Slide Writing

Slide Writing is a self-built, zero-dependency PowerPoint engine for automated consulting-deck creation. It reads .pptx templates as Office Open XML packages, extracts their design model, and writes new slides that reuse the original masters, layouts, theme colors, fonts, and placeholders.

The goal is simple: turn structured content into editable, CI-consistent PowerPoint slides without python-pptx, without pip install, and without opening PowerPoint.

What It Can Do

• Parse PPTX packages, relationships, content types, themes, slide masters, layouts, placeholders, and existing slide metadata.
• Create new slides against existing layouts so backgrounds, logos, placeholders, fonts, and colors are inherited from the source template.
• Build native editable PowerPoint content: textboxes, shapes, lines, images, and formatted tables.
• Render consulting components from code: agenda, KPI tiles, Harvey balls, waterfall charts, bar charts, timelines, comparison matrices, traffic lights, callouts, and section numbers.
• Generate complete decks from a declarative JSON spec.
• Run entirely on the Python standard library.

PDF Import (Web Tool)

The browser tool (docs/index.html) accepts PDF uploads in addition to .pptx. Each PDF page is analysed client-side with a vendored copy of pdf.js (docs/vendor/, no server, no API) and rebuilt into native, editable PowerPoint slides — then shown in the same preview and exported as .pptx exactly like an uploaded deck:

• Text → native editable text boxes (position + font size).
• Vectors → native shapes (prstGeom rectangles, custGeom freeform paths).
• Tables → native PowerPoint tables (a:tbl), reconstructed from the page's ruling lines and cell text.
• Charts → real, editable chart objects (c:barChart / c:lineChart) with a reconstructed category/value data table (bar height ↔ axis-tick scale).
• Images → embedded pictures; any page that can't be safely reconstructed falls back to a full-page raster image so fidelity is never lost.

The PDF is synthesised into the same in-memory OOXML package an uploaded .pptx produces, so analysis, SVG preview, and .pptx export all reuse the existing pipeline.

Repository Status

This is a V1 engineering baseline:

• slidewriting/ contains the PPTX parser, model, builder, writer, component library, JSON spec renderer, and CLI.
• tests/ contains unit and end-to-end tests using unittest.
• examples/consulting_deck.json is a ready-to-run consulting demo spec.
• docs/ZIELBILD.md describes the product vision in German.
• ARCHITECTURE.md documents the internal module contracts and OOXML rules.

Quick Start

Run the package directly from the repository root:

python3 -m slidewriting bootstrap --out out/template.pptx
python3 -m slidewriting inspect out/template.pptx
python3 -m slidewriting validate examples/consulting_deck.json
python3 -m slidewriting demo --out out/showcase.pptx
python3 -m slidewriting generate --spec examples/consulting_deck.json --out out/demo.pptx

After installation, use the branded CLI command:

slide-writing generate --spec examples/consulting_deck.json --out out/demo.pptx

Use an existing corporate PowerPoint master as the design source:

python3 -m slidewriting generate \
  --template path/to/company-master.pptx \
  --spec examples/consulting_deck.json \
  --out out/company-demo.pptx

Python API

from slidewriting import Deck, build_from_spec

deck = Deck.create(accent="206EFB")
slide = deck.add_slide("obj")
slide.title("Market analysis 2026")
slide.placeholder("body").bullets([
    "Demand is shifting toward automated delivery.",
    {"text": "Template fidelity remains non-negotiable.", "level": 1},
])
deck.save("out/manual-api-demo.pptx")

spec_deck = build_from_spec({
    "title": "Generated strategy deck",
    "slides": [
        {"type": "title", "title": "Strategy 2026"},
        {"type": "agenda", "items": ["Context", "Options", "Roadmap"]},
    ],
})
spec_deck.save("out/spec-demo.pptx")

The consulting components are also available as an explicit module API:

from slidewriting.components import add_agenda, add_kpi_row, add_waterfall

JSON Spec Slide Types

Slide Writing currently supports:

• title
• section
• bullets
• agenda
• kpi
• table
• waterfall
• bars
• timeline
• comparison
• blank

Unknown JSON keys are ignored for forward compatibility. Invalid required fields are reported with slide indexes by slide-writing validate.

Design Principles

• Standard library only.
• Native OOXML package reading and writing.
• Theme-first output: generated content uses PowerPoint theme slots wherever possible.
• Editable output: components are built from native shapes and text, not image screenshots.
• Deterministic output for reliable testing and reproducible automation.

Test

python3 -m unittest discover -s tests -v

Important Limitations

• V1 reads templates and appends new slides; it does not edit existing slide content in place.
• Shape-based charts are implemented; native PowerPoint chart parts are a planned later extension.
• Visual rendering is delegated to PowerPoint or compatible office software.