@ai-matrx/print 0.8.1 → 0.8.3

package/CHANGELOG.md

@@ -1,5 +1,27 @@
 # Changelog
 
+## 0.8.3
+
+- **Fixed:** an `<artifact>` envelope (how chat answers store tables and code)
+  printed as raw pipes in documents. The document pipeline now unwraps it with
+  the same rules the print skins use (`unwrapArtifactsToMarkdown`, exported
+  from `/markdown`): a table stays a table, code stays code.
+- **Fixed (EPUB):** a `$$…$$` formula inside a sentence rendered as a `<div>`
+  inside `<p>` (epubcheck RSC-005); it is now a block-styled span.
+- **Word:** East-Asian text carries a CJK font hint, so Word/LibreOffice pick
+  an installed Japanese/Chinese face.
+- Consumer action: none — refresh to latest.
+
+## 0.8.2
+
+- **PDF font fallback covers Japanese / Chinese / Korean.** Characters the
+  body, emoji and symbol faces cannot draw now fall back to Noto Sans JP
+  (SIL OFL, fontsource) — loading only the unicode-range subsets the document
+  uses (`fonts.cjk` loader; default jsDelivr). Found on a real conversation
+  whose citations carried Japanese source titles. A glyph is announced only
+  when no face in the chain can draw it.
+- Consumer action: none — refresh to latest.
+
 ## 0.8.1
 
 - **Fixed (Word):** exporting DOCX laid out a whole PDF to number the table of

package/dist/document.cjs

@@ -35,12 +35,14 @@ __export(document_exports, {
   CitationEngine: () => CitationEngine,
   DEFAULT_FOOTER: () => DEFAULT_FOOTER,
   DOCUMENT_FORMATS: () => DOCUMENT_FORMATS,
+  NOTO_CJK_BASE: () => NOTO_CJK_BASE,
   NOTO_EMOJI_URL: () => NOTO_EMOJI_URL,
   NOTO_SYMBOLS_URL: () => NOTO_SYMBOLS_URL,
   PAGE_SIZES: () => PAGE_SIZES,
   buildPdfDefinition: () => buildPdfDefinition,
   countUnplacedWords: () => countUnplacedWords,
   cslStyleId: () => cslStyleId,
+  defaultCjkFontLoader: () => defaultCjkFontLoader,
   defaultCslFetcher: () => defaultCslFetcher,
   defaultEmojiFontLoader: () => defaultEmojiFontLoader,
   defaultImageLoader: () => defaultImageLoader,
@@ -63,6 +65,7 @@ __export(document_exports, {
   parseLength: () => parseLength,
   parseLocator: () => parseLocator,
   parseMathML: () => parseMathML,
+  parseUnicodeRangeCss: () => parseUnicodeRangeCss,
   plainText: () => plainText,
   prepareDocument: () => prepareDocument,
   renderBlocksHtml: () => renderBlocksHtml,
@@ -3104,7 +3107,11 @@ async function documentToDocx(tree, options = {}) {
     numbering: { config: [...c.numbering.entries()].map(([reference, levels2]) => ({ reference, levels: levels2 })) },
     styles: {
       default: {
-        document: { run: { font: "Cambria", size, color: HEX.text }, paragraph: { spacing: { after: 140, line: 300 } } }
+        // East-Asian text gets a CJK face (Word and LibreOffice substitute an installed one).
+        document: {
+          run: { font: { ascii: "Cambria", hAnsi: "Cambria", cs: "Cambria", eastAsia: "Yu Gothic" }, size, color: HEX.text },
+          paragraph: { spacing: { after: 140, line: 300 } }
+        }
       },
       paragraphStyles: [
         heading("Heading1", "Heading 1", s.fontSize * 1.8, 0),
@@ -3247,7 +3254,7 @@ function renderInlineHtml(nodes, x) {
       case "cite":
         return n.resolved ? `<a class="doc-cite" href="${(0, import_html_escape.escapeHtml)(href("ref-" + (n.keys[0] ?? ""), x))}">${(0, import_html_escape.escapeHtml)(n.label)}</a>` : `<span class="doc-cite doc-cite-missing">${(0, import_html_escape.escapeHtml)(n.label)}</span>`;
       case "math":
-        return mathHtml(n.tex, n.display, x);
+        return mathHtml(n.tex, n.display, x, true);
       case "noteRef":
         return `<sup class="doc-note-ref"><a href="${(0, import_html_escape.escapeHtml)(href(`note-${n.n}`, x))}" id="noteref-${n.n}">${n.n}</a></sup>`;
       default:
@@ -3255,10 +3262,11 @@ function renderInlineHtml(nodes, x) {
     }
   }).join("");
 }
-function mathHtml(tex, display, x) {
+function mathHtml(tex, display, x, inText = false) {
   const m = x.tree.math.get(mathKey(tex, display));
   if (m?.mathml && !m.error) {
-    return display ? `<div class="doc-math-display">${m.mathml}</div>` : `<span class="doc-math">${m.mathml}</span>`;
+    if (!display) return `<span class="doc-math">${m.mathml}</span>`;
+    return inText ? `<span class="doc-math-display">${m.mathml}</span>` : `<div class="doc-math-display">${m.mathml}</div>`;
   }
   return `<code class="doc-math-tex">${(0, import_html_escape.escapeHtml)(display ? `$$${tex}$$` : `$${tex}$`)}</code>`;
 }
@@ -3420,7 +3428,7 @@ table caption{caption-side:top;text-align:left;margin-bottom:4pt;}
 .doc-figure figcaption{margin-top:4pt;}
 .doc-image-missing{border:1px dashed var(--matrx-print-rule,#d1d5db);padding:16pt;color:var(--matrx-print-muted,#6b7280);}
 .doc-inline-img{max-height:1.2em;vertical-align:middle;}
-.doc-math-display{margin:0.6em 0 0.9em;text-align:center;break-inside:avoid;overflow-x:auto;}
+.doc-math-display{display:block;margin:0.6em 0 0.9em;text-align:center;break-inside:avoid;overflow-x:auto;}
 .doc-math-display math{font-size:1.15em;}
 .doc-math-tex{color:var(--matrx-print-danger,#b91c1c);}
 .doc-task-list{list-style:none;padding-left:0.2em;}
@@ -3499,7 +3507,7 @@ blockquote{margin:0 0 0.8em;padding-left:1em;border-left:3px solid #d1d5db;color
 .doc-cover-logo{max-height:5em;}
 .doc-cover-title{font-size:2em;}
 .matrx-page-break{page-break-after:always;break-after:page;}
-.doc-math-display{margin:0.6em 0;text-align:center;}
+.doc-math-display{display:block;margin:0.6em 0;text-align:center;}
 `.trim();
 function extOf(mime) {
   return mime === "image/png" ? "png" : mime === "image/gif" ? "gif" : "jpg";
@@ -4533,12 +4541,13 @@ async function loadTypography(tree, loaders = {}) {
   const base = await loadFaces(vfs, s.fontSize);
   let needEmoji = false;
   let needSymbols = false;
+  const cjk = /* @__PURE__ */ new Set();
   for (const ch of documentText(tree)) {
     const cp = ch.codePointAt(0);
     if (cp < 32 || cp === 8205 || cp === 65039 || base.body(cp)) continue;
     if (EMOJI_CP.test(ch)) needEmoji = true;
+    else if (isCjk(cp)) cjk.add(cp);
     else needSymbols = true;
-    if (needEmoji && needSymbols) break;
   }
   const fallbacks = [];
   if (needEmoji) {
@@ -4549,6 +4558,26 @@ async function loadTypography(tree, loaders = {}) {
     const b = await registerFace("NotoSymbols", "NotoSansMath.woff", loaders.symbols ?? defaultSymbolFontLoader, tree, "symbols");
     if (b) fallbacks.push({ font: "NotoSymbols", bytes: b });
   }
+  if (cjk.size) {
+    let subsets = [];
