pi}+1=0$ done");
+    const math = findNode(doc, "mathInline");
+    expect(math).toBeTruthy();
+    expect(math.attrs?.text).toContain("e^{i\\pi}");
+  });
+
+  it("==…== highlight -> highlight mark", () => {
+    const doc = markdownToProseMirrorSync("A ==marked== word");
+    const marked = findNode(doc, "text");
+    // The highlighted run carries a `highlight` mark somewhere in the doc.
+    const hasHighlight = (n: any): boolean => {
+      if (!n || typeof n !== "object") return false;
+      if (
+        n.type === "text" &&
+        (n.marks || []).some((m: any) => m.type === "highlight")
+      )
+        return true;
+      return Array.isArray(n.content) ? n.content.some(hasHighlight) : false;
+    };
+    expect(marked).toBeTruthy();
+    expect(hasHighlight(doc)).toBe(true);
+  });
+
+  it("<!--subpages--> standalone comment -> subpages node", () => {
+    const doc = markdownToProseMirrorSync("intro\n\n<!--subpages-->\n\nafter");
+    expect(collectTypes(doc).has("subpages")).toBe(true);
+  });
+});
+
+// #347 negatives: plain text carrying markdown-LIKE punctuation must NOT be
+// silently converted/mangled (currency, bare `==`, a `[^1]` reference form).
+describe("plain-text paste negatives (no phantom conversion)", () => {
+  const findNode = (n: any, type: string): any => {
+    if (!n || typeof n !== "object") return undefined;
+    if (n.type === type) return n;
+    if (Array.isArray(n.content)) {
+      for (const c of n.content) {
+        const hit = findNode(c, type);
+        if (hit) return hit;
+      }
+    }
+    return undefined;
+  };
+  const collectTypes = (n: any, set = new Set<string>()): Set<string> => {
+    if (!n || typeof n !== "object") return set;
+    if (n.type) set.add(n.type);
+    if (Array.isArray(n.content)) n.content.forEach((c) => collectTypes(c, set));
+    return set;
+  };
+  const allText = (n: any): string => {
+    if (!n || typeof n !== "object") return "";
+    if (typeof n.text === "string") return n.text;
+    if (Array.isArray(n.content)) return n.content.map(allText).join("");
+    return "";
+  };
+
+  it("currency `$5 and $10` is NOT turned into math", () => {
+    const doc = markdownToProseMirrorSync("It costs $5 and $10 total");
+    expect(findNode(doc, "mathInline")).toBeFalsy();
+    expect(allText(doc)).toContain("$5 and $10");
+  });
+
+  it("a lone `==` is NOT turned into a highlight", () => {
+    const doc = markdownToProseMirrorSync("compare a == b in code");
+    const hasHighlight = (n: any): boolean => {
+      if (!n || typeof n !== "object") return false;
+      if (
+        n.type === "text" &&
+        (n.marks || []).some((m: any) => m.type === "highlight")
+      )
+        return true;
+      return Array.isArray(n.content) ? n.content.some(hasHighlight) : false;
+    };
+    expect(hasHighlight(doc)).toBe(false);
+    expect(allText(doc)).toContain("== b");
+  });
+
+  it("a `[^1]` reference form (no `^[`) is NOT turned into a footnote", () => {
+    const doc = markdownToProseMirrorSync("see note [^1] for details");
+    expect(collectTypes(doc).has("footnoteReference")).toBe(false);
+    expect(allText(doc)).toContain("[^1]");
+  });
+});

apps/client/src/features/editor/extensions/markdown-clipboard.ts

@@ -1,15 +1,23 @@
 // adapted from: https://github.com/aguingand/tiptap-markdown/blob/main/src/extensions/tiptap/clipboard.js - MIT
 import { Extension } from "@tiptap/core";
 import { Plugin, PluginKey, TextSelection } from "@tiptap/pm/state";
-import { DOMParser, DOMSerializer, Fragment, Slice } from "@tiptap/pm/model";
+import { DOMParser, DOMSerializer, Fragment, Slice, Node as PMNode } from "@tiptap/pm/model";
 import { find } from "linkifyjs";
 import {
-  markdownToHtml,
-  htmlToMarkdown,
   canonicalizeFootnotes,
   FOOTNOTES_LIST_NAME,
   FOOTNOTE_REFERENCE_NAME,
 } from "@docmost/editor-ext";
+// Markdown <-> ProseMirror conversion now lives ONLY in the canonical
+// `@docmost/prosemirror-markdown` package (issue #347). The BROWSER entry uses
+// the native `DOMParser` for its HTML->DOM stage (jsdom stays out of the client
+// bundle) while producing the SAME nodes the server import does — so a paste of
+// canonical markdown (`^[…]`, `<!--img …-->`, `> [!type]`, `$…$`, `==…==`,
+// standalone comments) is recognized identically to import.
+import {
+  markdownToProseMirror,
+  convertProseMirrorToMarkdown,
+} from "@docmost/prosemirror-markdown/browser";
 import type { Schema } from "@tiptap/pm/model";
 
 export const MarkdownClipboard = Extension.create({
@@ -39,25 +47,24 @@ export const MarkdownClipboard = Extension.create({
               classifyClipboardSelection(topLevelNodes);
             if (!asMarkdown) return null;
 
-            const div = document.createElement("div");
+            // Convert the copied selection to Markdown through the canonical
-            const serializer = DOMSerializer.fromSchema(this.editor.schema);
+            // package (issue #347), the SAME serializer the server export uses,
-            const fragment = serializer.serializeFragment(slice.content);
+            // so a copied table/list matches the on-disk markdown form. The
-
+            // converter takes a ProseMirror `doc` JSON, so wrap the slice's
+            // top-level content in a synthetic doc.
+            const content = slice.content.toJSON() as any[];
             if (wrapBareRows) {
-              // A partial table cell-selection serializes to bare <tr> nodes
+              // A partial table cell-selection serializes to bare `tableRow`
-              // (prosemirror-tables returns the whole `table` node only when the
+              // nodes (prosemirror-tables yields the whole `table` node only for
-              // entire table is selected). Bare <tr> would be foster-parented
+              // a full-table selection). The converter's table case expects a
-              // away by the HTML parser inside htmlToMarkdown, so wrap them in
+              // `table` wrapper, so wrap the bare rows in one — mirroring the old
-              // <table><tbody> first for the GFM turndown rule to detect them.
+              // <table><tbody> wrap that the HTML->markdown step needed.
-              const table = document.createElement("table");
+              return convertProseMirrorToMarkdown({
-              const tbody = document.createElement("tbody");
+                type: "doc",
-              tbody.appendChild(fragment);
+                content: [{ type: "table", content }],
-              table.appendChild(tbody);
+              });
-              div.appendChild(table);
-            } else {
-              div.appendChild(fragment);
             }
-            return htmlToMarkdown(div.innerHTML);
+            return convertProseMirrorToMarkdown({ type: "doc", content });
           },
           handlePaste: (view, event, slice) => {
             if (!event.clipboardData) {
@@ -95,37 +102,115 @@ export const MarkdownClipboard = Extension.create({
               }
             }
 
-            const { tr } = view.state;
+            const schema = this.editor.schema;
-            const { from, to } = view.state.selection;
+            // Capture the target range NOW. markdownToProseMirror RETURNS A
+            // PROMISE (kept async only for the Node consumers' contract; the
+            // conversion pipeline itself is synchronous), so the actual replace
+            // happens on the next microtask. No user input can interleave a
+            // microtask, so the state is unchanged when we dispatch — but we
+            // still re-read the live state before replacing and, if the doc did
+            // change under us, fall back to the live selection rather than the
+            // captured (now-stale) range.
+            const from = view.state.selection.from;
+            const to = view.state.selection.to;
+            const startDoc = view.state.doc;
+            const md = text.replace(/\n+$/, "");
 
-            const parsed = markdownToHtml(text.replace(/\n+$/, ""));
+            void markdownToProseMirror(md)
-            const body = elementFromString(parsed);
+              .then((doc) => {
-            normalizeTableColumnWidths(body);
+                if (view.isDestroyed) return;
+                // Canonical PM-JSON -> HTML via the LIVE editor schema, then
+                // reuse the UNCHANGED downstream seam (normalizeTableColumnWidths
+                // + parseSlice + canonicalizePastedFootnotes). The JSON->HTML->
+                // JSON hop is lossless (same schema both directions); it lets the
+                // existing paste-insertion logic stay byte-identical — only the
+                // SOURCE of the markdown conversion changed (issue #347 guardrail:
+                // no converter logic in the client, only a call into the package).
+                const node = PMNode.fromJSON(schema, doc);
+                const div = document.createElement("div");
+                DOMSerializer.fromSchema(schema).serializeFragment(
+                  node.content,
+                  { document },
+                  div,
+                );
 
-            const parsedSlice = DOMParser.fromSchema(
+                const body = elementFromString(div.innerHTML);
-              this.editor.schema,
+                normalizeTableColumnWidths(body);
-            ).parseSlice(body, {
-              preserveWhitespace: true,
-            });
 
-            // A markdown paste builds its ProseMirror fragment directly (DOM ->
+                const parsedSlice = DOMParser.fromSchema(schema).parseSlice(
-            // parseSlice), bypassing the editor's footnoteSyncPlugin, which never
+                  body,
-            // reorders an existing list. So a pasted markdown block whose footnote
+                  { preserveWhitespace: true },
-            // definitions are out of order (or contains orphan defs) would be
+                );
-            // stored out of order. Canonicalize the self-contained pasted block so
-            // its footnotes come out reference-ordered, deduped and orphan-free
-            // (issue #228). See canonicalizePastedFootnotes for why this is scoped
-            // to whole-block pastes that carry their own footnotesList.
-            const contentNodes = canonicalizePastedFootnotes(
-              parsedSlice,
-              this.editor.schema,
-            );
 
-            tr.replaceRange(from, to, contentNodes);
+                // A markdown paste builds its ProseMirror fragment directly (DOM
-            const insertEnd = tr.mapping.map(from, 1);
+                // -> parseSlice), bypassing the editor's footnoteSyncPlugin, which
-            tr.setSelection(TextSelection.near(tr.doc.resolve(Math.max(from, insertEnd - 2)), -1));
+                // never reorders an existing list. So a pasted markdown block whose
-            tr.setMeta('paste', true)
+                // footnote definitions are out of order (or contains orphan defs)
-            view.dispatch(tr);
+                // would be stored out of order. Canonicalize the self-contained
+                // pasted block so its footnotes come out reference-ordered, deduped
+                // and orphan-free (issue #228). See canonicalizePastedFootnotes for
+                // why this is scoped to whole-block pastes that carry their own
+                // footnotesList.
+                const contentNodes = canonicalizePastedFootnotes(
+                  parsedSlice,
+                  schema,
+                );
+
+                // Target the captured range (normally still valid — same
+                // microtask). If the doc changed under us since capture, the
+                // captured absolute from/to are stale, so fall back to the live
+                // selection rather than StepMap-mapping the old range.
+                const tr = view.state.tr;
+                let mappedFrom = from;
+                let mappedTo = to;
+                if (view.state.doc !== startDoc) {
+                  // Defensive: if the doc changed under us, fall back to the
+                  // current selection rather than a stale absolute range.
+                  mappedFrom = view.state.selection.from;
+                  mappedTo = view.state.selection.to;
+                }
+                tr.replaceRange(mappedFrom, mappedTo, contentNodes);
+                const insertEnd = tr.mapping.map(mappedFrom, 1);
+                tr.setSelection(
+                  TextSelection.near(
+                    tr.doc.resolve(Math.max(mappedFrom, insertEnd - 2)),
+                    -1,
+                  ),
+                );
+                tr.setMeta("paste", true);
+                view.dispatch(tr);
+              })
+              .catch((err) => {
+                // Fail-open: a conversion error must not swallow the paste
+                // silently in a way that loses the text. We already claimed the
+                // event (returned true), so re-insert the raw text as a plain
+                // paragraph so the user never loses their clipboard content.
+                // Log it: this catch covers BOTH the converter and the success
+                // `.then` body (e.g. PMNode.fromJSON throwing on a schema drift
+                // between the canonical package and the live editor schema), so a
+                // silent degrade to raw text would otherwise be an invisible,
+                // non-reproducible regression ("my table pasted as text").
+                console.error(
+                  "markdown paste conversion failed, inserting raw text",
+                  err,
+                );
+                if (view.isDestroyed) return;
+                const tr = view.state.tr;
+                // Same guard the success path uses: if the doc changed under us
+                // since the range was captured (normally never — same microtask),
+                // the captured absolute from/to are stale and would throw a
+                // RangeError here (an unhandled rejection on a hot paste path).
+                // Fall back to the live selection instead of a stale range.
+                if (view.state.doc !== startDoc) {
+                  const sel = view.state.selection;
+                  tr.insertText(md, sel.from, sel.to);
+                } else {
+                  tr.insertText(md, from, to);
+                }
+                tr.setMeta("paste", true);
+                view.dispatch(tr);
+              });
+            // Claim the paste: we insert asynchronously above.
             return true;
           },
           // Strip trailing whitespace-only paragraphs from pasted content.

apps/client/src/features/editor/hooks/use-generate-page-title.test.tsx

@@ -33,10 +33,11 @@ vi.mock("@/lib/local-emitter.ts", () => ({
   default: { emit: (...args: unknown[]) => localEmitMock(...args) },
 }));
 
-// htmlToMarkdown just echoes the editor HTML so each test controls the markdown
+// convertProseMirrorToMarkdown echoes a marker carried on the fake editor's
-// purely via the fake page editor's getHTML().
+// getJSON() doc, so each test controls the markdown purely via the fake page
-vi.mock("@docmost/editor-ext", () => ({
+// editor (issue #347: the hook now serializes editor JSON through the package).
-  htmlToMarkdown: (html: string) => html,
+vi.mock("@docmost/prosemirror-markdown/browser", () => ({
+  convertProseMirrorToMarkdown: (doc: { __md?: string }) => doc?.__md ?? "",
 }));
 
 const notificationsShowMock = vi.fn();
@@ -53,10 +54,12 @@ import { useGeneratePageTitle } from "./use-generate-page-title.ts";
 
 // --- Test helpers -------------------------------------------------------------
 
-function makePageEditor(pageId: string, html = "<p>content</p>"): Editor {
+function makePageEditor(pageId: string, md = "content"): Editor {
   return {
     isDestroyed: false,
-    getHTML: () => html,
+    // The mocked convertProseMirrorToMarkdown reads `__md` back off this doc,
+    // so `md` is exactly the markdown the hook will send to the title service.
+    getJSON: () => ({ type: "doc", __md: md }),
     storage: { pageId },
   } as unknown as Editor;
 }

apps/client/src/features/editor/hooks/use-generate-page-title.ts

@@ -3,7 +3,7 @@ import { useMutation } from "@tanstack/react-query";
 import { useAtomValue } from "jotai";
 import { notifications } from "@mantine/notifications";
 import { useTranslation } from "react-i18next";
-import { htmlToMarkdown } from "@docmost/editor-ext";
+import { convertProseMirrorToMarkdown } from "@docmost/prosemirror-markdown/browser";
 import {
   pageEditorAtom,
   titleEditorAtom,
@@ -49,7 +49,9 @@ export function useGeneratePageTitle(pageId: string) {
     mutationFn: async () => {
       if (!pageEditor || pageEditor.isDestroyed) return;
 
-      const markdown = htmlToMarkdown(pageEditor.getHTML()).trim();
+      // Serialize the live editor content to markdown through the canonical
+      // converter (issue #347), matching the on-disk/export markdown form.
+      const markdown = convertProseMirrorToMarkdown(pageEditor.getJSON()).trim();
       if (!markdown) {
         notifications.show({ message: t("The note is empty"), color: "yellow" });
         return;

apps/client/src/features/page/components/header/page-header-menu.tsx

@@ -37,7 +37,7 @@ import { useTreeMutation } from "@/features/page/tree/hooks/use-tree-mutation.ts
 import { PageWidthToggle } from "@/features/user/components/page-width-pref.tsx";
 import { Trans, useTranslation } from "react-i18next";
 import ExportModal from "@/components/common/export-modal";
-import { htmlToMarkdown } from "@docmost/editor-ext";
+import { convertProseMirrorToMarkdown } from "@docmost/prosemirror-markdown/browser";
 import {
   pageEditorAtom,
   yjsConnectionStatusAtom,
@@ -199,8 +199,9 @@ function PageActionMenu({ readOnly }: PageActionMenuProps) {
 
   const handleCopyAsMarkdown = () => {
     if (!pageEditor) return;
-    const html = pageEditor.getHTML();
+    // Copy the page as canonical markdown through the shared converter (issue
-    const markdown = htmlToMarkdown(html);
+    // #347), so "Copy as markdown" matches the server export byte-for-byte.
+    const markdown = convertProseMirrorToMarkdown(pageEditor.getJSON());
     const title = page?.title ? `# ${page.title}\n\n` : "";
     clipboard.copy(`${title}${markdown}`);
     notifications.show({ message: t("Copied") });

apps/server/src/common/helpers/prosemirror/html-embed-import-detect.spec.ts

@@ -1,4 +1,5 @@
-import { markdownToHtml, encodeHtmlEmbedSource } from '@docmost/editor-ext';
+import { markdownToProseMirror } from '@docmost/prosemirror-markdown';
+import { encodeHtmlEmbedSource } from '@docmost/editor-ext';
 import { htmlToJson } from '../../../collaboration/collaboration.util';
 import { hasHtmlEmbedNode, stripHtmlEmbedNodes } from './html-embed.util';
 
@@ -10,13 +11,12 @@ import { hasHtmlEmbedNode, stripHtmlEmbedNodes } from './html-embed.util';
  *
  * The block renders inside a sandboxed iframe, so this is not an XSS surface;
  * this exercises the REAL server import conversion path that ImportService uses
- * (`markdownToHtml` then `htmlToJson`; `processHTML` adds only a cheerio
+ * (`markdownToProseMirror`, the canonical converter — issue #345/#347) and
- * link/iframe normalize pass which does not touch htmlEmbed divs) and asserts
+ * asserts that such a node is DETECTED and STRIPPABLE — so the share read path's
- * that such a node is DETECTED and STRIPPABLE — so the share read path's
  * master-toggle strip can remove it when the workspace toggle is OFF.
  */
 describe('htmlEmbed smuggled via the raw serialized div in imported markdown/HTML', () => {
-  it('round-trips through markdownToHtml -> htmlToJson and is DETECTED (base64 data-source)', async () => {
+  it('round-trips through markdownToProseMirror and is DETECTED (base64 data-source)', async () => {
     const source = '<script>steal()</script>';
     const encoded = encodeHtmlEmbedSource(source);
     const md = [
@@ -27,12 +27,9 @@ describe('htmlEmbed smuggled via the raw serialized div in imported markdown/HTM
       'World',
     ].join('\n');
 
-    const html = await markdownToHtml(md);
+    // The canonical importer parses the raw block-level div into a real
-    // marked preserves the raw block-level div verbatim.
+    // htmlEmbed node carrying the decoded source.
-    expect(html).toContain('data-type="htmlEmbed"');
+    const json = await markdownToProseMirror(md);
-
-    const json = htmlToJson(html);
-    // The div parses into a real htmlEmbed node carrying the decoded source.
     expect(hasHtmlEmbedNode(json)).toBe(true);
 
     // Because it is detected, the share master-toggle strip can remove it.
@@ -59,8 +56,7 @@ describe('htmlEmbed smuggled via the raw serialized div in imported markdown/HTM
     // therefore stripping) does not depend on the source being well-formed, so
     // the bypass cannot be hidden by sending a malformed data-source.
     const md = `<div data-type="htmlEmbed" data-source="&lt;script&gt;x&lt;/script&gt;"></div>`;
-    const html = await markdownToHtml(md);
+    const json = await markdownToProseMirror(md);
-    const json = htmlToJson(html);
     expect(hasHtmlEmbedNode(json)).toBe(true);
     expect(hasHtmlEmbedNode(stripHtmlEmbedNodes(json))).toBe(false);
   });

apps/server/src/integrations/import/utils/foreign-markdown.ts

@@ -238,8 +238,9 @@ function convertReferenceFootnotes(markdown: string): string {
  *
  * LINE-ANCHORED (the same shape the canonical parser uses in
  * prosemirror-markdown/page-file.ts): the block opens only on `---\n` at the
- * very start and closes only on a `\n---` line. The retired `markdownToHtml`
+ * very start and closes only on a `\n---` line. The retired editor-ext
- * strip closed on the FIRST `---` ANYWHERE (an unanchored close), so a value
+ * `markdownToHtml` front-matter strip (removed in #347) closed on the FIRST
+ * `---` ANYWHERE (an unanchored close), so a value
  * containing a triple-dash (e.g. `title: Q1 --- Q2`) truncated the front-matter
  * and leaked the rest into the body. An optional leading BOM is tolerated.
  */

apps/server/src/integrations/mcp/mcp.service.ts

@@ -34,8 +34,6 @@ import {
   isMetricsEnabled,
   observeMcpTool,
   incConnectTimeout,
-  incGetPageCacheHit,
-  incGetPageCacheMiss,
 } from '../metrics/metrics.registry';
 
