Skip to content

AI turns documents or topics into real, native PowerPoint decks—with native shapes, transitions and animations, data-backed charts and tables on demand, audio narration from speaker notes, and support for your own .pptx templates. · by Hugo He

fix(doc-to-md): keep notebook output images inside the per-output _files dir

hugohe3e4a4c084aa45
diff --git a/skills/ppt-master/scripts/docs/conversion.md b/skills/ppt-master/scripts/docs/conversion.md
index 5580cab5..a02498ce 100644
--- a/skills/ppt-master/scripts/docs/conversion.md
+++ b/skills/ppt-master/scripts/docs/conversion.md
@@ -132,7 +132,8 @@ Native path (no external binary required):
- `.docx` — via `mammoth`; text-only tables (with footnotes) and chart data become pipe Markdown, and OMML / Office Math equations (Word-native or MathType "Convert to Office Math") are rewritten to inline LaTeX. Classic MathType OLE objects carry no OMML and are kept only as their preview image.
- `.html` / `.htm` — via `markdownify` + `beautifulsoup4`
- `.epub` — via `ebooklib` + `markdownify`
-- `.ipynb` — via `nbconvert`
+- `.ipynb` — via `nbconvert`; code-cell output images and Markdown attachments
+ use `<output-stem>_files/` with relative Markdown references
Pandoc fallback (only if you need these):
- `.doc`, `.odt`, `.rtf`, `.tex`/`.latex`, `.rst`, `.org`, `.typ`
diff --git a/skills/ppt-master/scripts/source_to_md/doc_to_md.py b/skills/ppt-master/scripts/source_to_md/doc_to_md.py
index 56f20522..212ec031 100644
--- a/skills/ppt-master/scripts/source_to_md/doc_to_md.py
+++ b/skills/ppt-master/scripts/source_to_md/doc_to_md.py
@@ -1498,18 +1498,23 @@ def _convert_ipynb(input_file: Path, out_file: Path) -> str:
print("[ERROR] nbconvert not installed. Run: pip install nbconvert")
return ""
- # Pre-process cell-level markdown attachments: nbconvert leaves
- # `attachment:<name>` references intact but doesn't write the files.
- # Extract them into our outputs dict so FilesWriter picks them up.
+ # Extract attachments once with notebook-wide names; nbconvert's own
+ # attachment extractor can reuse names from different Markdown cells.
nb = nbformat.read(str(input_file), as_version=4)
extra_outputs: dict[str, bytes] = {}
rel_media_dir = f"{out_file.stem}_files"
+ output_stems = {
+ Path(output.metadata["filename"]).stem.casefold()
+ for cell in nb.cells
+ for output in cell.get("outputs", [])
+ if output.get("metadata", {}).get("filename")
+ }
attach_counter = 0
for cell in nb.cells:
if cell.cell_type != "markdown":
continue
- attachments = getattr(cell, "attachments", None) or {}
+ attachments = cell.pop("attachments", None) or {}
if not attachments:
continue
for att_name, mime_data in attachments.items():
@@ -1519,6 +1524,9 @@ def _convert_ipynb(input_file: Path, out_file: Path) -> str:
if ext == ".jpe":
ext = ".jpg"
filename = f"attachment_{attach_counter:03d}{ext}"
+ while Path(filename).stem.casefold() in output_stems:
+ attach_counter += 1
+ filename = f"attachment_{attach_counter:03d}{ext}"
out_path = f"{rel_media_dir}/{filename}"
try:
extra_outputs[out_path] = base64.b64decode(b64)
@@ -1530,7 +1538,9 @@ def _convert_ipynb(input_file: Path, out_file: Path) -> str:
cell.source = src
exporter = MarkdownExporter()
- body, resources = exporter.from_notebook_node(nb)
+ body, resources = exporter.from_notebook_node(
+ nb, resources={"output_files_dir": rel_media_dir},
+ )
# Merge attachment outputs with whatever nbconvert collected
resources.setdefault("outputs", {}).update(extra_outputs)
diff --git a/skills/ppt-master/scripts/tests/test_source_to_md_content_loss_docx.py b/skills/ppt-master/scripts/tests/test_source_to_md_content_loss_docx.py
index bd61f3b0..6eb3aed4 100644
--- a/skills/ppt-master/scripts/tests/test_source_to_md_content_loss_docx.py
+++ b/skills/ppt-master/scripts/tests/test_source_to_md_content_loss_docx.py
@@ -5,8 +5,11 @@ Usage: python3 -m unittest tests.test_source_to_md_content_loss_docx
Dependencies: python-docx, mammoth
"""
+import base64
import io
import json
+import re
+import subprocess
import sys
import tempfile
import unittest
@@ -26,6 +29,128 @@ sys.path.insert(0, str(SCRIPTS_DIR / "source_to_md"))
import doc_to_md # noqa: E402
+class NotebookContentLossTests(unittest.TestCase):
+ def test_attachments_do_not_overwrite_named_outputs_or_each_other(self) -> None:
+ try:
+ import nbconvert # noqa: F401
+ import nbformat
+ except ImportError:
+ self.skipTest("optional nbconvert/nbformat unavailable")
+ from PIL import Image
+
+ for filename in ("attachment_001", "attachment_001.png"):
+ with self.subTest(filename=filename), tempfile.TemporaryDirectory() as tmp:
+ root = Path(tmp)
+ expected = []
+ encoded = []