+    try {
+      subsets = await (loaders.cjk ?? defaultCjkFontLoader)([...cjk]);
+    } catch {
+      subsets = [];
+    }
+    for (const sub of subsets) {
+      const font = `NotoCJK_${sub.name.replace(/[^\w]/g, "_")}`;
+      const b = await registerFace(font, `${sub.name}.woff`, async () => sub.bytes, tree, "Japanese / Chinese");
+      if (b) fallbacks.push({ font, bytes: b });
+    }
+    if (!subsets.length) {
+      tree.notices.push({
+        code: "glyph-substituted",
+        message: "The Japanese / Chinese font could not be loaded, so those characters were left out of the PDF (Word, EPUB and HTML keep them).",
+        remedy: "Check the internet connection (the font comes from cdn.jsdelivr.net) and export again."
+      });
+    }
+  }
   const faces = fallbacks.length ? await loadFaces(vfs, s.fontSize, fallbacks) : base;
   if (!s.hyphenate) return { faces };
   const split = await loadHyphenator(s.lang);
@@ -4571,6 +4600,37 @@ var urlLoader = (url) => async () => {
 };
 var defaultEmojiFontLoader = urlLoader(NOTO_EMOJI_URL);
 var defaultSymbolFontLoader = urlLoader(NOTO_SYMBOLS_URL);
+var NOTO_CJK_BASE = "https://cdn.jsdelivr.net/npm/@fontsource/noto-sans-jp@5.3.0";
+function parseUnicodeRangeCss(css) {
+  const out = [];
+  for (const block3 of css.split("@font-face").slice(1)) {
+    const file = /url\(\.\/files\/([\w-]+)\.woff\)/.exec(block3)?.[1];
+    const range = /unicode-range:\s*([^;]+);/.exec(block3)?.[1];
+    if (!file || !range) continue;
+    const ranges = range.split(",").map((r) => {
+      const [a, b] = r.trim().replace(/^U\+/i, "").split("-");
+      const lo = parseInt(a, 16);
+      return [lo, b ? parseInt(b, 16) : lo];
+    });
+    out.push({ file, ranges });
+  }
+  return out;
+}
+var defaultCjkFontLoader = async (cps) => {
+  if (typeof fetch !== "function" || cps.length === 0) return [];
+  const res = await fetch(`${NOTO_CJK_BASE}/400.css`);
+  if (!res.ok) return [];
+  const faces = parseUnicodeRangeCss(await res.text()).filter((f) => cps.some((cp) => f.ranges.some(([a, b]) => cp >= a && cp <= b)));
+  const out = [];
+  await Promise.all(
+    faces.map(async (f) => {
+      const r = await fetch(`${NOTO_CJK_BASE}/files/${f.file}.woff`);
+      if (r.ok) out.push({ name: f.file, bytes: new Uint8Array(await r.arrayBuffer()) });
+    })
+  );
+  return out;
+};
+var isCjk = (cp) => cp >= 11904 && cp <= 40959 || cp >= 44032 && cp <= 55215 || cp >= 63744 && cp <= 64255 || cp >= 65072 && cp <= 65103 || cp >= 65280 && cp <= 65519 || cp >= 131072 && cp <= 201551;
 var EMOJI_CP = /\p{Extended_Pictographic}|\p{Regional_Indicator}|\u20E3/u;
 function documentText(tree) {
   const s = tree.settings;
@@ -4737,6 +4797,70 @@ function removeThinkingContent(content) {
   return result.trim();
 }
 var PAGE_BREAK_HTML = `<div class="${PAGE_BREAK_CLASS}" role="separator" aria-label="${PAGE_BREAK_LABEL}" data-label="${PAGE_BREAK_LABEL}"></div>`;
+var CODE_ARTIFACT_LANGUAGE = {
+  html: "html",
+  react: "tsx",
+  svg: "svg",
+  mermaid: "mermaid",
+  diff: "diff",
+  iframe: "html",
+  code: ""
+};
+function parseTagAttributes(tag) {
+  const attrs = {};
+  for (const m of tag.matchAll(/([a-zA-Z_][\w-]*)\s*=\s*(?:"([^"]*)"|'([^']*)')/g)) {
+    attrs[m[1].toLowerCase()] = m[2] ?? m[3] ?? "";
+  }
+  return attrs;
+}
+function unwrapArtifactsToMarkdown(markdown) {
+  if (!/<artifact\b/i.test(markdown)) return markdown;
+  return markdown.split(/(^[ \t]*(?:`{3,}|~{3,})[\s\S]*?^[ \t]*(?:`{3,}|~{3,})[ \t]*$)/m).map(
+    (part, i) => i % 2 === 1 ? part : part.replace(/<artifact\b((?:[^>"']|"[^"]*"|'[^']*')*?)>([\s\S]*?)<\/artifact>/gi, (_m, tag, body) => {
+      const attrs = parseTagAttributes(tag);
+      const type = (attrs.type ?? "").toLowerCase();
+      const trimmed = body.replace(/^\n+|\n+$/g, "");
+      if (!trimmed.trim()) return "";
+      if (/^\s*(`{3,}|~{3,})/.test(trimmed)) return `
+
+${trimmed}
+
+`;
+      if (type in CODE_ARTIFACT_LANGUAGE) {
+        const lang = CODE_ARTIFACT_LANGUAGE[type] || attrs.language || attrs.lang || "";
+        return `
+
+\`\`\`\`${lang}
+${trimmed}
+\`\`\`\`
+
+`;
+      }
+      if (looksLikeJson(trimmed)) return `
+
+\`\`\`\`json
+${JSON.stringify(JSON.parse(trimmed), null, 2)}
+\`\`\`\`
+
+`;
+      return `
+
+${trimmed}
+
+`;
+    })
+  ).join("");
+}
+function looksLikeJson(body) {
+  const t = body.trim();
+  if (!/^[[{]/.test(t)) return false;
+  try {
+    JSON.parse(t);
+    return true;
+  } catch {
+    return false;
+  }
+}
 var LINK_URL = String.raw`((?:[^()\s]|\([^()\s]*\))+)(?:\s+"[^"]*")?`;
 var IMAGE_RE = new RegExp(String.raw`!\[([^\]]*)\]\(` + LINK_URL + String.raw`\)`, "g");
 var LINK_RE = new RegExp(String.raw`\[([^\]]+)\]\(` + LINK_URL + String.raw`\)`, "g");
@@ -5191,7 +5315,7 @@ function parseDocument(source, options = {}) {
   const settings = resolveSettings(fm.data, sink, options);
   for (const n of sink) notify(n);
   const body = removeThinkingContent(fm.body);
-  const lifted = extractMath(body);
+  const lifted = extractMath(unwrapArtifactsToMarkdown(body));
   const root = (0, import_mdast_util_from_markdown.fromMarkdown)(prepass(lifted.text), {
     extensions: [(0, import_micromark_extension_gfm.gfm)()],
     mdastExtensions: [(0, import_mdast_util_gfm.gfmFromMarkdown)()]