 // Minimal shape of the embedded MCP HTTP handler exported by @docmost/mcp/http.
@@ -359,10 +357,6 @@ export class McpService implements OnModuleDestroy {
                       observeMcpTool(labels?.tool ?? 'other', value);
                     } else if (name === 'collab_connect_timeouts_total') {
                       incConnectTimeout();
-                    } else if (name === 'mcp_getpage_cache_hits_total') {
-                      incGetPageCacheHit();
-                    } else if (name === 'mcp_getpage_cache_misses_total') {
-                      incGetPageCacheMiss();
                     }
                   }
                 : undefined,

apps/server/src/integrations/metrics/metrics.constants.ts

@@ -25,15 +25,6 @@ export const METRIC_COLLAB_CONNECT_TIMEOUTS_TOTAL =
 export const METRIC_COLLAB_AUTH_DURATION = 'collab_auth_duration_seconds';
 export const METRIC_MCP_TOOL_DURATION = 'mcp_tool_duration_seconds';
 
-// #479 — getPage PM→Markdown conversion cache hit/miss counters. Emitted by the
-// MCP package via its dependency-neutral onMetric sink and routed onto these two
-// prom counters by the mcp.service onMetric callback; a >50% hit-rate is the
-// success signal for the getPage perf work. Same "do not rename" contract.
-export const METRIC_MCP_GETPAGE_CACHE_HITS_TOTAL =
-  'mcp_getpage_cache_hits_total';
-export const METRIC_MCP_GETPAGE_CACHE_MISSES_TOTAL =
-  'mcp_getpage_cache_misses_total';
-
 // Histogram buckets (seconds). Chosen to give useful p50/p95/p99 resolution
 // for typical web/DB latencies without exploding series cardinality.
 export const HTTP_BUCKETS = [

apps/server/src/integrations/metrics/metrics.registry.ts

@@ -24,8 +24,6 @@ import {
   METRIC_DB_QUERY_DURATION,
   METRIC_HTTP_REQUEST_DURATION,
   METRIC_MCP_TOOL_DURATION,
-  METRIC_MCP_GETPAGE_CACHE_HITS_TOTAL,
-  METRIC_MCP_GETPAGE_CACHE_MISSES_TOTAL,
   sizeBucket,
 } from './metrics.constants';
 
@@ -63,9 +61,6 @@ let connectTimeoutsCounter: Counter | null = null;
 let collabConnectHist: Histogram | null = null;
 let collabAuthHist: Histogram | null = null;
 let mcpToolHist: Histogram<'tool'> | null = null;
-// #479 — getPage conversion-cache hit/miss counters.
-let getPageCacheHitsCounter: Counter | null = null;
-let getPageCacheMissesCounter: Counter | null = null;
 
 // #402 — read-on-scrape source for collab_docs_open. The gauge is NEVER
 // inc/dec'd (that drifts under crashes/handoffs); instead its collect() callback
@@ -180,18 +175,6 @@ function init(): void {
     buckets: MCP_TOOL_BUCKETS,
     registers: [registry],
   });
-
-  getPageCacheHitsCounter = new Counter({
-    name: METRIC_MCP_GETPAGE_CACHE_HITS_TOTAL,
-    help: 'Total getPage PM→Markdown conversions served from the cache (skipped)',
-    registers: [registry],
-  });
-
-  getPageCacheMissesCounter = new Counter({
-    name: METRIC_MCP_GETPAGE_CACHE_MISSES_TOTAL,
-    help: 'Total getPage PM→Markdown conversions computed (cache misses)',
-    registers: [registry],
-  });
 }
 
 // Runs once when this module is first imported. Safe to call again (idempotent).
@@ -264,14 +247,6 @@ export function observeCollabAuth(seconds: number): void {
   collabAuthHist?.observe(seconds);
 }
 
-export function incGetPageCacheHit(): void {
-  getPageCacheHitsCounter?.inc();
-}
-
-export function incGetPageCacheMiss(): void {
-  getPageCacheMissesCounter?.inc();
-}
-
 export function observeMcpTool(tool: string, seconds: number): void {
   // `tool` MUST be a bounded, registration-derived MCP tool name (the caller
   // guarantees it comes from the registered-tool set) — never free-form input —

apps/server/src/integrations/metrics/metrics.spec.ts

@@ -10,8 +10,6 @@ import {
   incConnectTimeout,
   incDocLoad,
   incDocUnload,
-  incGetPageCacheHit,
-  incGetPageCacheMiss,
   isMetricsEnabled,
   observeCollabAuth,
   observeCollabConnect,
@@ -199,8 +197,6 @@ describe('metrics helpers are safe no-ops when METRICS_PORT is unset', () => {
       incDocLoad();
       incDocUnload();
       incConnectTimeout();
-      incGetPageCacheHit();
-      incGetPageCacheMiss();
       // Registering a source must not create the gauge or invoke the fn.
       registerDocsOpenSource(() => {
         throw new Error('docsOpenSource must NOT be called when disabled');

packages/editor-ext/package.json

@@ -11,9 +11,6 @@
   "main": "dist/index.js",
   "module": "./src/index.ts",
   "types": "dist/index.d.ts",
-  "dependencies": {
-    "marked": "17.0.5"
-  },
   "devDependencies": {
     "@vitest/coverage-v8": "4.1.6",
     "vitest": "4.1.6"

packages/editor-ext/src/index.ts

@@ -18,7 +18,6 @@ export * from "./lib/excalidraw";
 export * from "./lib/embed";
 export * from "./lib/html-embed/html-embed";
 export * from "./lib/mention";
-export * from "./lib/markdown";
 export * from "./lib/search-and-replace";
 export * from "./lib/embed-provider";
 export * from "./lib/subpages";

packages/editor-ext/src/lib/footnote/footnote-canonicalize.ts

@@ -14,7 +14,8 @@ import {
  * ProseMirror JSON directly (never running the editor's plugins), so the
  * canonical footnote topology was never enforced on those writes. The consumers
  * of this editor-ext copy are: the server markdown/HTML import
- * (`markdownToHtml -> htmlToJson` in import.service / file-import-task.service),
+ * (`markdownToProseMirror` from @docmost/prosemirror-markdown in import.service /
+ * file-import-task.service),
  * `PageService` create/update (`parseProsemirrorContent` for the JSON/markdown/
  * HTML REST write paths), and the client markdown PASTE path
  * (`markdown-clipboard.ts`). (The MCP package mirrors this canonicalizer in

packages/editor-ext/src/lib/footnote/footnote-markdown.test.ts

@@ -1,131 +0,0 @@
-import { describe, it, expect } from "vitest";
-import { htmlToMarkdown } from "../markdown/utils/turndown.utils";
-import { markdownToHtml } from "../markdown/utils/marked.utils";
-import { extractFootnoteDefinitions } from "../markdown/utils/footnote.marked";
-
-// HTML the editor-ext nodes render (sup[data-footnote-ref], section/div).
-const HTML =
-  `<p>Water<sup data-footnote-ref data-id="fn1"></sup> and clay<sup data-footnote-ref data-id="fn2"></sup>.</p>` +
-  `<section data-footnotes>` +
-  `<div data-footnote-def data-id="fn1"><p>First note.</p></div>` +
-  `<div data-footnote-def data-id="fn2"><p>Second note.</p></div>` +
-  `</section>`;
-
-describe("footnote markdown round-trip", () => {
-  it("HTML -> Markdown produces pandoc footnote syntax", () => {
-    const md = htmlToMarkdown(HTML);
-    expect(md).toContain("[^fn1]");
-    expect(md).toContain("[^fn2]");
-    expect(md).toContain("[^fn1]: First note.");
-    expect(md).toContain("[^fn2]: Second note.");
-  });
-
-  it("Markdown -> HTML rebuilds the footnote nodes' HTML", async () => {
-    const md = htmlToMarkdown(HTML);
-    const html = await markdownToHtml(md);
-    expect(html).toContain('data-footnote-ref data-id="fn1"');
-    expect(html).toContain('data-footnote-ref data-id="fn2"');
-    expect(html).toContain("data-footnotes");
-    expect(html).toContain('data-footnote-def data-id="fn1"');
-    expect(html).toContain("First note.");
-    expect(html).toContain("Second note.");
-  });
-
-  it("preserves a [^id]: line shown inside a fenced code block (not a definition)", async () => {
-    // A document that DOCUMENTS footnote syntax inside a code fence. The
-    // `[^demo]: ...` line is example text, not a real definition, and must
-    // survive the Markdown -> HTML conversion verbatim.
-    const md = [
-      "Here is how footnotes look:",
-      "",
-      "```markdown",
-      "Some text[^demo]",
-      "",
-      "[^demo]: this is the definition",
-      "```",
-      "",
-      "End of doc.",
-    ].join("\n");
-
-    const html = await markdownToHtml(md);
-    // The example definition line is kept inside the rendered code block.
-    expect(html).toContain("[^demo]: this is the definition");
-    // It did NOT get pulled out into a real footnotes section.
-    expect(html).not.toContain("data-footnotes");
-    expect(html).not.toContain("data-footnote-def");
-  });
-
-  it("extractFootnoteDefinitions keeps the FIRST duplicate definition and reuses markers", () => {
-    // Two definitions share id `d`, and the body has two `[^d]` markers. Under
-    // the import model (#166) duplicate definition ids are FIRST-WINS: only the
-    // first definition is kept; markers are NEVER rewritten, so the two `[^d]`
-    // references reuse the single footnote.
-    const md = [
-      "See here[^d] and there[^d].",
-      "",
-      "[^d]: first",
-      "[^d]: second",
-    ].join("\n");
-
-    const { body, section } = extractFootnoteDefinitions(md);
-
-    const defIds = Array.from(
-      section.matchAll(/data-footnote-def data-id="([^"]+)"/g),
-    ).map((m) => m[1]);
-    expect(defIds).toEqual(["d"]); // first-wins: one definition
-    expect(section).toContain("first");
-    expect(section).not.toContain("second"); // duplicate dropped
-
-    // Both markers stay `[^d]` (reuse) — no `d__2` minting.
-    const refIds = Array.from(body.matchAll(/\[\^([^\]\s]+)\]/g)).map(
-      (m) => m[1],
-    );
-    expect(refIds).toEqual(["d", "d"]);
-  });
-
-  it("extractFootnoteDefinitions is DETERMINISTIC and stable (same input -> same output)", () => {
-    // The output must be a pure function of the input markdown so importing the
-    // same source twice (or via the editor and the MCP mirror) is identical.
-    const md = [
-      "See[^d] one[^d] two[^d].",
-      "",
-      "[^d]: first",
-      "[^d]: second",
-      "[^d]: third",
-    ].join("\n");
-
-    const run = () => {
-      const { body, section } = extractFootnoteDefinitions(md);
-      const defIds = Array.from(
-        section.matchAll(/data-footnote-def data-id="([^"]+)"/g),
-      ).map((m) => m[1]);
-      const refIds = Array.from(body.matchAll(/\[\^([^\]\s]+)\]/g)).map(
-        (m) => m[1],
-      );
-      return { defIds, refIds };
-    };
-
-    const a = run();
-    const b = run();
-    expect(a).toEqual(b);
-    // First-wins: one kept definition `d`; all three reuse markers stay `d`.
-    expect(a.defIds).toEqual(["d"]);
-    expect(a.refIds).toEqual(["d", "d", "d"]);
-  });
-
-  it("markdownToHtml with a reused id renders ONE shared footnote def", async () => {
-    const md = [
-      "See here[^d] and there[^d].",
-      "",
-      "[^d]: first",
-      "[^d]: second",
-    ].join("\n");
-    const html = await markdownToHtml(md);
-    const defIds = Array.from(
-      html.matchAll(/data-footnote-def data-id="([^"]+)"/g),
-    ).map((m) => m[1]);
-    expect(defIds).toEqual(["d"]); // one shared definition
-    expect(html).toContain("first");
-    expect(html).not.toContain("second");
-  });
-});

packages/editor-ext/src/lib/footnote/footnote-sync.ts

@@ -103,8 +103,9 @@ interface CollisionPlan {
  * `X__2`, `X__3`, collision-bumped) so it survives as a distinct footnote — which,
  * having no matching reference, then falls under the normal orphan policy. It is
  * only ever dropped for lacking a reference, never for colliding. The IMPORT
- * paths (footnote.marked.ts / MCP extractFootnotes) instead apply first-wins +
+ * paths (@docmost/prosemirror-markdown / MCP extractFootnotes) instead apply
- * drop + warn for duplicate definitions; that divergence is intentional — import
+ * first-wins + drop + warn for duplicate definitions; that divergence is
+ * intentional — import
  * is an agent-authored artifact we sanitize, the editor is live user data we must
  * not lose.
  *

packages/editor-ext/src/lib/footnote/footnote-util.derive-id.test.ts

@@ -6,8 +6,9 @@ import { deriveFootnoteId } from "./footnote-util";
  *
  * `deriveFootnoteId` lives ONLY in editor-ext now — it is used by
  * `resolveCollisions` (re-id of a duplicate definition) and `footnotePastePlugin`
- * (re-id of a pasted colliding definition). The MCP/marked import paths no longer
+ * (re-id of a pasted colliding definition). The MCP / @docmost/prosemirror-markdown
- * derive ids (duplicate definitions there are first-wins-dropped, #166), so there
+ * import paths no longer derive ids (duplicate definitions there are
+ * first-wins-dropped, #166), so there
  * is no cross-package copy and no parity test to keep in sync. This table pins the
  * deterministic scheme so a future change to it is a conscious one.
  */

packages/editor-ext/src/lib/footnote/footnote-util.ts

@@ -63,8 +63,9 @@ export function generateFootnoteId(): string {
  * its own seen-set before requesting the next derived id.
  *
  * Used only inside editor-ext now (resolveCollisions for a re-id'd duplicate
- * DEFINITION, and footnotePastePlugin). The MCP/marked import paths no longer
+ * DEFINITION, and footnotePastePlugin). The MCP / @docmost/prosemirror-markdown
- * derive ids — duplicate definitions there are first-wins-dropped (#166) — so
+ * import paths no longer derive ids — duplicate definitions there are
+ * first-wins-dropped (#166) — so
  * there is no cross-package copy to keep in sync. The golden table in
  * footnote-util.derive-id.test.ts pins the scheme.
  */

packages/editor-ext/src/lib/image/image-markdown.test.ts

@@ -1,68 +0,0 @@
-import { describe, it, expect } from "vitest";
-import { generateJSON } from "@tiptap/html";
-import { Document } from "@tiptap/extension-document";
-import { Paragraph } from "@tiptap/extension-paragraph";
-import { Text } from "@tiptap/extension-text";
-import { htmlToMarkdown } from "../markdown/utils/turndown.utils";
-import { markdownToHtml } from "../markdown/utils/marked.utils";
-import { TiptapImage } from "./image";
-
-// Minimal schema for parsing markdownToHtml output back to JSON (mirrors
-// image.spec.ts), so we can assert the recovered caption EXACTLY.
-const parseExtensions = [Document, Paragraph, Text, TiptapImage];
-
-// Lossless markdown round-trip for image captions (issue #221). An image WITH a
-// caption can't be expressed as `![alt](src)`, so it is emitted as a raw <img>
-// (carrying data-caption) wrapped in a block <div>, the same trick the <video>
-// rule uses. marked passes the raw HTML through, so markdownToHtml keeps the
-// data-caption, and the image extension's parseHTML restores the attribute.
-describe("image caption markdown round-trip", () => {
-  it("HTML -> Markdown emits a raw <img data-caption> for captioned images", () => {
-    const html = `<p><img src="/files/a.png" alt="cat" data-caption="A grey cat"></p>`;
-    const md = htmlToMarkdown(html);
-    expect(md).toContain("data-caption=\"A grey cat\"");
-    expect(md).toContain('src="/files/a.png"');
-    expect(md).toContain('alt="cat"');
-    // It must NOT degrade to the lossy ![]() form.
-    expect(md).not.toContain("![cat]");
-  });
-
-  it("Markdown -> HTML restores data-caption on the <img>", async () => {
-    const html = `<p><img src="/files/a.png" alt="cat" data-caption="A grey cat"></p>`;
-    const md = htmlToMarkdown(html);
-    const back = await markdownToHtml(md);
-    expect(back).toContain('data-caption="A grey cat"');
-    expect(back).toContain('src="/files/a.png"');
-  });
-
-  it("special characters in the caption survive the round-trip (escaped)", async () => {
-    // The source caption is the decoded string `Tom & "Jerry"` (both an `&` and
-    // a `"`). escapeHtmlAttr must encode `&` -> `&amp;` and `"` -> `&quot;`.
-    const html = `<p><img src="/files/a.png" data-caption='Tom &amp; &quot;Jerry&quot;'></p>`;
-    const md = htmlToMarkdown(html);
-
-    // (a) The intermediate Markdown must carry the EXACT escaped attribute. This
-    // fails if escapeHtmlAttr stopped escaping `"` (attribute break-out:
-    // data-caption="Tom & "Jerry"") or double-encoded `&` (`&amp;amp;`).
-    expect(md).toContain('data-caption="Tom &amp; &quot;Jerry&quot;"');
-
-    const back = await markdownToHtml(md);
-    expect(back).toContain("data-caption=");
-    expect(back).toContain("Jerry");
-    expect(back).toContain("Tom");
-
-    // (b) Re-parse the rendered HTML through the image extension's parseHTML and
-    // assert the recovered caption is EXACTLY the original (no corruption, loss,
-    // or double-encoding).
-    const json = generateJSON(back, parseExtensions);
-    expect(json.content?.[0]?.attrs?.caption).toBe('Tom & "Jerry"');
-  });
-
-  it("caption-less images stay a clean ![alt](src) with no raw HTML", () => {
-    const html = `<p><img src="/files/a.png" alt="cat"></p>`;
-    const md = htmlToMarkdown(html);
-    expect(md).toContain("![cat](/files/a.png)");
-    expect(md).not.toContain("data-caption");
-    expect(md).not.toContain("<img");
-  });
-});

packages/editor-ext/src/lib/markdown/html-embed-marked.spec.ts

@@ -1,105 +0,0 @@
-import { describe, expect, it } from "vitest";
-import { htmlEmbedExtension } from "./utils/html-embed.marked";
-import { markdownToHtml } from "./index";
-import { encodeHtmlEmbedSource } from "../html-embed/html-embed";
-
-// CONTRACT tests for the marked block tokenizer that rebuilds an htmlEmbed node
-// from the `<!--html-embed:BASE64-->` marker (html-embed.marked.ts), plus the
-// observable round-trip through markdownToHtml.
-//
-// These pin the REAL tokenizer behaviour the import path depends on:
-//   - the tokenizer rule is anchored (^) and only accepts the base64 alphabet
-//     [A-Za-z0-9+/=], so a marker with non-base64 chars is NOT tokenized and
-//     survives as a literal HTML comment (not silently turned into something the
-//     server's strip no longer recognizes);
-//   - start() reports the correct index of the next marker so marked invokes the
-//     tokenizer at the right offset when a marker sits mid-document / after text;
-//   - a marker with surrounding text on the SAME line is split out into its own
-//     embed div while the surrounding text becomes ordinary paragraphs.
-//
-// The contract is asserted against the actual exported extension and pipeline —
-// no behaviour is invented; the expectations were read off the real tokenizer.
-
-const SAMPLE = "<b>x</b>";
-const ENC = encodeHtmlEmbedSource(SAMPLE);
-
-describe("htmlEmbed marked tokenizer — start()", () => {
-  it("returns the index of a marker that sits mid-document", () => {
-    const src = `hello world <!--html-embed:${ENC}-->`;
-    expect(htmlEmbedExtension.start(src)).toBe(src.indexOf("<!--html-embed:"));
-  });
-
-  it("returns 0 when the marker is at the very start", () => {
-    expect(htmlEmbedExtension.start(`<!--html-embed:${ENC}-->`)).toBe(0);
-  });
-
-  it("returns -1 when there is no marker", () => {
-    expect(htmlEmbedExtension.start("no marker here")).toBe(-1);
-  });
-});
-
-describe("htmlEmbed marked tokenizer — tokenizer()", () => {
-  it("tokenizes a marker at the start of the input, capturing the base64 payload", () => {
-    const token = htmlEmbedExtension.tokenizer(`<!--html-embed:${ENC}-->`);
-    expect(token).toBeTruthy();
-    expect(token!.type).toBe("htmlEmbed");
-    expect(token!.raw).toBe(`<!--html-embed:${ENC}-->`);
-    expect(token!.encoded).toBe(ENC);
-  });
-
-  it("tokenizes an EMPTY marker (the [A-Za-z0-9+/=]* class allows zero chars)", () => {
-    const token = htmlEmbedExtension.tokenizer("<!--html-embed:-->");
-    expect(token).toBeTruthy();
-    expect(token!.encoded).toBe("");
-    expect(token!.raw).toBe("<!--html-embed:-->");
-  });
-
-  it("does NOT tokenize when text precedes the marker (rule is anchored ^)", () => {
-    // marked relies on start() to advance to the marker; the tokenizer itself
-    // only matches at offset 0, so a non-anchored call returns undefined.
-    expect(
-      htmlEmbedExtension.tokenizer(`hello <!--html-embed:${ENC}-->`),
-    ).toBeUndefined();
-  });
-
-  it("does NOT tokenize a marker containing a non-base64 char ('$')", () => {
-    expect(
-      htmlEmbedExtension.tokenizer("<!--html-embed:ab$cd-->"),
-    ).toBeUndefined();
-  });
-
-  it("does NOT tokenize a marker containing a space", () => {
-    expect(
-      htmlEmbedExtension.tokenizer("<!--html-embed:ab cd-->"),
-    ).toBeUndefined();
-  });
-
-  it("renderer emits the embed div the node's parseHTML recognizes", () => {
-    const token = htmlEmbedExtension.tokenizer(`<!--html-embed:${ENC}-->`)!;
-    const html = htmlEmbedExtension.renderer(token as any);
-    expect(html).toBe(
-      `<div data-type="htmlEmbed" data-source="${ENC}"></div>`,
-    );
-  });
-});
-
-describe("htmlEmbed marked tokenizer — markdownToHtml round-trip", () => {
-  it("splits a marker out of surrounding same-line text into its own embed div", async () => {
-    const html = await markdownToHtml(`before <!--html-embed:${ENC}--> after`);
-    // The marker became the embed div...
-    expect(html).toContain(
-      `<div data-type="htmlEmbed" data-source="${ENC}"></div>`,
-    );
-    // ...and the surrounding text survived as ordinary paragraph content.
-    expect(html).toContain("before");
-    expect(html).toContain("after");
-  });
-
-  it("leaves a marker with non-base64 chars as a literal comment (NOT an embed div)", async () => {
-    const html = await markdownToHtml("<!--html-embed:ab$cd-->");
-    // It is NOT tokenized into an embed div the server would strip...
-    expect(html).not.toContain('data-type="htmlEmbed"');
-    // ...it passes through unchanged as a literal HTML comment.
-    expect(html).toContain("<!--html-embed:ab$cd-->");
-  });
-});