+ for color in ("red", "blue", "green"):
+ buffer = io.BytesIO()
+ Image.new("RGB", (8, 8), color).save(buffer, format="PNG")
+ expected.append(buffer.getvalue())
+ encoded.append(base64.b64encode(buffer.getvalue()).decode("ascii"))
+ notebook = nbformat.v4.new_notebook(cells=[
+ nbformat.v4.new_code_cell("# Named output", outputs=[
+ nbformat.v4.new_output(
+ "display_data", data={"image/png": encoded[0]},
+ metadata={"filename": filename},
+ ),
+ ]),
+ *[
+ nbformat.v4.new_markdown_cell(
+ "![attachment](attachment:picture.png)",
+ attachments={"picture.png": {"image/png": data}},
+ )
+ for data in encoded[1:]
+ ],
+ ])
+ source = root / "book.ipynb"
+ nbformat.write(notebook, source)
+ with redirect_stdout(io.StringIO()):
+ markdown = doc_to_md.convert_to_markdown(str(source))
+ refs = re.findall(r"!\[[^\]]*\]\(([^)]+)\)", markdown)
+ self.assertEqual(len(refs), 3)
+ self.assertEqual(len(set(refs)), 3)
+ for ref, raw in zip(refs, expected):
+ self.assertEqual(Path(ref).parent, Path("book_files"))
+ self.assertEqual((root / ref).read_bytes(), raw)
+ manifest = json.loads((root / "book_files/image_manifest.json").read_text(encoding="utf-8"))
+ self.assertEqual({row["filename"] for row in manifest}, {Path(ref).name for ref in refs})
+ self.assertTrue(all(len(row["occurrences"]) == 1 for row in manifest))
+
+ def test_batch_images_and_attachments_remain_independent_after_import(self) -> None:
+ try:
+ import nbconvert # noqa: F401
+ import nbformat
+ except ImportError:
+ self.skipTest("optional nbconvert/nbformat unavailable")
+ from PIL import Image
+ from project_management.cli import ProjectManager
+
+ for same_stem in (False, True):
+ with self.subTest(same_stem=same_stem), tempfile.TemporaryDirectory() as tmp:
+ root = Path(tmp)
+ sources = []
+ expected = []
+ for index, color in enumerate(("red", "blue")):
+ source_dir = root / str(index)
+ source_dir.mkdir()
+ source = source_dir / ("book.ipynb" if same_stem else f"book{index}.ipynb")
+ buffer = io.BytesIO()
+ Image.new("RGB", (8, 8), color).save(buffer, format="PNG")
+ raw = buffer.getvalue()
+ encoded = base64.b64encode(raw).decode("ascii")
+ output = nbformat.v4.new_output(
+ "display_data", data={"image/png": encoded},
+ metadata={"filename": "figure"} if same_stem else {},
+ )
+ notebook = nbformat.v4.new_notebook(cells=[
+ nbformat.v4.new_code_cell(f"# Notebook {index}", outputs=[output]),
+ nbformat.v4.new_markdown_cell(
+ "![attachment](attachment:picture.png)",
+ attachments={"picture.png": {"image/png": encoded}},
+ ),
+ ])
+ nbformat.write(notebook, source)
+ sources.append(source)
+ expected.append(raw)
+ output_dir = root / "converted"
+ result = subprocess.run(
+ [sys.executable, str(SCRIPTS_DIR / "source_to_md/doc_to_md.py"),
+ *map(str, sources), "-o", str(output_dir)],
+ capture_output=True, text=True, encoding="utf-8", check=False,
+ )
+ self.assertEqual(result.returncode, 0, result.stdout + result.stderr)
+ markdown_paths = sorted(output_dir.glob("*.md"))
+ self.assertEqual(len(markdown_paths), 2)
+ self.assertEqual(list(output_dir.glob("*.png")), [])
+ for markdown, raw in zip(markdown_paths, expected):
+ refs = re.findall(r"!\[[^\]]*\]\(([^)]+)\)", markdown.read_text(encoding="utf-8"))
+ self.assertEqual(len(refs), 2)
+ for ref in refs:
+ self.assertEqual(Path(ref).parent, Path(f"{markdown.stem}_files"))
+ self.assertEqual((output_dir / ref).read_bytes(), raw)
+ profile = json.loads(markdown.with_suffix(".conversion_profile.json").read_text())
+ self.assertEqual(profile["outputs"]["image_count"], 2)
+ manifest = json.loads((output_dir / profile["outputs"]["image_manifest"]).read_text())
+ self.assertEqual(len(manifest), 2)
+ manager = ProjectManager(base_dir=root / "projects")
+ with redirect_stdout(io.StringIO()):
+ project = Path(manager.init_project("notebooks", quick_generate=True))
+ summary = manager.import_sources(str(project), list(map(str, markdown_paths)), copy=True)
+ self.assertEqual(summary["skipped"], [])
+ self.assertEqual(len(summary["assets"]), 2)
+ for path, raw in zip(summary["markdown"], expected):
+ markdown = Path(path)
+ refs = re.findall(r"!\[[^\]]*\]\(([^)]+)\)", markdown.read_text(encoding="utf-8"))
+ for ref in refs:
+ self.assertEqual((markdown.parent / ref).read_bytes(), raw)
+ manifest = json.loads((project / "images/image_manifest.json").read_text())
+ self.assertEqual({row["source_namespace"] for row in manifest},
+ {path.stem for path in markdown_paths})
+ self.assertEqual({(project / "images" / row["filename"]).read_bytes() for row in manifest},
+ set(expected))
+
+
class DocxContentLossTests(unittest.TestCase):
def _convert(self, document, patch_archive=None):
with tempfile.TemporaryDirectory() as tmp: