dsh-office-kit 0.4.5 → 0.4.6

package/CAPABILITY.md

@@ -138,10 +138,11 @@ never guessed.
 | PDF create / read / render / `metadata.py` / `page_ops.py` / `redact_region.py` | ok — CJK font applied, metadata written, merge→remove→rotate, region redaction removed 170 chars and 11 drawings, `page-1.png` produced |
 | Annotation geometry helpers | pymupdf rect conversion, python-pptx overlap scan, A1 text matching run against those files |
 | Forbidden-vocabulary grep over the delivered docs | empty |
+| macOS 26 arm64, DSH 0.2.1-rc.2, `DSH_OFFICE_PYTHON` = Homebrew 3.13 | full vitest suite with `DSH_OFFICE_REQUIRE_RENDERER=1` passes; docx/pptx/xlsx render and xlsx recalc report `engine: native`, also with the app's Electron binary as the recorded node; PDF create/stamp pick STHeiti |
 
 Not claimed (no evidence here): installer or self-heal timings, live GUI
-interaction (`/` menu, chip, sidebar preview, annotation overlay), and macOS
-renderer support.
+interaction (`/` menu, chip, sidebar preview, annotation overlay), and automatic
+Python discovery on macOS (see #23).
 
 ## Explicitly out of scope
 
@@ -152,6 +153,6 @@ renderer support.
   so); OCR and scanned-PDF text extraction.
 - PPTX animations, speaker notes, slide masters; XLSX pivots, macros, data
   validation, conditional formatting — no spec keys or verbs in the helper.
-- Formula evaluation without the DSH LibreOffice kit; macOS/Linux installation
-  was exercised in CI (Linux runs the WASM engine), but only Windows was checked
-  by hand.
+- Formula evaluation without the DSH LibreOffice kit; Linux installation was
+  exercised in CI (Linux runs the WASM engine); Windows and macOS were checked by
+  hand.

package/assets/skills/office-documents/scripts/internal_nav.py

@@ -5,7 +5,9 @@ Word's TOC field needs a refresh before its links work.  This script instead
 writes static hyperlinks that any renderer follows: bookmarks on every heading,
 a contents list at the top (with quick links and back links), and optional
 Top/Bottom anchors.  Because the entries are literal text there is nothing to
-update before the file is read.
+update before the file is read.  Rerunning on a file this script already
+processed strips the previous navigation first, so the output is rebuilt
+cleanly instead of stacking a second contents block.
 
 Exit codes: 0 success, 1 no headings, 2 bad input.
 """
@@ -59,6 +61,63 @@ def paragraph(style: str | None = None):
     return node
 
 
+def _remove_bookmark(node) -> None:
+    """Drop one bookmarkStart and its matching bookmarkEnd wherever they live."""
+    ident = node.get(kit.q("id"))
+    parent = node.getparent()
+    if parent is not None:
+        parent.remove(node)
+    if ident is None:
+        return
+    root = parent if parent is not None else node
+    while root.getparent() is not None:
+        root = root.getparent()
+    for end in list(root.iter(kit.q("bookmarkEnd"))):
+        if end.get(kit.q("id")) == ident:
+            end.getparent().remove(end)
+
+
+def strip_previous_nav(body) -> int:
+    """Remove navigation inserted by an earlier run, so reruns rebuild cleanly.
+
+    Without this, a second run stacks a second contents block, re-collects the
+    first block's title as a chapter, and duplicates the nav_top/nav_toc/
+    nav_bottom bookmark names, which makes every back link resolve to the stale
+    block. A paragraph is treated as generated when it links to a nav_ anchor or
+    carries a structural nav bookmark; a real heading only loses its nav_h
+    bookmarks. Returns the number of paragraphs removed.
+    """
+    removed = 0
+    for paragraph_node in list(body.iter(kit.q("p"))):
+        bookmarks = [
+            node
+            for node in paragraph_node.iter(kit.q("bookmarkStart"))
+            if (node.get(kit.q("name")) or "").startswith("nav_")
+        ]
+        links = [
+            node
+            for node in paragraph_node.iter(kit.q("hyperlink"))
+            if (node.get(kit.q("anchor")) or "").startswith("nav_")
+        ]
+        if not bookmarks and not links:
+            continue
+        generated = bool(links) or any(
+            node.get(kit.q("name")) in (TOP_BOOKMARK, BOTTOM_BOOKMARK, TOC_BOOKMARK)
+            for node in bookmarks
+        )
+        if generated:
+            for node in bookmarks:
+                _remove_bookmark(node)
+            parent = paragraph_node.getparent()
+            if parent is not None:
+                parent.remove(paragraph_node)
+                removed += 1
+        else:
+            for node in bookmarks:
+                _remove_bookmark(node)
+    return removed
+
+
 def main(argv: list[str] | None = None) -> int:
     parser = argparse.ArgumentParser(description="Add bookmark anchors and static internal links to headings.")
     parser.add_argument("input_docx", help="Input .docx")
@@ -77,6 +136,7 @@ def main(argv: list[str] | None = None) -> int:
         return 2
 
     body = doc.body()
+    stripped = strip_previous_nav(body)
     headings = []
     ident = doc.next_id()
     for paragraph_node in list(body.iter(kit.q("p"))):
@@ -147,7 +207,7 @@ def main(argv: list[str] | None = None) -> int:
         body.append(last)
 
     doc.save(args.out)
-    kit.emit({"ok": True, "output": str(Path(args.out)), "headings": headings, "entries": len(block)})
+    kit.emit({"ok": True, "output": str(Path(args.out)), "headings": headings, "entries": len(block), "stripped": stripped})
     return 0
 