packages/editor-ext/src/lib/markdown/index.ts

@@ -1,2 +0,0 @@
-export * from "./utils/marked.utils";
-export * from "./utils/turndown.utils";

packages/editor-ext/src/lib/markdown/markdown-html-embed.spec.ts

@@ -1,112 +0,0 @@
-import { describe, it, expect } from "vitest";
-import { markdownToHtml, htmlToMarkdown } from "./index";
-import {
-  encodeHtmlEmbedSource,
-  decodeHtmlEmbedSource,
-} from "../html-embed/html-embed";
-
-// SECURITY (Variant C admin gate, import attack surface).
-//
-// The markdown import path is the only write path where an htmlEmbed reaches
-// the server purely from file bytes (no editor / collab socket). The marked
-// tokenizer in `html-embed.marked.ts` and the turndown rule in
-// `turndown.utils.ts` are what materialize the `<!--html-embed:BASE64-->`
-// marker into the `<div data-type="htmlEmbed" data-source="BASE64">` element
-// that the server then parses into an htmlEmbed node and the admin gate strips.
-//
-// If either the tokenizer regex or the turndown rule shape drifts, the marker
-// would either (a) stop becoming an htmlEmbed node (silently dropping admin
-// content) or (b) become some OTHER tag the server's `hasHtmlEmbedNode` no
-// longer recognizes (a strip bypass). These tests pin the marker <-> embed-div
-// contract that the server-side strip relies on. editor-ext had ZERO tests
-// before this file; this adds the runner + the round-trip coverage.
-
-// The server parses the embed div by matching `data-type="htmlEmbed"` and
-// decoding `data-source`; mirror that here so the assertion is exactly what the
-// real `htmlToJson` -> htmlEmbed node parse depends on (the node's parseHTML in
-// html-embed.ts uses the same selector + decodeHtmlEmbedSource).
-const EMBED_DIV_RE = /<div[^>]*\bdata-type="htmlEmbed"[^>]*>/;
-function extractEmbedSource(html: string): string | undefined {
-  const div = EMBED_DIV_RE.exec(html);
-  if (!div) return undefined;
-  const enc = /data-source="([^"]*)"/.exec(div[0]);
-  if (!enc) return undefined;
-  return decodeHtmlEmbedSource(enc[1]);
-}
-
-// Replicates the server's `hasHtmlEmbedNode` decision against the embed *div*
-// (the HTML form the server immediately converts to JSON). If this matches, the
-// server's JSON-level `hasHtmlEmbedNode` will too, because htmlToJson maps this
-// exact div to an htmlEmbed node.
-function htmlHasHtmlEmbed(html: string): boolean {
-  return EMBED_DIV_RE.test(html);
-}
-
-describe("markdown <!--html-embed--> import round-trip", () => {
-  const source = "<script>x</script>";
-
-  it("markdownToHtml turns the marker into an htmlEmbed div carrying the source", async () => {
-    const md = "<!--html-embed:" + encodeHtmlEmbedSource(source) + "-->";
-    const html = await markdownToHtml(md);
-
-    // The marker became the embed div the server recognizes as an htmlEmbed
-    // node (so the server's hasHtmlEmbedNode would match it after htmlToJson).
-    expect(htmlHasHtmlEmbed(html)).toBe(true);
-    // The decoded source is the original script, intact.
-    expect(extractEmbedSource(html)).toBe(source);
-    // The raw script is NOT inlined into the HTML — it stays base64 in the
-    // attribute (the marker itself must not be a direct injection vector).
-    expect(html).not.toContain("<script>x</script>");
-  });
-
-  it("preserves UTF-8 / special chars in the embedded source", async () => {
-    const utf8 = '<script>console.log("héllo → 世界")</script>';
-    const md = "<!--html-embed:" + encodeHtmlEmbedSource(utf8) + "-->";
-    const html = await markdownToHtml(md);
-    expect(htmlHasHtmlEmbed(html)).toBe(true);
-    expect(extractEmbedSource(html)).toBe(utf8);
-  });
-
-  it("an empty marker still produces an htmlEmbed div (empty source)", async () => {
-    const html = await markdownToHtml("<!--html-embed:-->");
-    expect(htmlHasHtmlEmbed(html)).toBe(true);
-    expect(extractEmbedSource(html)).toBe("");
-  });
-
-  it("round-trips htmlToMarkdown -> markdownToHtml preserving the embed marker", async () => {
-    const encoded = encodeHtmlEmbedSource(source);
-    // NOTE: turndown drops a *blank* (childless) element before any custom rule
-    // runs, and the htmlEmbed div is normally childless. The export pipeline
-    // therefore must give the rule a non-blank div to fire on; we add an inert
-    // text child here to exercise the real turndown htmlEmbed rule. (A blank
-    // embed div serializing to "" is asserted separately below as a documented
-    // edge so this contract drift is visible.)
-    const startHtml = `<div data-type="htmlEmbed" data-source="${encoded}">x</div>`;
-
-    // Export to markdown: the turndown rule emits the <!--html-embed:..-->
-    // marker (lossless, inert in plain markdown viewers).
-    const md = htmlToMarkdown(startHtml);
-    expect(md).toContain("<!--html-embed:" + encoded + "-->");
-
-    // Re-import: the marker round-trips back into an embed div with the same
-    // decoded source — this is the marker <-> embed-div contract the server's
-    // import strip depends on.
-    const html = await markdownToHtml(md);
-    expect(htmlHasHtmlEmbed(html)).toBe(true);
-    expect(extractEmbedSource(html)).toBe(source);
-  });
-
-  it("documents that a BLANK embed div serializes to empty markdown (turndown drops childless blocks)", () => {
-    const encoded = encodeHtmlEmbedSource(source);
-    const blank = `<div data-type="htmlEmbed" data-source="${encoded}"></div>`;
-    // This pins current behavior so a future change to the turndown rule (e.g.
-    // making it fire on blank nodes) is caught rather than silently shipping.
-    expect(htmlToMarkdown(blank)).toBe("");
-  });
-
-  it("the base64 codec itself round-trips (no '<' leaks into the attribute)", () => {
-    const encoded = encodeHtmlEmbedSource(source);
-    expect(encoded).not.toContain("<");
-    expect(decodeHtmlEmbedSource(encoded)).toBe(source);
-  });
-});

packages/editor-ext/src/lib/markdown/utils/basename.ts

@@ -1,29 +0,0 @@
-/**
- * Flexible `basename` implementation for node and the browser
- * @see https://stackoverflow.com/a/59907288/2228771
- */
-export function getBasename(path: string) {
-  // make sure the basename is not empty, if string ends with separator
-  let end = path.length - 1;
-  while (path[end] === '/' || path[end] === '\\') {
-    --end;
-  }
-
-  // support mixing of Win + Unix path separators
-  const i1 = path.lastIndexOf('/', end);
-  const i2 = path.lastIndexOf('\\', end);
-
-  let start: number;
-  if (i1 === -1) {
-    if (i2 === -1) {
-      // no separator in the whole thing
-      return path;
-    }
-    start = i2;
-  } else if (i2 === -1) {
-    start = i1;
-  } else {
-    start = Math.max(i1, i2);
-  }
-  return path.substring(start + 1, end + 1);
-}

packages/editor-ext/src/lib/markdown/utils/callout-common.marked.ts

@@ -1,33 +0,0 @@
-/**
- * Shared pieces for the two callout tokenizers — `callout.marked.ts` (the
- * `:::type` fenced form) and `github-callout.marked.ts` (the `> [!type]` GitHub
- * alert form). Both emit the SAME callout node, so the banner type dictionary
- * and the HTML renderer live here once instead of drifting apart in two files.
- * The tokenizers themselves stay separate (different syntaxes / source matching).
- */
-
-/** The four callout banner types the editor schema supports. */
-export const CALLOUT_TYPES = ['info', 'success', 'warning', 'danger'] as const;
-
-export type CalloutType = (typeof CALLOUT_TYPES)[number];
-
-/**
- * Coerce an arbitrary type name onto a supported banner type, defaulting to
- * `info` for anything unrecognized (the shared fallback both tokenizers use).
- */
-export function normalizeCalloutType(type: string): CalloutType {
-  return (CALLOUT_TYPES as readonly string[]).includes(type)
-    ? (type as CalloutType)
-    : 'info';
-}
-
-/**
- * Render a callout node to the editor's HTML shape. `body` is the already
- * markdown-parsed inner content (marked may hand back a string synchronously).
- */
-export function renderCalloutHtml(
-  type: string,
-  body: string | Promise<string>,
-): string {
-  return `<div data-type="callout" data-callout-type="${type}">${body}</div>`;
-}

packages/editor-ext/src/lib/markdown/utils/callout.marked.ts

@@ -1,37 +0,0 @@
-import { Token, marked } from 'marked';
-import { normalizeCalloutType, renderCalloutHtml } from './callout-common.marked';
-
-interface CalloutToken {
-  type: 'callout';
-  calloutType: string;
-  text: string;
-  raw: string;
-}
-
-export const calloutExtension = {
-  name: 'callout',
-  level: 'block',
-  start(src: string) {
-    return src.match(/:::/)?.index ?? -1;
-  },
-  tokenizer(src: string): CalloutToken | undefined {
-    const rule = /^:::([a-zA-Z0-9]+)\s+([\s\S]+?):::/;
-    const match = rule.exec(src);
-
-    if (match) {
-      return {
-        type: 'callout',
-        calloutType: normalizeCalloutType(match[1]),
-        raw: match[0],
-        text: match[2].trim(),
-      };
-    }
-  },
-  renderer(token: Token) {
-    const calloutToken = token as CalloutToken;
-    return renderCalloutHtml(
-      calloutToken.calloutType,
-      marked.parse(calloutToken.text),
-    );
-  },
-};

packages/editor-ext/src/lib/markdown/utils/footnote.marked.orphan.test.ts

@@ -1,72 +0,0 @@
-import { describe, it, expect } from "vitest";
-import { extractFootnoteDefinitions } from "./footnote.marked";
-
-/** Pull the ordered list of `data-footnote-def` ids out of the rendered section. */
-function defIds(section: string): string[] {
-  return [...section.matchAll(/data-footnote-def data-id="([^"]+)"/g)].map(
-    (m) => m[1],
-  );
-}
-
-/** Pull the ordered list of `[^id]` markers that remain in the body. */
-function bodyMarkers(body: string): string[] {
-  return [...body.matchAll(/\[\^([^\]\s]+)\]/g)].map((m) => m[1]);
-}
-
-describe("extractFootnoteDefinitions: duplicate definition ids (first-wins)", () => {
-  // Body has ONE `[^d]` reference but THREE `[^d]:` definitions. Under the
-  // import model (#166) a duplicate definition id is FIRST-WINS: only the first
-  // definition is kept; the rest are DROPPED (and surfaced by analyzeFootnotes,
-  // not silently re-id'd into orphan footnotes as before). Reference markers are
-  // never rewritten, so repeated references would reuse the single footnote.
-  const md = ["See[^d].", "", "[^d]: a", "[^d]: b", "[^d]: c"].join("\n");
-
-  it("keeps only the FIRST definition for the id (first-wins)", () => {
-    const { section } = extractFootnoteDefinitions(md);
-    const ids = defIds(section);
-    expect(ids).toEqual(["d"]);
-  });
-
-  it("keeps the first definition's text and drops the duplicates", () => {
-    const { section } = extractFootnoteDefinitions(md);
-    expect(section).toContain('data-footnote-def data-id="d"><p>a</p>');
-    // No derived `d__2` / `d__3` ids are emitted anymore.
-    expect(section).not.toContain("d__2");
-    expect(section).not.toContain("d__3");
-    // The dropped duplicate texts are not in the section.
-    expect(section).not.toContain("<p>b</p>");
-    expect(section).not.toContain("<p>c</p>");
-  });
-
-  it("leaves the SINGLE body marker as [^d] (markers are never rewritten)", () => {
-    const { body } = extractFootnoteDefinitions(md);
-    expect(bodyMarkers(body)).toEqual(["d"]);
-    expect(body).toContain("See[^d].");
-    // The definition lines themselves were pulled OUT of the body.
-    expect(body).not.toContain("[^d]: a");
-    expect(body).not.toContain("[^d]: b");
-    expect(body).not.toContain("[^d]: c");
-  });
-
-  it("does not crash and produces a well-formed footnotes section", () => {
-    const { section } = extractFootnoteDefinitions(md);
-    expect(section.startsWith("<section data-footnotes>")).toBe(true);
-    expect(section.endsWith("</section>")).toBe(true);
-    // Exactly one definition div (first-wins).
-    expect([...section.matchAll(/<div data-footnote-def/g)]).toHaveLength(1);
-  });
-});
-
-describe("extractFootnoteDefinitions: reuse (repeated references, one definition)", () => {
-  // Pandoc semantics: many `[^a]` references + one `[^a]:` definition = one
-  // footnote, shared. Markers are left intact so the editor numbers them as one.
-  const md = ["A[^a] B[^a] C[^a].", "", "[^a]: shared note"].join("\n");
-
-  it("emits exactly one definition and leaves every reference marker as [^a]", () => {
-    const { section, body } = extractFootnoteDefinitions(md);
-    expect(defIds(section)).toEqual(["a"]);
-    expect(section).toContain('data-footnote-def data-id="a"><p>shared note</p>');
-    // All three reference markers stay `a` (no `a__2`/`a__3` minting).
-    expect(bodyMarkers(body)).toEqual(["a", "a", "a"]);
-  });
-});

packages/editor-ext/src/lib/markdown/utils/footnote.marked.ts

@@ -1,131 +0,0 @@
-import { marked } from "marked";
-
-/**
- * Pandoc/GFM footnote support for the marked (Markdown -> HTML) pipeline.
- *
- * Two pieces:
- *  - an INLINE tokenizer for `[^id]` references -> <sup data-footnote-ref
- *    data-id="id"> (matches the editor-ext FootnoteReference renderHTML);
- *  - a document hook (`preprocess`/`walkTokens` is awkward for collecting +
- *    removing definitions, so we use a regex preprocessing step instead) that
- *    pulls every `[^id]: text` definition line out of the body and appends a
- *    single <section data-footnotes> with one <div data-footnote-def> per
- *    definition, so the round-trip rebuilds footnotesList + footnoteDefinition.
- *
- * Every FIRST definition line is emitted — duplicate ids are first-wins (the
- * rest are dropped, and surfaced via analyzeFootnotes), and reference markers are
- * left untouched so repeated `[^a]` references reuse the one footnote (#166).
- * Orphan definitions (no matching reference) are still emitted here; the editor's
- * sync plugin reconciles the final reference/definition set (drops orphans,
- * synthesizes a single empty definition for a reference that lacks one).
- */
-
-const DEFINITION_RE = /^\[\^([^\]\s]+)\]:[ \t]*(.*)$/;
-const REFERENCE_RE = /\[\^([^\]\s]+)\]/;
-
-interface FootnoteRefToken {
-  type: "footnoteRef";
-  raw: string;
-  id: string;
-}
-
-export const footnoteReferenceExtension = {
-  name: "footnoteRef",
-  level: "inline" as const,
-  start(src: string) {
-    return src.match(/\[\^/)?.index ?? -1;
-  },
-  tokenizer(src: string): FootnoteRefToken | undefined {
-    const match = REFERENCE_RE.exec(src);
-    // Only match at the very start of the remaining inline source.
-    if (match && match.index === 0) {
-      return {
-        type: "footnoteRef",
-        raw: match[0],
-        id: match[1],
-      };
-    }
-    return undefined;
-  },
-  renderer(token: FootnoteRefToken) {
-    return `<sup data-footnote-ref data-id="${escapeAttr(token.id)}"></sup>`;
-  },
-};
-
-function escapeAttr(value: string): string {
-  return String(value).replace(/&/g, "&amp;").replace(/"/g, "&quot;");
-}
-
-/**
- * Extract `[^id]: text` definition lines from the markdown body, returning the
- * cleaned body plus a rendered <section data-footnotes> (empty string when no
- * definitions). Call this BEFORE marked.parse and append the section to the
- * resulting HTML.
- */
-export function extractFootnoteDefinitions(markdown: string): {
-  body: string;
-  section: string;
-} {
-  const lines = markdown.split("\n");
-  const bodyLines: string[] = [];
-  const definitions: Array<{ id: string; text: string }> = [];
-
-  // Track fenced-code state so a `[^id]: ...` line that merely SHOWS footnote
-  // syntax inside a ``` / ~~~ code block is left in the body verbatim and not
-  // mistaken for a real definition.
-  let fence: string | null = null;
-
-  for (const line of lines) {
-    const fenceMatch = /^(\s*)(`{3,}|~{3,})/.exec(line);
-    if (fenceMatch) {
-      const marker = fenceMatch[2][0];
-      if (fence === null) {
-        fence = marker; // opening fence
-      } else if (marker === fence) {
-        fence = null; // closing fence (matching delimiter type)
-      }
-      bodyLines.push(line);
-      continue;
-    }
-
-    const m = fence === null ? DEFINITION_RE.exec(line) : null;
-    if (m) {
-      definitions.push({ id: m[1], text: m[2] });
-    } else {
-      bodyLines.push(line);
-    }
-  }
-
-  if (definitions.length === 0) {
-    return { body: markdown, section: "" };
-  }
-
-  // Duplicate definition ids (e.g. `[^d]: first` / `[^d]: second`): FIRST WINS,
-  // the rest are DROPPED. Reference markers are left UNTOUCHED so repeated `[^a]`
-  // references reuse the single footnote (Pandoc semantics, #166). This differs
-  // from the live editor's never-lose policy (resolveCollisions re-ids a
-  // duplicate definition into an orphan) on purpose: an import is an
-  // agent-authored artifact we sanitize, and the dropped duplicate is surfaced
-  // to the caller via analyzeFootnotes' `duplicateDefinitions` warning instead.
-  const firstById = new Map<string, string>(); // id -> first definition text
-  for (const def of definitions) {
-    if (!firstById.has(def.id)) firstById.set(def.id, def.text);
-  }
-
-  const defsHtml = [...firstById.entries()]
-    .map(([id, text]) => {
-      // Render the definition text as inline markdown so emphasis/links inside
-      // a footnote survive the round-trip; wrap in a paragraph (the node's
-      // content is paragraph+).
-      const inner = marked.parseInline(text || "");
-      return `<div data-footnote-def data-id="${escapeAttr(
-        id,
-      )}"><p>${inner}</p></div>`;
-    })
-    .join("");
-
-  return {
-    body: bodyLines.join("\n"),
-    section: `<section data-footnotes>${defsHtml}</section>`,
-  };
-}

packages/editor-ext/src/lib/markdown/utils/github-callout.marked.test.ts

@@ -1,54 +0,0 @@
-import { describe, it, expect } from "vitest";
-import { markdownToHtml } from "./marked.utils";
-
-/**
- * Regression for issue #192: pasting a GitHub-style `> [!type]` alert produced a
- * literal `<blockquote>` containing `[!info]` instead of a callout node, because
- * only the `:::type` form was tokenized. The editor paste path runs the same
- * `markdownToHtml`, so these assertions pin the conversion at the source.
- */
-function html(md: string): string {
-  const out = markdownToHtml(md);
-  if (typeof out !== "string") throw new Error("expected sync string output");
-  return out;
-}
-
-describe("markdownToHtml: GitHub `> [!type]` callouts", () => {
-  it("converts `> [!info]` to a callout node, not a literal blockquote", () => {
-    const out = html("> [!info]\n> Callout body text here");
-    expect(out).toContain('data-type="callout"');
-    expect(out).toContain('data-callout-type="info"');
-    expect(out).toContain("Callout body text here");
-    expect(out).not.toContain("[!info]");
-    expect(out).not.toContain("<blockquote");
-  });
-
-  it("maps GitHub alert aliases onto the supported banner types", () => {
-    expect(html("> [!NOTE]\n> x")).toContain('data-callout-type="info"');
-    expect(html("> [!TIP]\n> x")).toContain('data-callout-type="success"');
-    expect(html("> [!WARNING]\n> x")).toContain('data-callout-type="warning"');
-    expect(html("> [!CAUTION]\n> x")).toContain('data-callout-type="danger"');
-  });
-
-  it("accepts the editor's own type names directly", () => {
-    expect(html("> [!success]\n> x")).toContain('data-callout-type="success"');
-    expect(html("> [!danger]\n> x")).toContain('data-callout-type="danger"');
-  });
-
-  it("falls back to info for an unknown type", () => {
-    expect(html("> [!bogus]\n> x")).toContain('data-callout-type="info"');
-  });
-
-  it("preserves multi-line callout bodies", () => {
-    const out = html("> [!warning]\n> line one\n> line two");
-    expect(out).toContain('data-callout-type="warning"');
-    expect(out).toContain("line one");
-    expect(out).toContain("line two");
-  });
-
-  it("still converts the `:::type` form", () => {
-    const out = html(":::info\nbody\n:::");
-    expect(out).toContain('data-type="callout"');
-    expect(out).toContain('data-callout-type="info"');
-  });
-});

packages/editor-ext/src/lib/markdown/utils/github-callout.marked.ts

@@ -1,81 +0,0 @@
-import { Token, marked } from 'marked';
-import { renderCalloutHtml } from './callout-common.marked';
-
-interface GithubCalloutToken {
-  type: 'githubCallout';
-  calloutType: string;
-  text: string;
-  raw: string;
-}
-
-/**
- * Map GitHub "alert" blockquote markers (`> [!NOTE]`, `> [!WARNING]`, …) onto
- * the four callout banner types the editor schema supports. The editor's own
- * type names (`info`/`success`/`warning`/`danger`) are also accepted directly,
- * because users paste both forms. Anything unrecognized falls back to `info`,
- * matching the `:::type` callout tokenizer.
- */
-const GITHUB_ALERT_TYPE_MAP: Record<string, string> = {
-  note: 'info',
-  tip: 'success',
-  important: 'info',
-  warning: 'warning',
-  caution: 'danger',
-  info: 'info',
-  success: 'success',
-  danger: 'danger',
-};
-
-/**
- * Tokenizer for GitHub-flavored alert callouts written as a blockquote whose
- * first line is `[!type]`:
- *
- *   > [!info]
- *   > body line one
- *   > body line two
- *
- * Without this, the default blockquote tokenizer wins and the marker renders as
- * a literal `[!info]` inside a `<blockquote>`. The editor's paste path runs the
- * same `markdownToHtml`, so registering this here also fixes pasting the syntax
- * into the editor (issue #192), not just markdown import.
- */
-export const githubCalloutExtension = {
-  name: 'githubCallout',
-  level: 'block' as const,
-  start(src: string) {
-    return src.match(/^ {0,3}>[ \t]*\[!/m)?.index ?? -1;
-  },
-  tokenizer(src: string): GithubCalloutToken | undefined {
-    const rule =
-      /^ {0,3}>[ \t]*\[!([a-zA-Z]+)\][^\n]*(?:\n {0,3}>[^\n]*)*(?:\n|$)/;
-    const match = rule.exec(src);
-    if (!match) return undefined;
-
-    const rawType = match[1].toLowerCase();
-    const calloutType = GITHUB_ALERT_TYPE_MAP[rawType] ?? 'info';
-
-    const text = match[0]
-      .replace(/\n+$/, '')
-      .split('\n')
-      // Strip the blockquote marker (`>` + optional space) from every line.
-      .map((line) => line.replace(/^ {0,3}>[ \t]?/, ''))
-      // Drop the `[!type]` marker that opens the first line.
-      .map((line, i) => (i === 0 ? line.replace(/^\[![a-zA-Z]+\][ \t]*/, '') : line))
-      .join('\n')
-      .trim();
-
-    return {
-      type: 'githubCallout',
-      calloutType,
-      raw: match[0],
-      text,
-    };
-  },
-  renderer(token: Token) {
-    const calloutToken = token as GithubCalloutToken;
-    return renderCalloutHtml(
-      calloutToken.calloutType,
-      marked.parse(calloutToken.text),
-    );
-  },
-};

packages/editor-ext/src/lib/markdown/utils/html-embed.marked.ts

@@ -1,41 +0,0 @@
-import { Token } from "marked";
-
-interface HtmlEmbedToken {
-  type: "htmlEmbed";
-  raw: string;
-  encoded: string;
-}
-
-/**
- * Marked extension that rebuilds an `htmlEmbed` node from the HTML comment
- * marker produced by the turndown rule (`<!--html-embed:<base64>-->`).
- *
- * It emits the same marker div the node's `parseHTML` recognizes, so the
- * pipeline MD -> HTML -> ProseMirror JSON restores the node (and its
- * base64 `data-source`) exactly. We do NOT expand the raw markup here; the
- * source stays base64-encoded in the attribute and is only executed by the
- * client NodeView.
- */
-export const htmlEmbedExtension = {
-  name: "htmlEmbed",
-  level: "block" as const,
-  start(src: string) {
-    return src.indexOf("<!--html-embed:");
-  },
-  tokenizer(src: string): HtmlEmbedToken | undefined {
-    const rule = /^<!--html-embed:([A-Za-z0-9+/=]*)-->/;
-    const match = rule.exec(src);
-
-    if (match) {
-      return {
-        type: "htmlEmbed",
-        raw: match[0],
-        encoded: match[1] ?? "",
-      };
-    }
-  },
-  renderer(token: Token) {
-    const htmlEmbedToken = token as HtmlEmbedToken;
-    return `<div data-type="htmlEmbed" data-source="${htmlEmbedToken.encoded}"></div>`;
-  },
-};

packages/editor-ext/src/lib/markdown/utils/marked.utils.ts

@@ -1,76 +0,0 @@
-import { marked } from "marked";
-import { calloutExtension } from "./callout.marked";
-import { githubCalloutExtension } from "./github-callout.marked";
-import { mathBlockExtension } from "./math-block.marked";
-import { mathInlineExtension } from "./math-inline.marked";
-import {
-  footnoteReferenceExtension,
-  extractFootnoteDefinitions,
-} from "./footnote.marked";
-import { htmlEmbedExtension } from "./html-embed.marked";
-
-marked.use({
-  renderer: {
-    list({ ordered, start, items }) {
-      let body = "";
-      for (const item of items) {
-        body += this.listitem(item);
-      }
-
-      if (ordered) {
-        const startAttr = start !== 1 ? ` start="${start}"` : "";
-        return `<ol${startAttr}>\n${body}</ol>\n`;
-      }
-
-      const isTaskList = items.some((item) => item.task);
-      const dataType = isTaskList ? ' data-type="taskList"' : "";
-      return `<ul${dataType}>\n${body}</ul>\n`;
-    },
-    listitem({ tokens, task: isTask, checked: isChecked }) {
-      const text = this.parser.parse(tokens);
-      if (!isTask) {
-        return `<li>${text}</li>\n`;
-      }
-      const checkedAttr = isChecked
-        ? 'data-checked="true"'
-        : 'data-checked="false"';
-      return `<li data-type="taskItem" ${checkedAttr}>${text}</li>\n`;
-    },
-  },
-});
-
-marked.use({
-  extensions: [
-    calloutExtension,
-    githubCalloutExtension,
-    mathBlockExtension,
-    mathInlineExtension,
-    footnoteReferenceExtension,
-    htmlEmbedExtension,
-  ],
-});
-
-marked.setOptions({ breaks: true });
-
-export function markdownToHtml(
-  markdownInput: string,
-): string | Promise<string> {
-  const YAML_FONT_MATTER_REGEX = /^\s*---[\s\S]*?---\s*/;
-
-  const markdown = markdownInput
-    .replace(YAML_FONT_MATTER_REGEX, "")
-    .trimStart();
-
-  // Pull `[^id]: ...` definition lines out of the body, render the body, then
-  // append a single <section data-footnotes> so the round-trip rebuilds the
-  // footnotesList + footnoteDefinition nodes.
-  const { body, section } = extractFootnoteDefinitions(markdown);
-
-  const parsed = marked.parse(body);
-  if (!section) return parsed;
-
-  if (typeof parsed === "string") {
-    return parsed + section;
-  }
-  return parsed.then((html) => html + section);
-}

packages/editor-ext/src/lib/markdown/utils/math-block.marked.ts

@@ -1,37 +0,0 @@
-import { Token, marked } from 'marked';
-
-interface MathBlockToken {
-  type: 'mathBlock';
-  text: string;
-  raw: string;
-}
-
-export const mathBlockExtension = {
-  name: 'mathBlock',
-  level: 'block',
-  start(src: string) {
-    return src.match(/\$\$/)?.index ?? -1;
-  },
-  tokenizer(src: string): MathBlockToken | undefined {
-    const rule = /^\$\$(?!(\$))([\s\S]+?)\$\$/;
-    const match = rule.exec(src);
-
-    if (match) {
-      return {
-        type: 'mathBlock',
-        raw: match[0],
-        text: match[2]?.trim(),
-      };
-    }
-  },
-  renderer(token: Token) {
-    const mathBlockToken = token as MathBlockToken;
-    // parse to prevent escaping slashes
-    const latex = marked
-      .parse(mathBlockToken.text)
-      .toString()
-      .replace(/<(\/)?p>/g, '');
-
-    return `<div data-type="${mathBlockToken.type}" data-katex="true">${latex}</div>`;
-  },
-};

packages/editor-ext/src/lib/markdown/utils/math-inline.marked.falsepositive.test.ts

@@ -1,50 +0,0 @@
-import { describe, it, expect } from "vitest";
-import { markdownToHtml } from "./marked.utils";
-
-/**
- * Data-integrity regression (issue #204, Phase 2): plain prose that mentions
- * prices like `$5 and $6` must NOT be misread as inline math. The inline-math
- * tokenizer mutates a global `marked` singleton at import time
- * (`marked.utils.ts`), so math behaviour can only be exercised safely through
- * the public `markdownToHtml`; importing the tokenizer in isolation would give
- * a different, non-representative result. These assertions therefore drive the
- * real conversion path.
- */
-function html(md: string): string {
-  const out = markdownToHtml(md);
-  if (typeof out !== "string") throw new Error("expected sync string output");
-  return out;
-}
-
-const MATH_MARKERS = ['data-type="mathInline"', 'data-katex="true"'];
-
-function hasInlineMath(out: string): boolean {
-  return MATH_MARKERS.some((m) => out.includes(m));
-}
-
-describe("markdownToHtml: inline-math false positives", () => {
-  it("does not treat prices `$5 and $6` as inline math", () => {
-    const out = html("It costs $5 and $6 today.");
-    expect(hasInlineMath(out)).toBe(false);
-    // The text survives verbatim (no katex span swallowing it).
-    expect(out).toContain("$5 and $6");
-  });
-
-  it("does not treat a single trailing price `$5` as inline math", () => {
-    const out = html("Lunch was $5.");
-    expect(hasInlineMath(out)).toBe(false);
-    expect(out).toContain("$5");
-  });
-
-  it("does not treat `$5, $6, $7` (multiple prices) as inline math", () => {
-    const out = html("Choose $5, $6, $7 plans.");
-    expect(hasInlineMath(out)).toBe(false);
-  });
-
-  it("STILL converts a genuine inline-math expression `$x + y$`", () => {
-    // Guard the positive path so the false-positive guard above can't be
-    // satisfied by simply disabling math entirely.
-    const out = html("The sum $x + y$ is shown.");
-    expect(hasInlineMath(out)).toBe(true);
-  });
-});

packages/editor-ext/src/lib/markdown/utils/math-inline.marked.ts

@@ -1,55 +0,0 @@
-import { Token, marked } from 'marked';
-
-interface MathInlineToken {
-  type: 'mathInline';
-  text: string;
-  raw: string;
-}
-
-const inlineMathRegex = /^\$(?!\s)(.+?)(?<!\s)\$(?!\d)/;
-
-export const mathInlineExtension = {
-  name: 'mathInline',
-  level: 'inline',
-  start(src: string) {
-    let index: number;
-    let indexSrc = src;
-
-    while (indexSrc) {
-      index = indexSrc.indexOf('$');
-      if (index === -1) {
-        return;
-      }
-      const f = index === 0 || indexSrc.charAt(index - 1) === ' ';
-      if (f) {
-        const possibleKatex = indexSrc.substring(index);
-        if (possibleKatex.match(inlineMathRegex)) {
-          return index;
-        }
-      }
-
-      indexSrc = indexSrc.substring(index + 1).replace(/^\$+/, '');
-    }
-  },
-  tokenizer(src: string): MathInlineToken | undefined {
-    const match = inlineMathRegex.exec(src);
-
-    if (match) {
-      return {
-        type: 'mathInline',
-        raw: match[0],
-        text: match[1]?.trim(),
-      };
-    }
-  },
-  renderer(token: Token) {
-    const mathInlineToken = token as MathInlineToken;
-    // parse to prevent escaping slashes
-    const latex = marked
-      .parse(mathInlineToken.text)
-      .toString()
-      .replace(/<(\/)?p>/g, '');
-
-    return `<span data-type="${mathInlineToken.type}" data-katex="true">${latex}</span>`;
-  },
-};

packages/editor-ext/src/lib/markdown/utils/spoiler.marked.test.ts

@@ -1,128 +0,0 @@
-import { describe, it, expect } from "vitest";
-import { getSchema } from "@tiptap/core";
-import { generateHTML, generateJSON } from "@tiptap/html";
-import { Document } from "@tiptap/extension-document";
-import { Paragraph } from "@tiptap/extension-paragraph";
-import { Text } from "@tiptap/extension-text";
-import { Bold } from "@tiptap/extension-bold";
-import { htmlToMarkdown } from "./turndown.utils";
-import { markdownToHtml } from "./marked.utils";
-import { Spoiler } from "../../spoiler/spoiler";
-
-// The spoiler mark has no native Markdown syntax, so it is preserved losslessly
-// as raw inline HTML (`<span data-spoiler="true">…</span>`), the same approach
-// htmlEmbed uses. This test drives the full editor round-trip:
-//   JSON -> HTML -> Markdown -> HTML -> JSON
-// and asserts the `spoiler` mark survives end to end. We use the same
-// getSchema + @tiptap/html generateHTML/generateJSON utilities the other
-// editor-ext schema tests use.
-
-const extensions = [Document, Paragraph, Text, Bold, Spoiler];
-
-function html(md: string): string {
-  const out = markdownToHtml(md);
-  if (typeof out !== "string") throw new Error("expected sync string output");
-  return out;
-}
-
-// Count text nodes carrying a `spoiler` mark anywhere in a ProseMirror JSON doc.
-function countSpoilerMarks(doc: any): number {
-  let count = 0;
-  const walk = (node: any) => {
-    if (!node || typeof node !== "object") return;
-    if (Array.isArray(node.marks)) {
-      for (const mark of node.marks) {
-        if (mark?.type === "spoiler") count++;
-      }
-    }
-    if (Array.isArray(node.content)) node.content.forEach(walk);
-  };
-  walk(doc);
-  return count;
-}
-
-describe("Spoiler mark schema", () => {
-  it("registers the spoiler mark in the schema", () => {
-    const schema = getSchema(extensions);
-    expect(schema.marks.spoiler).toBeTruthy();
-  });
-
-  it("recovers the spoiler mark from span[data-spoiler] (HTML -> JSON)", () => {
-    const json = generateJSON(
-      '<p>before <span data-spoiler="true">hidden</span> after</p>',
-      extensions,
-    );
-    expect(countSpoilerMarks(json)).toBe(1);
-  });
-
-  it("emits data-spoiler + class on render (JSON -> HTML)", () => {
-    const doc = {
-      type: "doc",
-      content: [
-        {
-          type: "paragraph",
-          content: [
-            {
-              type: "text",
-              text: "hidden",
-              marks: [{ type: "spoiler" }],
-            },
-          ],
-        },
-      ],
-    };
-    const out = generateHTML(doc, extensions);
-    expect(out).toContain('data-spoiler="true"');
-    expect(out).toContain('class="spoiler"');
-  });
-});
-
-describe("Spoiler Markdown round-trip is lossless", () => {
-  const docWith = (textNode: any) => ({
-    type: "doc",
-    content: [
-      {
-        type: "paragraph",
-        content: [{ type: "text", text: "before " }, textNode, { type: "text", text: " after" }],
-      },
-    ],
-  });
-
-  it("preserves the spoiler mark through JSON -> MD -> HTML -> JSON", () => {
-    const startDoc = docWith({
-      type: "text",
-      text: "hidden",
-      marks: [{ type: "spoiler" }],
-    });
-
-    // JSON -> HTML
-    const html1 = generateHTML(startDoc, extensions);
-    expect(html1).toContain('data-spoiler="true"');
-
-    // HTML -> Markdown (raw inline HTML, lossless)
-    const md = htmlToMarkdown(html1);
-    expect(md).toContain('<span data-spoiler="true">hidden</span>');
-
-    // MD -> HTML -> JSON (mark restored via parseHTML)
-    const endJson = generateJSON(html(md), extensions);
-    expect(countSpoilerMarks(endJson)).toBe(1);
-    // The visible text survives.
-    expect(JSON.stringify(endJson)).toContain("hidden");
-  });
-
-  it("keeps the spoiler intact when it intersects a bold mark", () => {
-    const startDoc = docWith({
-      type: "text",
-      text: "secret",
-      marks: [{ type: "bold" }, { type: "spoiler" }],
-    });
-
-    const md = htmlToMarkdown(generateHTML(startDoc, extensions));
-    expect(md).toContain("data-spoiler=\"true\"");
-
-    const endJson = generateJSON(html(md), extensions);
-    expect(countSpoilerMarks(endJson)).toBe(1);
-    // Bold survives alongside the spoiler.
-    expect(JSON.stringify(endJson)).toContain('"bold"');
-  });
-});

packages/editor-ext/src/lib/markdown/utils/turndown.d.ts

@@ -1,12 +0,0 @@
-// Map @joplin/turndown types to @types/turndown
-declare module "@joplin/turndown" {
-  import TurndownService from "turndown";
-  export = TurndownService;
-}
-
-declare module "@joplin/turndown-plugin-gfm" {
-  import TurndownService from "turndown";
-  export const tables: TurndownService.Plugin;
-  export const strikethrough: TurndownService.Plugin;
-  export const highlightedCodeBlock: TurndownService.Plugin;
-}

packages/editor-ext/src/lib/markdown/utils/turndown.dataloss.test.ts

@@ -1,147 +0,0 @@
-import { describe, it, expect } from "vitest";
-import { htmlToMarkdown } from "./turndown.utils";
-import { markdownToHtml } from "./marked.utils";
-
-/**
- * #206 mdrt-2 — Markdown export must never SILENTLY drop a block. (FIXED)
- *
- * `htmlToMarkdown` (turndown) historically only registered rules for a fixed
- * set of custom nodes (callout, taskItem, details, math, iframe, htmlEmbed,
- * image, video, footnote). Any other custom node — `transclusionReference`,
- * `pageBreak`, `mention`, `status` — fell through to turndown's default
- * handling: an empty wrapper is "blank" and removed, so the block disappeared
- * from the exported Markdown with no trace, and `mention`/`status` collapsed to
- * bare text, losing their identity (data-id / data-color). The invariant
- * "never silently lose a block" was broken.
- *
- * The fix adds lossless turndown rules that re-emit each of these nodes as raw
- * HTML carrying every `data-*` attribute. Plain-Markdown viewers ignore the
- * inert tag; the import path round-trips it (`markdownToHtml` passes the raw
- * HTML through and each node's `parseHTML` rebuilds the ProseMirror node). These
- * tests assert the surviving contract (the block is preserved AND its identity
- * round-trips back through import).
- */
-describe("htmlToMarkdown — custom nodes are preserved losslessly (#206 mdrt-2)", () => {
-  const wrap = (inner: string) => `<p>before</p>${inner}<p>after</p>`;
-
-  it("preserves a pageBreak block on Markdown export", () => {
-    const md = htmlToMarkdown(
-      wrap('<div data-type="pageBreak" class="page-break"></div>'),
-    );
-    expect(md).toContain("before");
-    expect(md).toContain("after");
-    // The break survives as an inert raw-HTML tag, not silently dropped.
-    expect(md).toMatch(/data-type="pageBreak"/);
-    expect(md).toMatch(/page-?break/i);
-  });
-
-  it("preserves a transclusionReference's identity on Markdown export", () => {
-    const md = htmlToMarkdown(
-      wrap('<div data-type="transclusionReference" data-id="abc"></div>'),
-    );
-    expect(md).toContain("before");
-    expect(md).toContain("after");
-    // The data-id (the only thing that gives the reference identity) survives.
-    expect(md).toContain("abc");
-    expect(md).toMatch(/data-type="transclusionReference"/);
-  });
-
-  it("preserves a mention's data-id (stable identity) on Markdown export", () => {
-    const md = htmlToMarkdown(
-      '<p>hi <span data-type="mention" data-id="u1" data-label="Bob">@Bob</span> there</p>',
-    );
-    // The mention keeps its stable identity (data-id), not just the text.
-    expect(md).toContain("u1");
-    expect(md).toContain("Bob");
-    expect(md).toMatch(/data-type="mention"/);
-  });
-
-  it("preserves a status chip's color on Markdown export", () => {
-    const md = htmlToMarkdown(
-      '<p>s <span data-type="status" data-color="green">Done</span></p>',
-    );
-    // The chip's color (its identity) survives, not just the visible text.
-    expect(md).toContain("green");
-    expect(md).toContain("Done");
-    expect(md).toMatch(/data-type="status"/);
-  });
-
-  // The export form is only lossless if the import path can rebuild it. These
-  // assert the full MD -> HTML round-trip restores the node + its attributes,
-  // which is the marker <-> node contract each `parseHTML` relies on.
-  describe("import round-trip (markdownToHtml restores the node)", () => {
-    it("round-trips a pageBreak through export + import", async () => {
-      const md = htmlToMarkdown(
-        wrap('<div data-type="pageBreak" class="page-break"></div>'),
-      );
-      const html = await markdownToHtml(md);
-      expect(html).toMatch(/<div[^>]*data-type="pageBreak"[^>]*>/);
-      expect(html).toContain("before");
-      expect(html).toContain("after");
-    });
-
-    it("round-trips a transclusionReference (keeps data-id)", async () => {
-      const md = htmlToMarkdown(
-        wrap('<div data-type="transclusionReference" data-id="abc"></div>'),
-      );
-      const html = await markdownToHtml(md);
-      expect(html).toMatch(/<div[^>]*data-type="transclusionReference"[^>]*>/);
-      expect(html).toContain("abc");
-    });
-
-    it("round-trips a mention (keeps data-id + data-label)", async () => {
-      const md = htmlToMarkdown(
-        '<p>hi <span data-type="mention" data-id="u1" data-label="Bob">@Bob</span> there</p>',
-      );
-      const html = await markdownToHtml(md);
-      expect(html).toMatch(/<span[^>]*data-type="mention"[^>]*>/);
-      expect(html).toContain("u1");
-      expect(html).toContain("Bob");
-    });
-
-    it("round-trips a status chip (keeps data-color)", async () => {
-      const md = htmlToMarkdown(
-        '<p>s <span data-type="status" data-color="green">Done</span></p>',
-      );
-      const html = await markdownToHtml(md);
-      expect(html).toMatch(/<span[^>]*data-type="status"[^>]*>/);
-      expect(html).toContain("green");
-    });
-
-    // HTML special chars in an attribute value or in a node's text must be
-    // ESCAPED when re-emitted as raw HTML, otherwise the exported tag is
-    // malformed and `markdownToHtml`'s parser cannot restore the original value
-    // (the same silent data loss this PR fixes). Dropping `<`/`>` escaping is the
-    // dangerous regression: a stray `<` or `>` corrupts the tag (or injects new
-    // markup), so the test data carries ALL of `&`, `"`, `<`, `>` in BOTH the
-    // data-label attribute and the visible text. That fully exercises
-    // escapeHtmlAttr's `&,",<,>` branches and escapeHtmlText's `&,<,>` branches
-    // (escapeHtmlText leaves `"` literal); the alphanumeric-only cases above hit
-    // none of them.
-    it("escapes HTML special chars (& \" < >) in attrs + text and round-trips them", async () => {
-      const md = htmlToMarkdown(
-        `<p>hi <span data-type="mention" data-id="u1" data-label="A &amp; &lt;B&gt; &quot;C&quot;">@A &amp; &lt;B&gt; "C"</span> there</p>`,
-      );
-
-      // (a) The exported Markdown carries a WELL-FORMED, correctly-escaped tag:
-      // the attribute escapes `&`, `<`, `>` AND `"`; the text escapes `&`, `<`,
-      // `>` (a `"` inside text content is legal, so it stays literal).
-      expect(md).toContain('data-label="A &amp; &lt;B&gt; &quot;C&quot;"');
-      expect(md).toContain('>@A &amp; &lt;B&gt; "C"</span>');
-      // And explicitly NOT the raw, tag-corrupting forms: a literal `<B>` (would
-      // mean `<`/`>` escaping was dropped in either the attr or the text)...
-      expect(md).not.toContain("<B>");
-      // ...nor the malformed attribute that an unescaped `"` would produce.
-      expect(md).not.toContain('data-label="A &amp; &lt;B&gt; "C""');
-
-      // (b) Import restores the ORIGINAL (unescaped) values, attribute and text.
-      const html = await markdownToHtml(md);
-      const dom = new DOMParser().parseFromString(html as string, "text/html");
-      const span = dom.querySelector('span[data-type="mention"]');
-      expect(span).not.toBeNull();
-      expect(span!.getAttribute("data-id")).toBe("u1");
-      expect(span!.getAttribute("data-label")).toBe('A & <B> "C"');
-      expect(span!.textContent).toBe('@A & <B> "C"');
-    });
-  });
-});

packages/editor-ext/src/lib/markdown/utils/turndown.utils.ts

@@ -1,488 +0,0 @@
-import * as _TurndownService from '@joplin/turndown';
-import * as TurndownPluginGfm from '@joplin/turndown-plugin-gfm';
-import { getBasename } from './basename';
-
-// CJS/ESM interop: .default exists in Vite, not in NestJS
-const TurndownService = (_TurndownService as any).default || _TurndownService;
-
-function sanitizeMdLinkText(value: string): string {
-  return value
-    .replace(/\\/g, '\\\\')
-    .replace(/([\[\]!])/g, '\\$1')
-    .replace(/[\r\n]+/g, ' ');
-}
-
-// Tags turndown treats as void (self-closing). Footnote references render as an
-// empty <sup data-footnote-ref> whose meaning lives entirely in its data-id;
-// without marking it void, turndown's blank-node removal drops it before our
-// rule runs, losing the `[^id]` marker. Mirrors turndown's built-in list.
-const TURNDOWN_VOID_ELEMENTS = [
-  'AREA', 'BASE', 'BR', 'COL', 'COMMAND', 'EMBED', 'HR', 'IMG', 'INPUT',
-  'KEYGEN', 'LINK', 'META', 'PARAM', 'SOURCE', 'TRACK', 'WBR',
-];
-
-function isVoidNode(node: any): boolean {
-  const name = node?.nodeName?.toUpperCase?.();
-  if (!name) return false;
-  if (name === 'SUP' && node.hasAttribute?.('data-footnote-ref')) {
-    return true;
-  }
-  return TURNDOWN_VOID_ELEMENTS.indexOf(name) !== -1;
-}
-
-/**
- * An empty <sup data-footnote-ref> is "blank" to turndown, which removes blank
- * inline nodes (RootNode/Node use a module-level isVoid the options cannot
- * override). To survive, inject the id as text content so the node is non-blank;
- * the footnoteReference rule then reads data-id and emits `[^id]`.
- */
-function fillEmptyFootnoteRefs(html: string): string {
-  return html.replace(
-    /<sup\b([^>]*\bdata-footnote-ref\b[^>]*)>\s*<\/sup>/gi,
-    (_m, attrs) => `<sup${attrs}>​</sup>`,
-  );
-}
-
-/**
- * `pageBreak` and `transclusionReference` are childless atom <div>s. Like an
- * empty footnote ref (see above), turndown treats a childless block as "blank"
- * and replaces it with the blankRule BEFORE any custom rule can fire — so the
- * node disappears from the export with no trace (#206 mdrt-2). Inject a
- * zero-width space so the node is non-blank and our lossless rule runs; the
- * rule rebuilds the tag from the element's attributes, so the injected char
- * never reaches the output.
- */
-function fillEmptyAtomBlocks(html: string): string {
-  return html.replace(
-    /<div\b([^>]*\bdata-type="(?:pageBreak|transclusionReference)"[^>]*)>\s*<\/div>/gi,
-    (_m, attrs) => `<div${attrs}>​</div>`,
-  );
-}
-
-/** HTML-escape an attribute value so a re-emitted raw-HTML tag is well-formed. */
-function escapeHtmlAttr(value: string): string {
-  return value
-    .replace(/&/g, '&amp;')
-    .replace(/"/g, '&quot;')
-    .replace(/</g, '&lt;')
-    .replace(/>/g, '&gt;');
-}
-
-/** HTML-escape text placed inside a re-emitted raw-HTML element. */
-function escapeHtmlText(value: string): string {
-  return value
-    .replace(/&/g, '&amp;')
-    .replace(/</g, '&lt;')
-    .replace(/>/g, '&gt;');
-}
-
-/**
- * Serialize ALL of an element's attributes back to a raw-HTML attribute string
- * (leading space included). Generic on purpose: a custom node's identity lives
- * entirely in its `data-*` attributes (data-id, data-color, data-source-page-id,
- * data-transclusion-id, …), and serializing every attribute keeps the export
- * lossless regardless of which attributes a given node carries.
- */
-function serializeAttrs(node: any): string {
-  const attrs = node?.attributes;
-  if (!attrs) return '';
-  return Array.from(attrs as ArrayLike<{ name: string; value: string }>)
-    .map((attr) => ` ${attr.name}="${escapeHtmlAttr(attr.value ?? '')}"`)
-    .join('');
-}
-
-export function htmlToMarkdown(html: string): string {
-  const turndownService = new TurndownService({
-    headingStyle: 'atx',
-    codeBlockStyle: 'fenced',
-    hr: '---',
-    bulletListMarker: '-',
-    isVoid: isVoidNode,
-  });
-
-  turndownService.use([
-    TurndownPluginGfm.tables,
-    TurndownPluginGfm.strikethrough,
-    TurndownPluginGfm.highlightedCodeBlock,
-    taskList,
-    callout,
-    preserveDetail,
-    listParagraph,
-    orderedListItem,
-    mathInline,
-    mathBlock,
-    iframeEmbed,
-    htmlEmbed,
-    spoiler,
-    image,
-    video,
-    footnoteReference,
-    footnotesList,
-    pageBreak,
-    transclusionReference,
-    mention,
-    status,
-  ]);
-  return turndownService
-    .turndown(fillEmptyAtomBlocks(fillEmptyFootnoteRefs(html)))
-    .replaceAll('<br>', ' ');
-}
-
-/**
- * Lossless export rules for custom nodes that have NO native Markdown syntax
- * (#206 mdrt-2). Markdown cannot represent a page break, a transclusion
- * reference, a mention's stable id, or a status chip's color — so rather than
- * letting turndown silently drop them, each rule re-emits the node as raw HTML
- * carrying every `data-*` attribute. Plain-Markdown viewers ignore the inert
- * tag, and the import path round-trips it: `markdownToHtml` passes raw HTML
- * through and each node's `parseHTML` (`div[data-type="…"]`, `span[…]`) rebuilds
- * the ProseMirror node with its attributes intact.
- */
-function pageBreak(turndownService: _TurndownService) {
-  turndownService.addRule('pageBreak', {
-    filter: function (node: HTMLInputElement) {
-      return (
-        node.nodeName === 'DIV' &&
-        node.getAttribute('data-type') === 'pageBreak'
-      );
-    },
-    replacement: function (_content: string, node: HTMLInputElement) {
-      return `\n\n<div${serializeAttrs(node)}></div>\n\n`;
-    },
-  });
-}
-
-function transclusionReference(turndownService: _TurndownService) {
-  turndownService.addRule('transclusionReference', {
-    filter: function (node: HTMLInputElement) {
-      return (
-        node.nodeName === 'DIV' &&
-        node.getAttribute('data-type') === 'transclusionReference'
-      );
-    },
-    replacement: function (_content: string, node: HTMLInputElement) {
-      return `\n\n<div${serializeAttrs(node)}></div>\n\n`;
-    },
-  });
-}
-
-function mention(turndownService: _TurndownService) {
-  turndownService.addRule('mention', {
-    filter: function (node: HTMLInputElement) {
-      return (
-        node.nodeName === 'SPAN' &&
-        node.getAttribute('data-type') === 'mention'
-      );
-    },
-    replacement: function (_content: string, node: HTMLInputElement) {
-      const text = escapeHtmlText(node.textContent || '');
-      return `<span${serializeAttrs(node)}>${text}</span>`;
-    },
-  });
-}
-
-function status(turndownService: _TurndownService) {
-  turndownService.addRule('status', {
-    filter: function (node: HTMLInputElement) {
-      return (
-        node.nodeName === 'SPAN' && node.getAttribute('data-type') === 'status'
-      );
-    },
-    replacement: function (_content: string, node: HTMLInputElement) {
-      const text = escapeHtmlText(node.textContent || '');
-      return `<span${serializeAttrs(node)}>${text}</span>`;
-    },
-  });
-}
-
-/**
- * Serialize the `htmlEmbed` node to Markdown.
- *
- * Markdown has no native representation for an arbitrary-HTML block, so we
- * preserve the node losslessly as an HTML comment carrying the base64-encoded
- * source (the same `data-source` payload the node stores). `markdownToHtml`
- * recognizes the same marker and rebuilds the node, so the round-trip
- * MD -> HTML -> JSON keeps the source intact. The comment also keeps the raw
- * markup inert in the exported `.md` file (it does not render in plain Markdown
- * viewers).
- */
-function htmlEmbed(turndownService: _TurndownService) {
-  turndownService.addRule('htmlEmbed', {
-    filter: function (node: HTMLInputElement) {
-      return (
-        node.nodeName === 'DIV' &&
-        node.getAttribute('data-type') === 'htmlEmbed'
-      );
-    },
-    replacement: function (_content: string, node: HTMLInputElement) {
-      const encoded = node.getAttribute('data-source') || '';
-      return `\n\n<!--html-embed:${encoded}-->\n\n`;
-    },
-  });
-}
-
-/**
- * Serialize the `spoiler` inline mark to lossless raw inline HTML.
- *
- * Markdown has no native spoiler syntax, so we emit the same `<span
- * data-spoiler="true">…</span>` the mark renders. `marked` passes inline raw HTML
- * through untouched, and `generateJSON` restores the mark via its parseHTML, so
- * the round-trip MD -> HTML -> JSON keeps the spoiler intact. The UI-only
- * `is-revealed` state is never serialized.
- */
-function spoiler(turndownService: _TurndownService) {
-  turndownService.addRule('spoiler', {
-    filter: function (node: HTMLInputElement) {
-      return (
-        node.nodeName === 'SPAN' &&
-        node.getAttribute('data-spoiler') === 'true'
-      );
-    },
-    replacement: function (content: string) {
-      return `<span data-spoiler="true">${content}</span>`;
-    },
-  });
-}
-
-function listParagraph(turndownService: _TurndownService) {
-  turndownService.addRule('paragraph', {
-    filter: ['p'],
-    replacement: (content: string, node: HTMLInputElement) => {
-      if (node.parentElement?.nodeName === 'LI') {
-        return content;
-      }
-      return `\n\n${content}\n\n`;
-    },
-  });
-}
-
-function orderedListItem(turndownService: _TurndownService) {
-  turndownService.addRule('orderedListItem', {
-    filter: function (node: HTMLInputElement) {
-      return node.nodeName === 'LI' && node.getAttribute('data-type') !== 'taskItem';
-    },
-    replacement: (content: string, node: HTMLInputElement, options: any) => {
-      const parent = node.parentNode as HTMLElement;
-      if (parent.nodeName !== 'OL' && parent.nodeName !== 'UL') {
-        return content;
-      }
-
-      content = content
-        .replace(/^\n+/, '')
-        .replace(/\n+$/, '\n')
-        .replace(/\n/gm, '\n  ');
-
-      let prefix: string;
-      if (parent.nodeName === 'OL') {
-        const start = parseInt(parent.getAttribute('start') || '1', 10);
-        const index = Array.prototype.indexOf.call(parent.children, node);
-        prefix = `${start + index}. `;
-      } else {
-        prefix = `${options.bulletListMarker} `;
-      }
-
-      return (
-        prefix +
-        content +
-        (node.nextSibling && !/\n$/.test(content) ? '\n' : '')
-      );
-    },
-  });
-}
-
-function callout(turndownService: _TurndownService) {
-  turndownService.addRule('callout', {
-    filter: function (node: HTMLInputElement) {
-      return (
-        node.nodeName === 'DIV' && node.getAttribute('data-type') === 'callout'
-      );
-    },
-    replacement: function (content: string, node: HTMLInputElement) {
-      const calloutType = node.getAttribute('data-callout-type');
-      return `\n\n:::${calloutType}\n${content.trim()}\n:::\n\n`;
-    },
-  });
-}
-
-function taskList(turndownService: _TurndownService) {
-  turndownService.addRule('taskListItem', {
-    filter: function (node: HTMLInputElement) {
-      return (
-        node.getAttribute('data-type') === 'taskItem' &&
-        node.parentNode.nodeName === 'UL'
-      );
-    },
-    replacement: function (_content: string, node: HTMLInputElement) {
-      const isChecked = node.getAttribute('data-checked') === 'true';
-      const div = node.querySelector('div');
-      const text = div ? div.textContent.trim() : node.textContent.trim();
-
-      const prefix = `- ${isChecked ? '[x]' : '[ ]'} `;
-
-      return (
-        prefix +
-        text +
-        (node.nextSibling && !/\n$/.test(text) ? '\n' : '')
-      );
-    },
-  });
-}
-
-function preserveDetail(turndownService: _TurndownService) {
-  turndownService.addRule('preserveDetail', {
-    filter: function (node: HTMLInputElement) {
-      return node.nodeName === 'DETAILS';
-    },
-    replacement: function (_content: string, node: HTMLInputElement) {
-      const summary = node.querySelector(':scope > summary');
-      let detailSummary = '';
-
-      if (summary) {
-        detailSummary = `<summary>${turndownService.turndown(summary.innerHTML)}</summary>`;
-      }
-
-      const detailsContent = Array.from(node.childNodes)
-        .filter((child) => child.nodeName !== 'SUMMARY')
-        .map((child) =>
-          child.nodeType === 1
-            ? turndownService.turndown((child as HTMLElement).outerHTML)
-            : child.textContent,
-        )
-        .join('');
-
-      return `\n<details>\n${detailSummary}\n\n${detailsContent}\n\n</details>\n`;
-    },
-  });
-}
-
-function mathInline(turndownService: _TurndownService) {
-  turndownService.addRule('mathInline', {
-    filter: function (node: HTMLInputElement) {
-      return (
-        node.nodeName === 'SPAN' &&
-        node.getAttribute('data-type') === 'mathInline'
-      );
-    },
-    replacement: function (content: string) {
-      return `$${content}$`;
-    },
-  });
-}
-
-function mathBlock(turndownService: _TurndownService) {
-  turndownService.addRule('mathBlock', {
-    filter: function (node: HTMLInputElement) {
-      return (
-        node.nodeName === 'DIV' &&
-        node.getAttribute('data-type') === 'mathBlock'
-      );
-    },
-    replacement: function (content: string) {
-      return `\n$$\n${content}\n$$\n`;
-    },
-  });
-}
-
-function iframeEmbed(turndownService: _TurndownService) {
-  turndownService.addRule('iframeEmbed', {
-    filter: function (node: HTMLInputElement) {
-      return node.nodeName === 'IFRAME';
-    },
-    replacement: function (_content: string, node: HTMLInputElement) {
-      const src = node.getAttribute('src');
-      return '[' + src + '](' + src + ')';
-    },
-  });
-}
-
-function image(turndownService: _TurndownService) {
-  turndownService.addRule('image', {
-    filter: 'img',
-    replacement: function (_content: string, node: HTMLInputElement) {
-      const src = node.getAttribute('src') || '';
-      if (!src) return '';
-      const caption = node.getAttribute('data-caption') || '';
-      if (caption) {
-        // ![]() can't carry a caption, so emit a raw <img> wrapped in a block
-        // <div>. marked passes it through and the image extension's parseHTML
-        // restores the caption from data-caption.
-        const parts = [`src="${escapeHtmlAttr(src)}"`];
-        const alt = node.getAttribute('alt') || '';
-        if (alt) parts.push(`alt="${escapeHtmlAttr(alt)}"`);
-        parts.push(`data-caption="${escapeHtmlAttr(caption)}"`);
-        return `<div><img ${parts.join(' ')}></div>`;
-      }
-      const alt = sanitizeMdLinkText(node.getAttribute('alt') || '');
-      const title = node.getAttribute('title') || '';
-      const titlePart = title ? ' "' + title.replace(/"/g, '\\"') + '"' : '';
-      return '![' + alt + '](' + src + titlePart + ')';
-    },
-  });
-}
-
-/**
- * Footnote reference (inline atom) -> pandoc/GFM marker `[^id]`.
- * The visible number is derived (not stored), so the id is the stable anchor.
- */
-function footnoteReference(turndownService: _TurndownService) {
-  turndownService.addRule('footnoteReference', {
-    filter: function (node: HTMLInputElement) {
-      return (
-        node.nodeName === 'SUP' && node.hasAttribute('data-footnote-ref')
-      );
-    },
-    replacement: function (_content: string, node: HTMLInputElement) {
-      const id = node.getAttribute('data-id') || '';
-      return id ? `[^${id}]` : '';
-    },
-  });
-}
-
-/**
- * Footnotes container -> the list of `[^id]: text` definitions at the end of
- * the document (one per line). Each footnoteDefinition inside emits its own
- * `[^id]: ...` line; turndown joins them with the surrounding block spacing.
- */
-function footnotesList(turndownService: _TurndownService) {
-  turndownService.addRule('footnoteDefinition', {
-    filter: function (node: HTMLInputElement) {
-      return (
-        node.nodeName === 'DIV' && node.hasAttribute('data-footnote-def')
-      );
-    },
-    replacement: function (content: string, node: HTMLInputElement) {
-      const id = node.getAttribute('data-id') || '';
-      // Collapse internal newlines so the definition stays a single MD line;
-      // continuation lines are a v2 refinement.
-      const text = content.replace(/\s*\n+\s*/g, ' ').trim();
-      return id ? `\n[^${id}]: ${text}\n` : '';
-    },
-  });
-
-  turndownService.addRule('footnotesList', {
-    filter: function (node: HTMLInputElement) {
-      return (
-        node.nodeName === 'SECTION' && node.hasAttribute('data-footnotes')
-      );
-    },
-    replacement: function (content: string) {
-      return `\n\n${content.trim()}\n`;
-    },
-  });
-}
-
-function video(turndownService: _TurndownService) {
-  turndownService.addRule('video', {
-    filter: function (node: HTMLInputElement) {
-      return node.tagName === 'VIDEO';
-    },
-    replacement: function (_content: string, node: HTMLInputElement) {
-      const src = node.getAttribute('src') || '';
-      const ariaLabel = node.getAttribute('aria-label');
-      const name = sanitizeMdLinkText(
-        ariaLabel || getBasename(src) || src,
-      );
-      return '[' + name + '](' + src + ')';
-    },
-  });
-}

