AutoPPT

中文说明

Generate polished, editable PowerPoint decks from a topic, an outline, or a template.

AutoPPT is a PowerPoint-native presentation generator built for people who need real .pptx files, not a web-only slide toy. It combines LLM planning, optional research, layout-aware rendering, and template support so the output stays editable in PowerPoint, Keynote, or Google Slides.

Why AutoPPT

• Native .pptx output that stays editable after generation
• Template-aware rendering for branded decks and repeatable formatting
• Layout planning before rendering, including richer slide types such as comparison and quote slides
• Slide-level workbench for regenerate and remix workflows
• Mock mode and offline mode for CI, demos, and local development without live API keys

See It In Action

Showcase Samples

These committed decks are the fastest way to judge current output quality.

Feature-oriented validation decks are documented in samples/README.md.

30-Second Quick Try

Install AutoPPT and generate a local deck without any external API key:

pip install autoppt
AUTOPPT_OFFLINE=1 autoppt --provider mock --topic "The Future of AI"

Run the web UI:

streamlit run autoppt/app.py

Core Capabilities

Installation

Install from PyPI:

pip install autoppt

Install from source:

git clone https://github.com/yeasy/autoppt.git
cd autoppt
pip install -e .

Install development dependencies:

pip install -e ".[dev]"

The dependency source of truth is pyproject.toml. requirements.txt is kept as a thin local install wrapper.

Configuration

Copy the example environment file and add the providers you want to use:

cp .env.example .env

Typical variables:

OPENAI_API_KEY=sk-...
GOOGLE_API_KEY=AIza...
ANTHROPIC_API_KEY=sk-ant-...

# Optional local OpenAI-compatible endpoint
OPENAI_API_BASE=http://localhost:1234/v1

# Optional Anthropic-compatible endpoint
ANTHROPIC_BASE_URL=https://api.anthropic.com

# Optional offline mode
AUTOPPT_OFFLINE=1

Usage

CLI

# Generate with default settings
autoppt --topic "The Future of AI"

# Let AutoPPT infer the best theme
autoppt --topic "Machine Learning Tutorial" --auto-style

# Preview the outline before rendering
autoppt --topic "Startup Pitch" --confirm-outline

# Generate outline only
autoppt --topic "Q1 Report" --outline-only

# Use a specific provider, style, and slide count
autoppt --topic "Planets in Solar System" --provider google --style dark --slides 8

# Use a custom template and render thumbnails
autoppt --topic "Q3 Report" --template templates/your-template.pptx --thumbnails

# Stay fully local for testing and CI
AUTOPPT_OFFLINE=1 autoppt --provider mock --topic "System Design Review"

Web App

streamlit run autoppt/app.py

Then open http://localhost:8501.

Docker

Run with Docker Compose:

docker-compose up -d
docker-compose logs -f

Run with Docker CLI:

docker build -t autoppt .
docker run --rm \
  -p 8501:8501 \
  --env-file .env \
  -v $(pwd)/output:/app/output \
  autoppt

Themes

Built-in themes:

minimalist, technology, nature, creative, corporate, academic, startup, dark, luxury, magazine, tech_gradient, ocean, sunset, chalkboard, blueprint, sketch, retro, neon

Use --style <theme> to force a specific look, or --auto-style to let AutoPPT choose.

Samples

Refresh all committed samples deterministically:

python scripts/generate_samples.py --category all --output-dir samples

Refresh one sample:

python scripts/generate_sample.py en_visual_showcase --output-dir samples

Refresh README preview assets:

python scripts/generate_readme_previews.py --output-dir docs/assets

For the full sample catalog, see samples/README.md.

Testing

Run the full test suite:

pip install -e ".[dev]"
pytest

Run coverage:

pytest --cov=autoppt --cov-report=term-missing

Run a focused test file:

pytest tests/test_renderer.py -v

Templates

Template guidance lives in templates/README.md.

Architecture

Architecture notes live in docs/architecture.md.

Contributing

1. Fork the repository.
2. Create a branch: git checkout -b feature/awesome
3. Install dev dependencies: pip install -e ".[dev]"
4. Run validation: pytest && python3 scripts/check_sensitive.py
5. Build and smoke test: python -m build && pip install dist/*.whl && autoppt --help
6. Commit your changes.
7. Push your branch and open a pull request.

License

Apache 2.0. See LICENSE.

Changelog

See CHANGELOG.md.