 

package/assets/skills/office-pdf/scripts/pdf_fonts.py

@@ -25,9 +25,9 @@ CJK_FONT_CANDIDATES = {
         "/usr/share/fonts/opentype/noto/NotoSerifCJK-Regular.ttc",
     ],
     "darwin": [
-        "/System/Library/Fonts/PingFang.ttc",
         "/System/Library/Fonts/STHeiti Medium.ttc",
-        "/System/Library/Fonts/Hiragino Sans GB.ttc",
+        "/System/Library/Fonts/STHeiti Light.ttc",
+        "/System/Library/Fonts/Supplemental/Songti.ttc",
     ],
 }
 

package/dist/index.js

@@ -2,7 +2,7 @@ import { createRequire } from "node:module";
 import { execFile } from "node:child_process";
 import { existsSync, mkdirSync, readFileSync, renameSync, statSync, writeFileSync } from "node:fs";
 import { readFile } from "node:fs/promises";
-import { dirname, join } from "node:path";
+import { dirname, join, sep } from "node:path";
 import process from "node:process";
 import { promisify } from "node:util";
 import { BUNDLED_SKILL_RANK } from "@deepseek-ai/dsh-skill";
@@ -55,6 +55,16 @@ const OFFICE_FEATURES = [
 const PACKAGE_ROOT$1 = fileURLToPath(new URL("..", import.meta.url));
 const KIT_CLI = "@deepseek-ai/libreoffice-kit/cli";
 /**
+* Electron reads `app.asar` as a directory but nothing else can, and the helper
+* runs the recorded path from Python. The packaged app ships the kit unpacked.
+*/
+function outsideAsar(path) {
+	const packed = `${sep}app.asar${sep}`;
+	if (!path.includes(packed)) return path;
+	const unpacked = path.replace(packed, `${sep}app.asar.unpacked${sep}`);
+	return existsSync(unpacked) ? unpacked : path;
+}
+/**
 * Locate the LibreOffice kit CLI that DeepSeek Harness ships.
 *
 * DSH installs third-party plugins into a profile (`~/.dsh/profiles/<name>/node_modules`)
@@ -71,7 +81,7 @@ function resolveKitCli(entry = process.argv[1]) {
 		if (anchor == null) continue;
 		try {
 			const cli = createRequire(anchor).resolve(KIT_CLI);
-			if (existsSync(cli)) return cli;
+			if (existsSync(cli)) return outsideAsar(cli);
 		} catch {}
 	}
 	return null;

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "dsh-office-kit",
   "type": "module",
-  "version": "0.4.5",
+  "version": "0.4.6",
   "description": "Office documents, PDF, presentations, and spreadsheets for DeepSeek Harness",
   "author": "Hairy",
   "license": "MIT",

package/python/office_cli.py

@@ -184,6 +184,11 @@ def find_kit() -> tuple[str, str] | None:
     return node, str(cli)
 
 
+def _kit_env() -> dict[str, str]:
+    """Environment for a kit run: the recorded node may be the harness's Electron binary."""
+    return {**os.environ, "ELECTRON_RUN_AS_NODE": "1"}
+
+
 def _kit_call(arguments: list[str], timeout: int = 300) -> dict[str, Any]:
     """Run one kit command and return its JSON report, failing with the kit's own words."""
     kit = find_kit()
@@ -195,6 +200,7 @@ def _kit_call(arguments: list[str], timeout: int = 300) -> dict[str, Any]:
             [node, cli, *arguments],
             capture_output=True,
             text=True,
+            env=_kit_env(),
             # The kit always speaks UTF-8; the locale code page would mangle CJK paths.
             encoding="utf-8",
             errors="replace",
@@ -231,6 +237,7 @@ def kit_report() -> dict[str, Any]:
             [node, cli, "capabilities", "--json"],
             capture_output=True,
             text=True,
+            env=_kit_env(),
             encoding="utf-8",
             errors="replace",
             timeout=120,
@@ -763,8 +770,9 @@ _PDF_CJK_CANDIDATES = [
     r"C:\Windows\Fonts\msyh.ttf",
     r"C:\Windows\Fonts\simhei.ttf",
     r"C:\Windows\Fonts\simsun.ttc",
-    "/System/Library/Fonts/PingFang.ttc",
     "/System/Library/Fonts/STHeiti Light.ttc",
+    "/System/Library/Fonts/STHeiti Medium.ttc",
+    "/System/Library/Fonts/Supplemental/Songti.ttc",
     "/usr/share/fonts/truetype/winfonts/msyh.ttc",
     "/usr/share/fonts/truetype/winfonts/msyhl.ttc",
     "/usr/share/fonts/truetype/winfonts/simhei.ttf",