packages/editor-ext/vitest.config.ts

@@ -14,10 +14,15 @@ export default defineConfig({
       provider: "v8",
       reporter: ["text-summary", "text"],
       all: false,
+      // functions lowered 60 -> 57 after issue #347 removed the editor-ext
+      // markdown layer (src/lib/markdown) and its image/footnote round-trip
+      // specs: that markdown behavior now lives in — and is tested by —
+      // @docmost/prosemirror-markdown, so the editor-ext baseline shifts down.
+      // Still a real gate (a few points below the post-removal measured level).
       thresholds: {
         statements: 54,
         branches: 44,
-        functions: 60,
+        functions: 57,
         lines: 54,
       },
     },

packages/mcp/src/client/context.ts

@@ -23,7 +23,6 @@ import { acquireCollabSession } from "../lib/collab-session.js";
 import { withPageLock, isUuid } from "../lib/page-lock.js";
 import { getCollabToken, performLogin } from "../lib/auth-utils.js";
 import { formatDocmostAxiosError } from "./errors.js";
-import { GetPageConversionCache } from "./getpage-cache.js";
 
 // A generic mixin base constructor (issue #450). Each domain mixin is a factory
 // `<T extends GConstructor<DocmostClientContext>>(Base: T) => class extends Base`
@@ -160,13 +159,6 @@ export abstract class DocmostClientContext {
   // bypassed on a forced refresh (the 401/403 reauth path). null = no token yet.
   protected collabTokenCache: { token: string; mintedAt: number } | null = null;
 
-  // Content-addressed conversion cache for getPage (issue #479). Keyed on
-  // (canonical pageId, updatedAt, optionsHash) -> the converted Markdown, so a
-  // re-read of an UNCHANGED page skips the expensive convertProseMirrorToMarkdown
-  // tree walk. Per-instance (a DocmostClient is built per user / per chat), so a
-  // cached conversion can never leak across identities. See getpage-cache.ts.
-  protected getPageCache = new GetPageConversionCache();
-
   // Two construction forms:
   //  - new DocmostClient(config)                  // discriminated union (current)
   //  - new DocmostClient(baseURL, email, password) // legacy positional creds

packages/mcp/src/client/getpage-cache.ts

@@ -1,147 +0,0 @@
-// Content-addressed LRU cache for the PM->Markdown conversion in getPage
-// (issue #479). getPage is the dominant agent op (812 calls / 2h, p95 840ms);
-// the bulk of its cost is convertProseMirrorToMarkdown — a full ProseMirror-tree
-// walk over the page content (hundreds of KB of JSON on large pages) run on
-// EVERY read. Since agents re-read far more than they write (812 reads vs 28
-// writes in the sample), most conversions re-produce the SAME markdown from
-// UNCHANGED content. This cache skips the recomputation on a hit.
-//
-// KEY = (pageId, updatedAt, optionsHash):
-//   - pageId: the page's CANONICAL UUID (resultData.id), not the agent-supplied
-//     slugId — so a slugId read and a UUID read of the same page share one entry.
-//   - updatedAt: comes from the SAME /pages/info response as `content`, so the
-//     two are mutually consistent; a changed page yields a new updatedAt -> a new
-//     key -> automatic, precise invalidation (no stale markdown is ever served).
-//   - optionsHash: a stable hash of the conversion options. getPage passes
-//     `{dropResolvedCommentAnchors:true}` while exportPageMarkdown passes the
-//     defaults (#328) — DIFFERENT output for the same content, so the options
-//     MUST be part of the key or a hit would serve the wrong variant.
-//
-// BOUNDS: evict the LEAST-recently-used entry when EITHER the entry count OR the
-// total stored bytes would exceed its cap. Large pages are hundreds of KB, so a
-// byte cap (not just a count cap) is what actually bounds memory. A Map iterates
-// in insertion order, so the first key is the LRU entry; a hit re-inserts its key
-// to move it to the most-recently-used end.
-//
-// This module is dependency-neutral (no axios/client/prom-client): a plain class
-// the shared client context owns one instance of, so the cache persists across
-// getPage calls on a single DocmostClient instance (built per user / per chat).
-
-/** A stable, order-insensitive hash of the conversion options object. */
-export function hashConvertOptions(options: unknown): string {
-  // JSON.stringify with SORTED keys makes the hash independent of key order, so
-  // {a:1,b:2} and {b:2,a:1} collapse to one entry. undefined/null options -> a
-  // fixed empty-object key, matching a caller that passes no options at all.
-  if (options === undefined || options === null) return "{}";
-  return stableStringify(options);
-}
-
-function stableStringify(value: any): string {
-  if (value === null || typeof value !== "object") return JSON.stringify(value);
-  if (Array.isArray(value)) return `[${value.map(stableStringify).join(",")}]`;
-  const keys = Object.keys(value).sort();
-  return `{${keys
-    .map((k) => `${JSON.stringify(k)}:${stableStringify(value[k])}`)
-    .join(",")}}`;
-}
-
-interface CacheEntry {
-  markdown: string;
-  bytes: number;
-}
-
-export interface GetPageCacheOptions {
-  /** Max number of entries before LRU eviction. Default 50. */
-  maxEntries?: number;
-  /** Max total stored bytes before LRU eviction. Default 10 MB. */
-  maxBytes?: number;
-}
-
-export class GetPageConversionCache {
-  private readonly maxEntries: number;
-  private readonly maxBytes: number;
-  // Insertion-ordered: the FIRST key is the least-recently-used entry.
-  private readonly map = new Map<string, CacheEntry>();
-  private totalBytes = 0;
-
-  constructor(opts: GetPageCacheOptions = {}) {
-    // A non-positive/NaN cap is treated as "use the default", never as an
-    // unbounded (or always-empty) cache — a silently unbounded cache would leak
-    // memory, and an always-empty one would defeat the whole optimization.
-    this.maxEntries =
-      Number.isFinite(opts.maxEntries) && (opts.maxEntries as number) > 0
-        ? Math.floor(opts.maxEntries as number)
-        : 50;
-    this.maxBytes =
-      Number.isFinite(opts.maxBytes) && (opts.maxBytes as number) > 0
-        ? Math.floor(opts.maxBytes as number)
-        : 10 * 1024 * 1024;
-  }
-
-  /** Compose the content-addressed key from its three parts. */
-  static key(pageId: string, updatedAt: string, optionsHash: string): string {
-    // A space separates the parts so no combination of values can collide by
-    // concatenation: a canonical UUID and an ISO updatedAt never contain a
-    // space, so the boundaries between the three parts are unambiguous.
-    return `${pageId} ${updatedAt} ${optionsHash}`;
-  }
-
-  /**
-   * Return the cached markdown for `key`, or undefined on a miss. A hit moves
-   * the entry to the most-recently-used end (delete + re-set) so the LRU order
-   * reflects real access, not just insertion.
-   */
-  get(key: string): string | undefined {
-    const entry = this.map.get(key);
-    if (entry === undefined) return undefined;
-    this.map.delete(key);
-    this.map.set(key, entry);
-    return entry.markdown;
-  }
-
-  /**
-   * Store `markdown` under `key`, then evict LRU entries until BOTH caps hold.
-   * Re-storing an existing key refreshes its value and recency (its old bytes
-   * are subtracted first, so totalBytes stays exact).
-   */
-  set(key: string, markdown: string): void {
-    // Byte size of the stored string (UTF-8). A single entry larger than the
-    // whole byte cap is still stored (so getPage always gets a hit next time),
-    // then the eviction loop below simply cannot shrink below it — accepted:
-    // one oversized page is bounded by the page itself, not a cache leak.
-    const bytes = Buffer.byteLength(markdown, "utf8");
-    const existing = this.map.get(key);
-    if (existing !== undefined) {
-      this.totalBytes -= existing.bytes;
-      this.map.delete(key);
-    }
-    this.map.set(key, { markdown, bytes });
-    this.totalBytes += bytes;
-    this.evict();
-  }
-
-  /** Evict the LRU entry until both the count and byte caps are satisfied. */
-  private evict(): void {
-    while (
-      this.map.size > this.maxEntries ||
-      (this.totalBytes > this.maxBytes && this.map.size > 1)
-    ) {
-      // The first key in insertion order is the least-recently-used.
-      const oldest = this.map.keys().next().value as string | undefined;
-      if (oldest === undefined) break;
-      const entry = this.map.get(oldest);
-      this.map.delete(oldest);
-      if (entry) this.totalBytes -= entry.bytes;
-    }
-  }
-
-  /** Current entry count (test/introspection). */
-  get size(): number {
-    return this.map.size;
-  }
-
-  /** Current total stored bytes (test/introspection). */
-  get bytes(): number {
-    return this.totalBytes;
-  }
-}

packages/mcp/src/client/read.ts

@@ -10,14 +10,7 @@ import {
   filterComment,
   filterSearchResult,
 } from "../lib/filters.js";
-import {
+import { convertProseMirrorToMarkdown } from "../lib/markdown-converter.js";
-  convertProseMirrorToMarkdown,
-  type ConvertProseMirrorToMarkdownOptions,
-} from "../lib/markdown-converter.js";
-import {
-  GetPageConversionCache,
-  hashConvertOptions,
-} from "./getpage-cache.js";
 import {
   collectInternalFileNodes,
   normalizeFileUrl,
@@ -402,19 +395,6 @@ export function ReadMixin<TBase extends GConstructor<DocmostClientContext>>(Base
 
   /** Raw page info including the ProseMirror JSON content and slugId. */
 
-  /**
-   * Overridable seam over convertProseMirrorToMarkdown (issue #479). Production
-   * just delegates; it exists as a method so a unit test can spy on it and
-   * assert the conversion is genuinely SKIPPED on a getPage cache HIT (the whole
-   * point of the cache) — an ESM named import cannot be intercepted otherwise.
-   */
-  protected convertPageMarkdown(
-    content: any,
-    options: ConvertProseMirrorToMarkdownOptions,
-  ): string {
-    return convertProseMirrorToMarkdown(content, options);
-  }
-
   async getPage(pageId: string) {
     await this.ensureAuthenticated();
     const resultData = await this.getPageRaw(pageId);
@@ -423,55 +403,13 @@ export function ReadMixin<TBase extends GConstructor<DocmostClientContext>>(Base
     // discussions. Active anchors are kept. (The lossless exportPageMarkdown
     // round-trip deliberately does NOT pass this flag — resolved anchors there
     // must be preserved.)
-    //
+    let content = resultData.content
-    // Content-addressed conversion cache (issue #479): the PM->Markdown walk is
+      ? convertProseMirrorToMarkdown(resultData.content, {
-    // the dominant cost of this hot read op. Key on the page's canonical UUID +
+          dropResolvedCommentAnchors: true,
-    // updatedAt (both from THIS /pages/info response, so mutually consistent) +
+        })
-    // a hash of the conversion options. A hit returns the cached markdown and
+      : "";
-    // skips the walk; a miss converts and stores. The cached value is the
-    // conversion output BEFORE the {{SUBPAGES}} substitution below, which uses
-    // live subpage data and stays outside the cache — so the final result is
-    // byte-identical to the uncached path.
-    const convertOptions = { dropResolvedCommentAnchors: true };
-    let content = "";
-    if (resultData.content) {
-      // Only cache when we have a stable identity+version for the key. Both come
-      // from the same response; if either is missing (unexpected server shape),
-      // fall back to converting uncached rather than keying on a partial tuple.
-      const cacheable =
-        typeof resultData.id === "string" &&
-        typeof resultData.updatedAt === "string";
-      const cacheKey = cacheable
-        ? GetPageConversionCache.key(
-            resultData.id,
-            resultData.updatedAt,
-            hashConvertOptions(convertOptions),
-          )
-        : null;
 
-      const cached = cacheKey ? this.getPageCache.get(cacheKey) : undefined;
+    // Always fetch subpages to provide context to the agent
-      if (cached !== undefined) {
-        content = cached;
-        this.onMetricFn?.("mcp_getpage_cache_hits_total", 1);
-      } else {
-        // Goes through the convertPageMarkdown seam (not the raw import) so a
-        // test can assert the conversion is SKIPPED on a hit (issue #479 F2).
-        content = this.convertPageMarkdown(resultData.content, convertOptions);
-        if (cacheKey) this.getPageCache.set(cacheKey, content);
-        // A non-cacheable page (missing id/updatedAt) is still a genuine
-        // conversion, so it counts as a miss for an honest hit-rate.
-        this.onMetricFn?.("mcp_getpage_cache_misses_total", 1);
-      }
-    }
-
-    // Always fetch subpages to provide context to the agent.
-    //
-    // NOT parallelizable with the page fetch (issue #479 asked to check): the
-    // sidebar-pages endpoint REQUIRES spaceId in its POST body, and spaceId is
-    // only known FROM this page fetch's response (resolvePageId yields the UUID
-    // but never the spaceId). So `Promise.all([pageFetch, subpagesFetch])` would
-    // have to invent a spaceId it does not have — the two calls are inherently
-    // sequential. Correctness wins; the conversion cache above is the real speedup.
     let subpages: any[] = [];
     try {
       // `pageId` may be a slugId, but the sidebar-pages endpoint requires the

packages/mcp/test/mock/getpage-conversion-cache.test.mjs

@@ -1,251 +0,0 @@
-// Mock-HTTP integration tests for the getPage conversion cache (issue #479).
-// A local http.createServer stands in for Docmost (same harness style as
-// get-page-context.test.mjs) so everything is deterministic and offline.
-//
-// Verifies end-to-end through the real client that:
-//   - the FIRST getPage of a page is a MISS (mcp_getpage_cache_misses_total)
-//     and converts the content (the server's convert-representative counter);
-//   - a SECOND getPage of the same (pageId, updatedAt) is a HIT
-//     (mcp_getpage_cache_hits_total) and returns BYTE-IDENTICAL output while
-//     skipping the conversion;
-//   - a changed updatedAt is a fresh key -> MISS again;
-//   - the returned shape still resolves page + subpages.
-import { test, after, mock } from "node:test";
-import assert from "node:assert/strict";
-import http from "node:http";
-import { DocmostClient } from "../../build/client.js";
-
-function readBody(req) {
-  return new Promise((resolve) => {
-    let raw = "";
-    req.on("data", (chunk) => (raw += chunk));
-    req.on("end", () => resolve(raw));
-  });
-}
-
-function sendJson(res, status, obj, extraHeaders = {}) {
-  res.writeHead(status, { "Content-Type": "application/json", ...extraHeaders });
-  res.end(JSON.stringify(obj));
-}
-
-const openServers = [];
-after(async () => {
-  await Promise.all(openServers.map((s) => new Promise((r) => s.close(r))));
-});
-
-const PAGE_UUID = "00000000-0000-4000-8000-000000000010";
-const SPACE_UUID = "00000000-0000-4000-8000-0000000000aa";
-const CHILD_UUID = "00000000-0000-4000-8000-0000000000bb";
-
-// A small ProseMirror doc so the converter produces non-trivial markdown.
-function makeDoc(text) {
-  return {
-    type: "doc",
-    content: [
-      {
-        type: "paragraph",
-        content: [{ type: "text", text }],
-      },
-    ],
-  };
-}
-
-// state.info counts /pages/info hits; state.updatedAt / state.text drive the
-// content+version returned; state.sidebar counts sidebar-pages hits;
-// state.subpages (when set) drives the child list the sidebar endpoint returns,
-// so a test can vary the live subpages across two reads of the same page.
-function spawn(state) {
-  return new Promise((resolve) => {
-    const server = http.createServer(async (req, res) => {
-      await readBody(req);
-      if (req.url === "/api/auth/login") {
-        return sendJson(res, 200, { success: true }, {
-          "Set-Cookie": "authToken=t; Path=/; HttpOnly",
-        });
-      }
-      if (req.url === "/api/pages/info") {
-        state.info++;
-        return sendJson(res, 200, {
-          success: true,
-          data: {
-            id: PAGE_UUID,
-            slugId: "slug123456",
-            title: "Cached Page",
-            parentPageId: null,
-            spaceId: SPACE_UUID,
-            updatedAt: state.updatedAt,
-            content: makeDoc(state.text),
-          },
-        });
-      }
-      if (req.url === "/api/pages/sidebar-pages") {
-        state.sidebar++;
-        const items = state.subpages ?? [
-          { id: CHILD_UUID, title: "Child", hasChildren: false },
-        ];
-        return sendJson(res, 200, {
-          success: true,
-          data: {
-            items,
-            meta: { hasNextPage: false, nextCursor: null },
-          },
-        });
-      }
-      return sendJson(res, 404, { message: "not found" });
-    });
-    server.listen(0, "127.0.0.1", () => {
-      openServers.push(server);
-      resolve(`http://127.0.0.1:${server.address().port}/api`);
-    });
-  });
-}
-
-function makeClient(baseURL, metrics) {
-  return new DocmostClient({
-    apiUrl: baseURL,
-    getToken: async () => "access",
-    onMetric: (name, value) => {
-      metrics[name] = (metrics[name] ?? 0) + value;
-    },
-  });
-}
-
-test("first read MISS, second read HIT with byte-identical output; convert runs once", async () => {
-  const state = { info: 0, sidebar: 0, updatedAt: "2026-01-01T00:00:00Z", text: "Hello world" };
-  const baseURL = await spawn(state);
-  const metrics = {};
-  const client = makeClient(baseURL, metrics);
-
-  const first = await client.getPage(PAGE_UUID);
-  assert.equal(metrics["mcp_getpage_cache_misses_total"], 1, "first read is a miss");
-  assert.equal(metrics["mcp_getpage_cache_hits_total"] ?? 0, 0, "no hit yet");
-
-  const second = await client.getPage(PAGE_UUID);
-  assert.equal(metrics["mcp_getpage_cache_hits_total"], 1, "second read is a hit");
-  assert.equal(metrics["mcp_getpage_cache_misses_total"], 1, "still one miss");
-
-  // BYTE-IDENTICAL: the cache only skips recomputation, never changes output.
-  assert.deepEqual(second, first, "cached result is identical to the uncached one");
-  assert.equal(
-    JSON.stringify(second),
-    JSON.stringify(first),
-    "serialized output is byte-identical",
-  );
-
-  // The page fetch + subpages fetch still happen every call (only the CPU
-  // conversion is cached); both reads hit /pages/info and sidebar-pages.
-  assert.equal(state.info, 2, "both reads still fetch /pages/info");
-  assert.equal(state.sidebar, 2, "both reads still fetch subpages");
-
-  // Shape sanity: content present, subpages resolved.
-  assert.equal(typeof second.data.content, "string");
-  assert.ok(second.data.content.includes("Hello world"));
-  assert.deepEqual(second.data.subpages, [{ id: CHILD_UUID, title: "Child" }]);
-});
-
-test("a changed updatedAt is a fresh key -> MISS again, with the NEW content", async () => {
-  const state = { info: 0, sidebar: 0, updatedAt: "2026-01-01T00:00:00Z", text: "Version one" };
-  const baseURL = await spawn(state);
-  const metrics = {};
-  const client = makeClient(baseURL, metrics);
-
-  const a = await client.getPage(PAGE_UUID); // miss
-  const b = await client.getPage(PAGE_UUID); // hit
-  assert.equal(metrics["mcp_getpage_cache_misses_total"], 1);
-  assert.equal(metrics["mcp_getpage_cache_hits_total"], 1);
-  assert.ok(a.data.content.includes("Version one"));
-
-  // The page changes: new updatedAt AND new content.
-  state.updatedAt = "2026-02-02T00:00:00Z";
-  state.text = "Version two";
-
-  const c = await client.getPage(PAGE_UUID); // miss on the new key
-  assert.equal(metrics["mcp_getpage_cache_misses_total"], 2, "changed version -> miss");
-  assert.equal(metrics["mcp_getpage_cache_hits_total"], 1, "no stale hit");
-  assert.ok(c.data.content.includes("Version two"), "the NEW content is served");
-  assert.ok(!c.data.content.includes("Version one"), "no stale markdown");
-
-  const d = await client.getPage(PAGE_UUID); // hit on the new key
-  assert.equal(metrics["mcp_getpage_cache_hits_total"], 2, "the new snapshot caches too");
-});
-
-test("a slugId read and a UUID read of the same page share one cache entry", async () => {
-  // resolvePageId maps the slugId -> UUID via /pages/info; the cache keys on the
-  // canonical UUID (resultData.id), so both inputs land on the same entry.
-  const state = { info: 0, sidebar: 0, updatedAt: "2026-01-01T00:00:00Z", text: "Shared" };
-  const baseURL = await spawn(state);
-  const metrics = {};
-  const client = makeClient(baseURL, metrics);
-
-  await client.getPage(PAGE_UUID);          // miss (keyed on UUID)
-  await client.getPage("slug123456");        // the server returns the same id -> HIT
-  assert.equal(metrics["mcp_getpage_cache_misses_total"], 1, "one conversion total");
-  assert.equal(metrics["mcp_getpage_cache_hits_total"], 1, "slugId read hits the UUID entry");
-});
-
-test("on a conversion HIT, the {{SUBPAGES}} block reflects the LIVE subpages, not the cached ones", async () => {
-  // The whole byte-identity guarantee: the cache stores the conversion output
-  // BEFORE the {{SUBPAGES}} substitution, so a re-read of an UNCHANGED page still
-  // splices the FRESH subpage list. The page body itself contains {{SUBPAGES}}
-  // (converts to a literal placeholder); getPage replaces it with the live list.
-  const CHILD_A = "00000000-0000-4000-8000-0000000000a1";
-  const CHILD_B = "00000000-0000-4000-8000-0000000000b2";
-  const state = {
-    info: 0,
-    sidebar: 0,
-    updatedAt: "2026-01-01T00:00:00Z", // FIXED across both reads -> conversion cache HIT
-    text: "Body before {{SUBPAGES}} body after",
-    subpages: [{ id: CHILD_A, title: "Alpha", hasChildren: false }],
-  };
-  const baseURL = await spawn(state);
-  const metrics = {};
-  const client = makeClient(baseURL, metrics);
-
-  // Read 1: MISS (converts). The substitution runs with list A.
-  const first = await client.getPage(PAGE_UUID);
-  assert.equal(metrics["mcp_getpage_cache_misses_total"], 1, "first read converts (miss)");
-  assert.ok(first.data.content.includes("[Alpha](page:" + CHILD_A + ")"), "list A spliced in");
-  assert.ok(!first.data.content.includes("{{SUBPAGES}}"), "placeholder consumed");
-
-  // The subpages change while the PAGE CONTENT/updatedAt do NOT: same conversion
-  // cache key -> a HIT that skips the CPU walk, but the live substitution must
-  // still run on the NEW list B.
-  state.subpages = [{ id: CHILD_B, title: "Beta", hasChildren: false }];
-
-  const second = await client.getPage(PAGE_UUID);
-  assert.equal(metrics["mcp_getpage_cache_hits_total"], 1, "second read is a conversion HIT");
-  assert.equal(metrics["mcp_getpage_cache_misses_total"], 1, "no second conversion");
-
-  // The cache did NOT freeze the subpages block: list B is present, list A gone.
-  assert.ok(second.data.content.includes("[Beta](page:" + CHILD_B + ")"), "live list B spliced in on a HIT");
-  assert.ok(!second.data.content.includes("Alpha"), "stale list A is NOT frozen into the output");
-  assert.deepEqual(second.data.subpages, [{ id: CHILD_B, title: "Beta" }], "subpages field reflects list B");
-});
-
-test("a cache HIT SKIPS the convertProseMirrorToMarkdown CPU walk (called once across MISS+HIT)", async () => {
-  // The single reason the cache exists: on a hit the expensive PM-tree walk must
-  // NOT run. The miss counter alone can't prove this — a broken hit branch that
-  // re-converted (same output, misses=1) would leave every other assert green.
-  // So spy directly on the conversion seam and assert the call COUNT.
-  const state = { info: 0, sidebar: 0, updatedAt: "2026-01-01T00:00:00Z", text: "Body text" };
-  const baseURL = await spawn(state);
-  const metrics = {};
-  const client = makeClient(baseURL, metrics);
-
-  // Spy on the seam that wraps convertProseMirrorToMarkdown; it still delegates,
-  // so output stays real and byte-identical — we only count invocations.
-  const spy = mock.method(client, "convertPageMarkdown");
-
-  await client.getPage(PAGE_UUID); // MISS -> converts once
-  assert.equal(spy.mock.callCount(), 1, "the miss converts exactly once");
-
-  await client.getPage(PAGE_UUID); // HIT -> must NOT convert again
-  assert.equal(
-    spy.mock.callCount(),
-    1,
-    "the hit skips the conversion: still exactly one call across MISS+HIT",
-  );
-  assert.equal(metrics["mcp_getpage_cache_hits_total"], 1, "and it was recorded as a hit");
-
-  spy.mock.restore();
-});

