Agent-driven PPTX toolkit. Coding agents use the installed skill set, plugin tools, and predefined design context to plan and generate coordinate-explicit PowerPoint decks.
| Package | Purpose |
|---|---|
| pptify-plugin | Source ingestion, design context, image/SVG helpers, PPTX extraction, collision audit |
| pptify-core | Agent Skills and workflow prompts |
| pptify-design | Predefined design profiles and template context |
| pptify-cli | Installs the above into ./.agent/ |
See ARCHITECTURE.md for details.
uv sync # base dependencies
uv sync --extra plugins # add source conversion, image search, vector tracinguv run python pptify-cli install # → ./.agent/
uv run python pptify-cli install --home ~ # → ~/.agent/
uv run python pptify-cli install --home tests/pptify-install-test # → tests/pptify-install-test/.agent/See pptify-cli/README.md for uninstall, help, and --dry-run.
When OpenAI or Azure OpenAI image generation is needed, create a local .env from .env.template and fill the required provider values there. The image helper loads .env automatically; .env is git-ignored and must not be committed.
Copy-Item .env.template .envuv run python pptify-plugin/design/design_context_catalog.py --list --pretty
uv run python pptify-plugin/audit/audit.py tests/clustered_deck.json --json
uv run python pptify-plugin/images/iconfy_search.py --query governance --collection fluent --color 0078D4 --pretty
uv run python pptify-plugin/images/text_prompt_to_infographic.py --provider azure-openai --azure-endpoint "<endpoint>" --model "gpt-image-2" --prompt "..." --output-path out.png --prettyExtraction helpers (extraction/pptx_extractor.py, extraction/pptx_style_master.py) are import APIs — load them with importlib.util.spec_from_file_location(...).
The MiniLM ONNX model and tokenizer are not committed. Restore from the repository root:
.\pptify-plugin\download-external-assets.ps1