packages/mcp/test/unit/getpage-cache.test.mjs

@@ -1,113 +0,0 @@
-// Unit tests for the getPage content-addressed conversion cache (issue #479).
-// Exercises the GetPageConversionCache class in isolation: key composition,
-// hit/miss, LRU recency on read, and eviction by BOTH the count cap and the
-// byte cap. The getPage integration (counter emission, byte-identical output)
-// is covered separately in test/mock/getpage-conversion-cache.test.mjs.
-import { test } from "node:test";
-import assert from "node:assert/strict";
-import {
-  GetPageConversionCache,
-  hashConvertOptions,
-} from "../../build/client/getpage-cache.js";
-
-test("key: identical (pageId, updatedAt, optionsHash) collapse to one entry", () => {
-  const c = new GetPageConversionCache();
-  const k1 = GetPageConversionCache.key("uuid-1", "2026-01-01T00:00:00Z", "{}");
-  const k2 = GetPageConversionCache.key("uuid-1", "2026-01-01T00:00:00Z", "{}");
-  assert.equal(k1, k2, "same parts -> same key");
-  c.set(k1, "MD-A");
-  assert.equal(c.get(k2), "MD-A", "hit on the identical key");
-  assert.equal(c.size, 1);
-});
-
-test("miss after updatedAt changes (precise invalidation)", () => {
-  const c = new GetPageConversionCache();
-  const oldK = GetPageConversionCache.key("uuid-1", "v1", "{}");
-  const newK = GetPageConversionCache.key("uuid-1", "v2", "{}");
-  c.set(oldK, "OLD-MD");
-  assert.equal(c.get(newK), undefined, "a new updatedAt is a fresh key -> miss");
-  assert.equal(c.get(oldK), "OLD-MD", "the old snapshot is still addressable");
-});
-
-test("miss after options change (dropResolvedCommentAnchors true vs false)", () => {
-  const c = new GetPageConversionCache();
-  const hDrop = hashConvertOptions({ dropResolvedCommentAnchors: true });
-  const hKeep = hashConvertOptions({ dropResolvedCommentAnchors: false });
-  assert.notEqual(hDrop, hKeep, "different options hash to different keys");
-  const kDrop = GetPageConversionCache.key("uuid-1", "v1", hDrop);
-  const kKeep = GetPageConversionCache.key("uuid-1", "v1", hKeep);
-  c.set(kDrop, "AGENT-MD");
-  assert.equal(
-    c.get(kKeep),
-    undefined,
-    "the export variant must NOT be served the agent variant",
-  );
-});
-
-test("hashConvertOptions is order-insensitive and handles empty", () => {
-  assert.equal(
-    hashConvertOptions({ a: 1, b: 2 }),
-    hashConvertOptions({ b: 2, a: 1 }),
-    "key order does not change the hash",
-  );
-  assert.equal(hashConvertOptions(undefined), "{}");
-  assert.equal(hashConvertOptions(null), "{}");
-  assert.equal(hashConvertOptions({}), "{}");
-});
-
-test("LRU eviction by COUNT cap evicts the least-recently-used entry", () => {
-  const c = new GetPageConversionCache({ maxEntries: 2, maxBytes: 10 * 1024 * 1024 });
-  c.set("k1", "a");
-  c.set("k2", "b");
-  c.set("k3", "c"); // over the count cap -> evict k1 (oldest)
-  assert.equal(c.size, 2);
-  assert.equal(c.get("k1"), undefined, "k1 evicted");
-  assert.equal(c.get("k2"), "b");
-  assert.equal(c.get("k3"), "c");
-});
-
-test("a read refreshes recency so the OTHER entry is evicted next", () => {
-  const c = new GetPageConversionCache({ maxEntries: 2, maxBytes: 10 * 1024 * 1024 });
-  c.set("k1", "a");
-  c.set("k2", "b");
-  // Touch k1 so k2 becomes the least-recently-used.
-  assert.equal(c.get("k1"), "a");
-  c.set("k3", "c"); // evicts the LRU, which is now k2 (not k1)
-  assert.equal(c.get("k2"), undefined, "k2 was LRU and got evicted");
-  assert.equal(c.get("k1"), "a", "k1 survived because it was read");
-  assert.equal(c.get("k3"), "c");
-});
-
-test("LRU eviction by BYTE cap evicts oldest until under the cap", () => {
-  // Byte cap of 100; each value is 40 bytes -> at most 2 fit (80 < 100 < 120).
-  const big = "x".repeat(40);
-  const c = new GetPageConversionCache({ maxEntries: 50, maxBytes: 100 });
-  c.set("k1", big); // 40
-  c.set("k2", big); // 80
-  assert.equal(c.size, 2);
-  c.set("k3", big); // 120 > 100 -> evict k1 -> 80
-  assert.equal(c.size, 2, "byte cap forced an eviction despite the count cap");
-  assert.equal(c.get("k1"), undefined, "oldest evicted by bytes");
-  assert.equal(c.get("k2"), big);
-  assert.equal(c.get("k3"), big);
-  assert.ok(c.bytes <= 100, "total bytes stays within the cap");
-});
-
-test("re-setting an existing key updates value+recency and keeps bytes exact", () => {
-  const c = new GetPageConversionCache({ maxEntries: 5, maxBytes: 10 * 1024 * 1024 });
-  c.set("k1", "short");
-  const b1 = c.bytes;
-  assert.equal(b1, Buffer.byteLength("short", "utf8"));
-  c.set("k1", "a much longer value");
-  assert.equal(c.size, 1, "no duplicate entry");
-  assert.equal(c.bytes, Buffer.byteLength("a much longer value", "utf8"));
-  assert.equal(c.get("k1"), "a much longer value");
-});
-
-test("an oversized single entry is still stored (never a permanent miss)", () => {
-  const c = new GetPageConversionCache({ maxEntries: 5, maxBytes: 10 });
-  const big = "y".repeat(1000);
-  c.set("k1", big);
-  assert.equal(c.get("k1"), big, "the page bigger than the cap is still served");
-  assert.equal(c.size, 1);
-});

packages/prosemirror-markdown/README.md

@@ -1,12 +1,31 @@
 # @docmost/prosemirror-markdown
 
 The single, canonical **ProseMirror ↔ Markdown converter** plus the Docmost
-schema mirror (#293/#345). Headless and framework-free: no React, no browser
+schema mirror (#293/#345/#347). Headless and framework-free: no React. There is
-runtime. There is exactly ONE copy of this converter in the repo, consumed by:
+exactly ONE copy of this converter in the repo, consumed by:
 
 - `packages/mcp` (the MCP server),
 - `packages/git-sync` (two-way Git sync),
-- `apps/server` (server-side markdown import/export, #345).
+- `apps/server` (server-side markdown import/export, #345),
+- `apps/client` (markdown paste/copy + AI-chat render, #347).
+
+### Node vs browser entry
+
+The HTML→DOM stage of markdown import runs on `jsdom` in Node and the native
+`DOMParser` in the browser, injected per environment so **jsdom never enters a
+client bundle**:
+
+- default entry (`@docmost/prosemirror-markdown`) — Node: registers jsdom +
+  `@tiptap/html`'s happy-dom `server` `generateJSON`. Used by mcp / git-sync /
+  apps/server.
+- `browser` entry (`@docmost/prosemirror-markdown/browser`, via the `"browser"`
+  exports condition) — registers the native `DOMParser` + `@tiptap/html`'s
+  browser `generateJSON`. Used by `apps/client`; carries no jsdom/happy-dom.
+
+Both entries expose the identical converter surface; only the injected
+DOM/`generateJSON` implementations differ (`src/lib/dom-parser.ts`). A
+`markdownToProseMirrorSync` variant exists for callers that cannot await (the
+client's synchronous chat renderer).
 
 `src/lib/docmost-schema.ts` **mirrors** the upstream Tiptap schema that lives in
 `packages/editor-ext`. The mirror is not free-floating: `serializer-contract.test.ts`

packages/prosemirror-markdown/package.json

@@ -9,7 +9,12 @@
   "exports": {
     ".": {
       "types": "./build/index.d.ts",
+      "browser": "./build/browser.js",
       "default": "./build/index.js"
+    },
+    "./browser": {
+      "types": "./build/browser.d.ts",
+      "default": "./build/browser.js"
     }
   },
   "scripts": {
@@ -30,6 +35,7 @@
     "@tiptap/html": "3.20.4",
     "@tiptap/pm": "3.20.4",
     "@tiptap/starter-kit": "3.20.4",
+    "happy-dom": "20.8.9",
     "jsdom": "25.0.0",
     "marked": "17.0.5",
     "zod": "4.3.6"

packages/prosemirror-markdown/src/browser.ts

@@ -0,0 +1,17 @@
+/**
+ * BROWSER entry of `@docmost/prosemirror-markdown`.
+ *
+ * Selected via the package's `"browser"` exports condition (bundlers) so the
+ * client gets the SAME public converter surface as the Node entry, but the
+ * markdown-import DOM passes run on the native `window.DOMParser` instead of
+ * jsdom. This module installs the native parser as a side effect BEFORE
+ * re-exporting, and imports NO jsdom — so a client bundle that resolves this
+ * entry carries no `jsdom` (nor any transitive jsdom import).
+ *
+ * The re-exported surface is identical to the default (Node) entry — the only
+ * difference is which HTML-DOM parser is registered — so a browser consumer can
+ * call `markdownToProseMirror` (and everything else) exactly as the server does.
+ */
+import "./lib/dom-parser.browser.js";
+
+export * from "./lib/index.js";

packages/prosemirror-markdown/src/index.ts

@@ -6,4 +6,11 @@
  * this top-level barrel simply re-exports that surface so the package entry is
  * the converter surface.
  */
+// DEFAULT (Node) entry: install the jsdom-backed HTML parser as a side effect
+// BEFORE re-exporting the converter surface, so every Node consumer (server,
+// mcp, git-sync) keeps the identical jsdom import behaviour with no code change.
+// The browser entry (`./browser.js`) installs the native-`DOMParser` parser
+// instead and never loads this module, so jsdom stays out of client bundles.
+import "./lib/dom-parser.node.js";
+
 export * from "./lib/index.js";

packages/prosemirror-markdown/src/lib/docmost-schema.ts

@@ -63,10 +63,9 @@ function getStyleProperty(element: HTMLElement, propertyName: string): string |
  * The editor SCHEMA genuinely only supports these six banner types — there is no
  * `tip`/`caution`/`important`/`question` callout node. So those are NOT first-
  * class types we can round-trip literally; they are INPUT ALIASES (GitHub/Obsidian
- * alert syntax). The editor's own paste/import path maps them onto the supported
+ * alert syntax). This package's own `> [!type]` import path maps them onto the
- * set (see `GITHUB_ALERT_TYPE_MAP` in
+ * supported set (see `CALLOUT_TYPE_ALIASES` below: tip -> success, caution ->
- * `@docmost/editor-ext` markdown/utils/github-callout.marked.ts:
+ * danger, important -> info). We apply that aliasing
- * tip -> success, caution -> danger, important -> info). We mirror that aliasing
  * here so an ingested `> [!tip]` / `> [!caution]` lands on the closest real banner
  * (success / danger) instead of flatly collapsing to `info` — matching exactly how
  * the editor itself would interpret the same alias. A schema type always maps to
@@ -75,11 +74,11 @@ function getStyleProperty(element: HTMLElement, propertyName: string): string |
  */
 const CALLOUT_TYPES = ["default", "info", "note", "success", "warning", "danger"];
 /**
- * NON-schema callout aliases -> their closest supported banner. Mirrors the
+ * NON-schema callout aliases -> their closest supported banner, for the names
- * editor's `GITHUB_ALERT_TYPE_MAP` for the names that are NOT already schema
+ * that are NOT already schema types (a schema type is preserved as-is and never
- * types (a schema type is preserved as-is and never consulted here). Keeping
+ * consulted here). This is the single canonical alias map now that the editor's
- * these in lockstep means git-sync ingest and an editor paste interpret the same
+ * old marked layer is gone; git-sync ingest and an editor paste both go through
- * `> [!alias]` identically.
+ * this package, so they interpret the same `> [!alias]` identically.
  */
 const CALLOUT_TYPE_ALIASES: Record<string, string> = {
   tip: "success",

packages/prosemirror-markdown/src/lib/dom-parser.browser.ts

@@ -0,0 +1,27 @@
+/**
+ * BROWSER registration of the injectable HTML parser (native `DOMParser`).
+ *
+ * Importing this module for its SIDE EFFECT installs a `window.DOMParser`-backed
+ * {@link HtmlDocumentParser}. It is loaded by the package's `browser.ts` barrel
+ * (selected via the `"browser"` exports condition). It imports NO `jsdom`, so a
+ * client bundle that resolves the browser entry never pulls jsdom in.
+ *
+ * The returned `Document` is a real browser document, so it exposes exactly the
+ * same query/mutation surface the import passes use (`querySelector(All)`,
+ * `createElement`, `createTreeWalker`/`NodeFilter` via `defaultView`,
+ * `body.innerHTML`) as jsdom did on the Node path.
+ */
+// @tiptap/html's default (browser) entry: its `generateJSON` uses the native
+// `window.DOMParser`, so it carries NO jsdom/happy-dom — keeping the client
+// bundle free of Node-only DOM libs.
+import { generateJSON } from "@tiptap/html";
+import { setHtmlDocumentParser, setGenerateJson } from "./dom-parser.js";
+
+setHtmlDocumentParser((html: string): Document => {
+  // Native, always available in a browser (and in a jsdom/happy-dom test
+  // environment, which is what the client vitest suite runs under). `text/html`
+  // parsing matches jsdom's `new JSDOM(html)` behaviour for our fragments.
+  return new DOMParser().parseFromString(html, "text/html");
+});
+
+setGenerateJson(generateJSON);

packages/prosemirror-markdown/src/lib/dom-parser.node.ts

@@ -0,0 +1,30 @@
+/**
+ * NODE registration of the injectable HTML parser (jsdom-backed).
+ *
+ * Importing this module for its SIDE EFFECT installs a jsdom-backed
+ * {@link HtmlDocumentParser}. It is loaded by the package's default entry
+ * (`index.ts`) so every existing Node consumer (server, mcp, git-sync) keeps
+ * the identical jsdom behaviour with no code change. This is the ONLY module in
+ * the import chain that imports `jsdom`; the browser entry never loads it, so
+ * `jsdom` cannot reach the client bundle.
+ */
+import { JSDOM } from "jsdom";
+// Use @tiptap/html's EXPLICIT server entry (happy-dom backed): it builds its own
+// DOM internally and needs NO ambient global `window`, so the Node path never
+// depends on which of @tiptap/html's conditional exports a resolver picks (Jest
+// selects the browser entry, which would throw without a global window). This
+// avoids the old module-level `global.window` jsdom shim entirely — that shim
+// was timing-fragile (it had to be installed AFTER prosemirror-view's
+// import-time env detection, or prosemirror-view reads an undefined `navigator`).
+import { generateJSON } from "@tiptap/html/server";
+import { setHtmlDocumentParser, setGenerateJson } from "./dom-parser.js";
+
+setHtmlDocumentParser((html: string): Document => {
+  // A fresh JSDOM per call mirrors the previous `new JSDOM(html)` usage in each
+  // import pass — no shared mutable document between conversions, so concurrent
+  // conversions never interfere.
+  const dom = new JSDOM(html);
+  return dom.window.document as unknown as Document;
+});
+
+setGenerateJson(generateJSON);

packages/prosemirror-markdown/src/lib/dom-parser.ts

@@ -0,0 +1,87 @@
+/**
+ * Injectable HTML-string -> DOM parser for the markdown import path.
+ *
+ * The markdown -> ProseMirror chain (`markdown-to-prosemirror.ts`) does three
+ * post-`marked` DOM passes (task-list bridge, comment directives, footnote
+ * assembly) that need a real DOM to query/mutate an HTML fragment. On the NODE
+ * path that DOM comes from `jsdom`; in the BROWSER the platform already provides
+ * a native `DOMParser` and `document`, and `jsdom` must NOT be bundled (size +
+ * it is Node-only). So the concrete parser is INJECTED per environment rather
+ * than imported statically here — this module carries no `jsdom` (or any DOM)
+ * import, so nothing on the browser code path can transitively pull `jsdom` in.
+ *
+ * The Node entry (`./dom-parser.node.js`, loaded by the package's default
+ * `index.js`) registers a jsdom-backed parser; the browser entry
+ * (`./dom-parser.browser.js`, loaded by the `browser.js` barrel) registers a
+ * `window.DOMParser`-backed one. A consumer that forgets to load an entry (e.g.
+ * a raw deep import) gets a clear error instead of a silent wrong-environment
+ * crash.
+ */
+
+/**
+ * Parse an HTML string into a `Document`. The returned document must support the
+ * standard query/mutation surface the import passes use: `querySelector(All)`,
+ * `createElement`, `createTreeWalker` + `NodeFilter` (read off the document's
+ * `defaultView`), and `body.innerHTML`.
+ */
+export type HtmlDocumentParser = (html: string) => Document;
+
+/**
+ * Convert an HTML string to a ProseMirror JSON doc against the given TipTap
+ * extension set. This is `@tiptap/html`'s `generateJSON`, injected per
+ * environment so the Node path binds its happy-dom `server` entry and the
+ * browser path its native-`DOMParser` entry — neither leaking the other's DOM
+ * lib into the wrong bundle.
+ */
+export type GenerateJsonFn = (html: string, extensions: any[]) => any;
+
+let injectedParser: HtmlDocumentParser | null = null;
+let injectedGenerateJson: GenerateJsonFn | null = null;
+
+/**
+ * Register the environment's HTML parser. Called ONCE at import time by the
+ * Node or browser entry module. Idempotent-friendly: the last registration
+ * wins, so a test harness can override it.
+ */
+export function setHtmlDocumentParser(parser: HtmlDocumentParser): void {
+  injectedParser = parser;
+}
+
+/**
+ * Parse `html` into a `Document` using the registered parser. Throws a clear
+ * error when no environment entry has registered one (the caller imported the
+ * converter without going through the Node/browser barrel).
+ */
+export function parseHtmlDocument(html: string): Document {
+  if (!injectedParser) {
+    throw new Error(
+      "No HTML DOM parser registered. Import `@docmost/prosemirror-markdown` " +
+        "(Node) or `@docmost/prosemirror-markdown/browser` (browser) so the " +
+        "environment's DOM parser is installed before calling the converter.",
+    );
+  }
+  return injectedParser(html);
+}
+
+/**
+ * Register the environment's `generateJSON` (HTML -> ProseMirror JSON). Called
+ * ONCE at import time by the Node or browser entry module.
+ */
+export function setGenerateJson(fn: GenerateJsonFn): void {
+  injectedGenerateJson = fn;
+}
+
+/**
+ * Run the registered `generateJSON`. Throws a clear error when no environment
+ * entry has registered one (same cause as {@link parseHtmlDocument}).
+ */
+export function generateJsonWith(html: string, extensions: any[]): any {
+  if (!injectedGenerateJson) {
+    throw new Error(
+      "No generateJSON registered. Import `@docmost/prosemirror-markdown` " +
+        "(Node) or `@docmost/prosemirror-markdown/browser` (browser) so the " +
+        "environment's generateJSON is installed before calling the converter.",
+    );
+  }
+  return injectedGenerateJson(html, extensions);
+}

packages/prosemirror-markdown/src/lib/index.ts

@@ -18,7 +18,10 @@ export type { DocmostMdMeta } from "./markdown-document.js";
 export { convertProseMirrorToMarkdown } from "./markdown-converter.js";
 export type { ConvertProseMirrorToMarkdownOptions } from "./markdown-converter.js";
 
-export { markdownToProseMirror } from "./markdown-to-prosemirror.js";
+export {
+  markdownToProseMirror,
+  markdownToProseMirrorSync,
+} from "./markdown-to-prosemirror.js";
 
 // The Docmost tiptap schema mirror. Exposed so consumers (and the sync
 // engine's schema-validity regression tests) can build the exact ProseMirror

packages/prosemirror-markdown/src/lib/markdown-to-prosemirror.ts

@@ -7,9 +7,8 @@
  * natively through the collab gateway, so no websocket/Yjs write-path lives
  * here.
  */
-import { generateJSON } from "@tiptap/html";
-import { JSDOM } from "jsdom";
 import { Marked } from "marked";
+import { parseHtmlDocument, generateJsonWith } from "./dom-parser.js";
 import type { TokenizerExtension, RendererExtension } from "marked";
 import { docmostExtensions } from "./docmost-schema.js";
 import { parseAttachedComment } from "./attached-comment.js";
@@ -245,12 +244,13 @@ const markedInstance = new Marked().use({
   ],
 });
 
-// Setup DOM environment for Tiptap HTML parsing in Node.js
+// NOTE: this module no longer installs a module-level `global.window`/`document`
-const dom = new JSDOM("<!DOCTYPE html><html><body></body></html>");
+// jsdom shim. The HTML->DOM passes below (bridgeTaskLists / applyCommentDirectives
-global.window = dom.window as any;
+// / assembleFootnotes) parse via the INJECTED `parseHtmlDocument` (jsdom on the
-global.document = dom.window.document;
+// Node entry, native `DOMParser` on the browser entry), and `@tiptap/html`'s v3
-// @ts-ignore
+// `generateJSON` supplies its OWN DOM per environment (happy-dom in Node, native
-global.Element = dom.window.Element;
+// `DOMParser` in the browser) — so no ambient global DOM is needed here, and
+// nothing on the browser code path statically imports jsdom.
 
 /**
  * Hard ceiling above which we skip callout preprocessing entirely. The linear
@@ -295,7 +295,13 @@ const CODE_FENCE_RE = /^(\s*)(`{3,}|~{3,})/;
  *   - emits the same `<div data-type="callout" data-callout-type="TYPE">` output
  *     (inner rendered through marked) as the previous regex implementation.
  */
-async function preprocessCallouts(markdown: string): Promise<string> {
+// SYNCHRONOUS by construction: the only formerly-awaited call is
+// `markedInstance.parse`, which returns a string synchronously for this
+// instance (no async marked extensions are registered), so the whole callout
+// preprocess is sync. Keeping it sync lets a sync converter entry
+// (`markdownToProseMirrorSync`, used by the client's chat renderer which must
+// stay synchronous) share this exact logic with the async entry.
+function preprocessCallouts(markdown: string): string {
   // Defensive cap: skip preprocessing for pathologically large inputs.
   if (markdown.length > MAX_CALLOUT_PREPROCESS_BYTES) {
     return markdown;
@@ -304,7 +310,7 @@ async function preprocessCallouts(markdown: string): Promise<string> {
   // Recursively transform a slice of lines, converting top-level callouts in
   // that slice into <div> blocks and rendering their inner content (which may
   // itself contain nested callouts) through this same function.
-  const transform = async (lines: string[]): Promise<string> => {
+  const transform = (lines: string[]): string => {
     const out: string[] = [];
     let inCodeFence = false;
     let codeFenceMarker = ""; // the exact run of backticks/tildes that opened it
@@ -383,8 +389,8 @@ async function preprocessCallouts(markdown: string): Promise<string> {
         if (j < lines.length) {
           // Found the matching closing fence: render the body (recursively, so
           // nested callouts are handled) and emit the callout div.
-          const inner = await transform(bodyLines);
+          const inner = transform(bodyLines);
-          const renderedInner = await markedInstance.parse(inner);
+          const renderedInner = markedInstance.parse(inner) as string;
           out.push(
             `\n<div data-type="callout" data-callout-type="${type}">${renderedInner}</div>\n`,
           );
@@ -423,8 +429,8 @@ async function preprocessCallouts(markdown: string): Promise<string> {
           // Drop the prefix + `>` + one optional space, leaving the body content.
           bodyLines.push(lines[j].slice(prefix.length).replace(/^>\s?/, ""));
         }
-        const inner = await transform(bodyLines);
+        const inner = transform(bodyLines);
-        const renderedInner = await markedInstance.parse(inner);
+        const renderedInner = markedInstance.parse(inner) as string;
         const block = `<div data-type="callout" data-callout-type="${type}">${renderedInner}</div>`;
         if (prefix.length === 0) {
           // Top-level callout: blank lines isolate the HTML block.
@@ -491,8 +497,7 @@ function bridgeTaskLists(html: string): string {
   if (html.length > MAX_CALLOUT_PREPROCESS_BYTES) {
     return html;
   }
-  const dom = new JSDOM(html);
+  const document = parseHtmlDocument(html);
-  const document = dom.window.document;
   // Collect the checkbox(es) that belong to THIS <li> directly: either direct
   // child <input type="checkbox"> elements or ones inside the <li>'s direct <p>
   // child (the shape marked emits: `<li><p><input type="checkbox"> text</p></li>`).
@@ -662,15 +667,23 @@ function placeStandalone(
 function applyCommentDirectives(html: string): string {
   // Cheap early-out: no comments at all -> nothing to intercept.
   if (!html.includes("<!--")) return html;
-  const dom = new JSDOM(html);
+  const document = parseHtmlDocument(html);
-  const document = dom.window.document;
+  // `SHOW_COMMENT` (128) is a stable DOM constant. Read it from whichever holder
-  const nodeFilter = dom.window.NodeFilter;
+  // exists — the document's window (jsdom: `defaultView`), the ambient global
+  // `NodeFilter` (browsers / test envs), else the literal — because a document
+  // produced by `DOMParser.parseFromString` has NO browsing context, so its
+  // `defaultView` is `null` (unlike a jsdom `new JSDOM(html).window.document`).
+  // `createTreeWalker` takes the numeric `whatToShow` mask directly.
+  const SHOW_COMMENT =
+    (document.defaultView as any)?.NodeFilter?.SHOW_COMMENT ??
+    (globalThis as any).NodeFilter?.SHOW_COMMENT ??
+    0x80;
   // Walk the WHOLE document, not just <body>: when a standalone machinery
   // comment is the FIRST thing in the output (before any body content), the
   // HTML parser places it at document level (a child of `#document`, before
   // `<html>`), where it is outside `document.body` and would be lost. Attached
   // attrs comments always live inside body, so this wider walk still finds them.
-  const walker = document.createTreeWalker(document, nodeFilter.SHOW_COMMENT);
+  const walker = document.createTreeWalker(document, SHOW_COMMENT);
   const comments: any[] = [];
   let current: any;
   while ((current = walker.nextNode())) comments.push(current);
@@ -945,8 +958,7 @@ const MAX_FOOTNOTE_ROUNDS = 10000;
 function assembleFootnotes(html: string): string {
   // Cheap early-out: nothing carries a footnote body -> nothing to assemble.
   if (!html.includes("data-fn-text")) return html;
-  const dom = new JSDOM(html);
+  const document = parseHtmlDocument(html);
-  const document = dom.window.document;
   if (document.querySelector("sup[data-footnote-ref][data-fn-text]") == null) {
     return html;
   }
@@ -1060,12 +1072,18 @@ function stripEmptyParagraphs(node: any): any {
   return { ...node, content: cleaned };
 }
 
-/** Convert markdown to a ProseMirror doc using the full Docmost schema. */
+/**
-export async function markdownToProseMirror(
+ * Convert markdown to a ProseMirror doc using the full Docmost schema
-  markdownContent: string,
+ * (SYNCHRONOUS core). Every stage — callout preprocess, `marked` parse, the
-): Promise<any> {
+ * three DOM passes, and generateJSON — is synchronous for this configuration
-  const withCallouts = await preprocessCallouts(markdownContent);
+ * (no async marked extensions), so the conversion needs no `await`. The async
-  const html = await markedInstance.parse(withCallouts);
+ * `markdownToProseMirror` below delegates here (its Promise return is preserved
+ * for every existing Node consumer). A sync entry is REQUIRED by the client's
+ * chat renderer, which runs inside a React render/useMemo and cannot await.
+ */
+export function markdownToProseMirrorSync(markdownContent: string): any {
+  const withCallouts = preprocessCallouts(markdownContent);
+  const html = markedInstance.parse(withCallouts) as string;
   // Materialize comment directives (#293 #9 attached textAlign; #5 standalone
   // subpages/pageBreak) while the comment nodes still exist, before generateJSON
   // drops them.
@@ -1075,6 +1093,17 @@ export async function markdownToProseMirror(
   // generateJSON, so references + definitions materialize into the schema model.
   const withFootnotes = assembleFootnotes(withAttrs);
   const bridged = bridgeTaskLists(withFootnotes);
-  const doc = generateJSON(bridged, docmostExtensions);
+  const doc = generateJsonWith(bridged, docmostExtensions);
   return stripEmptyParagraphs(doc);
 }
+
+/**
+ * Convert markdown to a ProseMirror doc (async entry, unchanged contract). Kept
+ * async so every existing Node consumer (server, mcp, git-sync) that `await`s
+ * it is untouched; it simply delegates to the synchronous core.
+ */
+export async function markdownToProseMirror(
+  markdownContent: string,
+): Promise<any> {
+  return markdownToProseMirrorSync(markdownContent);
+}

packages/prosemirror-markdown/test/dom-parser-both-paths.test.ts

@@ -0,0 +1,101 @@
+import { afterEach, beforeEach, describe, expect, it } from "vitest";
+import { JSDOM } from "jsdom";
+import {
+  setHtmlDocumentParser,
+  parseHtmlDocument,
+} from "../src/lib/dom-parser.js";
+import { markdownToProseMirror } from "../src/lib/markdown-to-prosemirror.js";
+
+/**
+ * The markdown import path parses its post-`marked` HTML through an INJECTED
+ * DOM parser: jsdom on the Node entry, native `DOMParser` on the browser entry
+ * (see dom-parser.node.ts / dom-parser.browser.ts). These tests exercise BOTH
+ * registrations against the same canonical inputs — the three DOM passes
+ * (task-list bridge, comment directives, footnote assembly) — and assert the
+ * converter produces the IDENTICAL ProseMirror doc regardless of which DOM
+ * parser is installed. That is the guarantee the client paste path relies on:
+ * pasting in the browser must yield the same nodes the server import produces.
+ */
+
+// A jsdom-backed parser (the Node entry's registration).
+const jsdomParser = (html: string): Document =>
+  new JSDOM(html).window.document as unknown as Document;
+
+// A `DOMParser`-backed parser standing in for the BROWSER entry's registration.
+// We drive a real `DOMParser` from a jsdom window (jsdom exposes the same
+// `new DOMParser().parseFromString(html, "text/html")` API the browser and the
+// client's jsdom vitest environment provide), so this path uses the exact code
+// dom-parser.browser.ts runs — a native `DOMParser`, no `JSDOM` document glue.
+const browserWindow = new JSDOM("").window;
+const domParserBackedParser = (html: string): Document =>
+  new browserWindow.DOMParser().parseFromString(
+    html,
+    "text/html",
+  ) as unknown as Document;
+
+// Canonical inputs, each hitting a different post-marked DOM pass.
+const CASES: Record<string, string> = {
+  // footnote assembly (assembleFootnotes): `^[…]` -> sup + section/def
+  "inline footnote ^[…]": "Body^[a note].",
+  // task-list bridge (bridgeTaskLists): checkbox list -> taskList/taskItem
+  "task list": "- [x] done\n- [ ] todo",
+  // comment directives (applyCommentDirectives): standalone machinery comment
+  "standalone subpages comment": "text\n\n<!--subpages-->\n\ntext2",
+  // attached image comment (applyCommentDirectives img form)
+  "attached image comment": '![alt](img.png) <!--img {"align":"left"}-->',
+  // github callout (preprocessCallouts bq path) + comment pass
+  "obsidian callout": "> [!info]\n> hello",
+  // highlight + math (marked extensions; still parsed through the DOM stage)
+  "highlight + math": "A ==mark== and $x^2$ end",
+};
+
+async function convertWith(
+  parser: (html: string) => Document,
+  md: string,
+): Promise<any> {
+  setHtmlDocumentParser(parser);
+  return markdownToProseMirror(md);
+}
+
+describe("markdown import: Node (jsdom) and browser (DOMParser) DOM paths agree", () => {
+  afterEach(() => {
+    // Restore the jsdom parser the suite-wide setup file installs, so later
+    // tests in the run are unaffected by our per-case swaps.
+    setHtmlDocumentParser(jsdomParser);
+  });
+
+  for (const [name, md] of Object.entries(CASES)) {
+    it(`produces identical nodes for: ${name}`, async () => {
+      const viaJsdom = await convertWith(jsdomParser, md);
+      const viaDomParser = await convertWith(domParserBackedParser, md);
+      expect(viaDomParser).toEqual(viaJsdom);
+    });
+  }
+});
+
+describe("dom-parser injection contract", () => {
+  let saved: (html: string) => Document;
+  beforeEach(() => {
+    saved = jsdomParser;
+  });
+  afterEach(() => {
+    setHtmlDocumentParser(saved);
+  });
+
+  it("throws a clear error when no parser is registered", () => {
+    // Install a thrower to simulate the unregistered state, then assert
+    // parseHtmlDocument surfaces the guidance error (we cannot un-set the
+    // module singleton, so we assert via a registration that throws the same).
+    setHtmlDocumentParser(() => {
+      throw new Error("No HTML DOM parser registered.");
+    });
+    expect(() => parseHtmlDocument("<p>x</p>")).toThrow(/No HTML DOM parser/);
+  });
+
+  it("uses the most-recently registered parser", () => {
+    const marker = new JSDOM("<!DOCTYPE html><body><b>marker</b></body>").window
+      .document as unknown as Document;
+    setHtmlDocumentParser(() => marker);
+    expect(parseHtmlDocument("<i>ignored</i>")).toBe(marker);
+  });
+});

packages/prosemirror-markdown/test/setup.dom-parser.ts

@@ -0,0 +1,13 @@
+/**
+ * Vitest setup: register the Node (jsdom) HTML parser for the whole suite.
+ *
+ * The package's tests import the converter through the RELATIVE `src/lib/*`
+ * modules (or the `docmost-client`/`src/lib/index` barrel), NOT through the
+ * top-level `index.ts` entry that a real Node consumer imports — so the entry's
+ * side-effect registration of the jsdom parser never runs here. This setup file
+ * performs the SAME registration the Node entry does, so `markdownToProseMirror`
+ * has a DOM parser in the (node-environment) tests, matching production Node
+ * behaviour. The browser path is covered separately by the client paste tests
+ * (jsdom vitest env + native DOMParser) and the dedicated dom-parser test.
+ */
+import "../src/lib/dom-parser.node.js";

packages/prosemirror-markdown/vitest.config.ts

@@ -19,5 +19,9 @@ export default defineConfig({
   test: {
     environment: 'node',
     include: ['test/**/*.test.ts'],
+    // Register the Node (jsdom) HTML parser before any test runs. Tests import
+    // the converter via relative src/lib modules, bypassing the top-level entry
+    // that normally installs the parser as a side effect (see setup file).
+    setupFiles: ['test/setup.dom-parser.ts'],
   },
 });

pnpm-lock.yaml

@@ -284,6 +284,9 @@ importers:
       '@docmost/editor-ext':
         specifier: workspace:*
         version: link:../../packages/editor-ext
+      '@docmost/prosemirror-markdown':
+        specifier: workspace:*
+        version: link:../../packages/prosemirror-markdown
       '@excalidraw/excalidraw':
         specifier: 0.18.0-3a5ef40
         version: 0.18.0-3a5ef40(@types/react-dom@18.3.1)(@types/react@18.3.12)(react-dom@18.3.1(react@18.3.1))(react@18.3.1)
@@ -912,10 +915,6 @@ importers:
         version: 8.57.1(eslint@9.39.4(jiti@2.4.2))(typescript@5.9.3)
 
   packages/editor-ext:
-    dependencies:
-      marked:
-        specifier: 17.0.5
-        version: 17.0.5
     devDependencies:
       '@vitest/coverage-v8':
         specifier: 4.1.6
@@ -1117,6 +1116,9 @@ importers:
       '@tiptap/starter-kit':
         specifier: 3.20.4
         version: 3.20.4
+      happy-dom:
+        specifier: 20.8.9
+        version: 20.8.9
       jsdom:
         specifier: 25.0.0
         version: 25.0.0
@@ -18490,7 +18492,7 @@ snapshots:
 
   happy-dom@20.8.9:
     dependencies:
-      '@types/node': 22.19.1
+      '@types/node': 25.5.0
       '@types/whatwg-mimetype': 3.0.2
       '@types/ws': 8.18.1
       entities: 7.0.1