Agent-driven PPTX toolkit as a VS Code plugin. Agents use bundled skills, tools, and design context to generate coordinate-explicit PowerPoint decks.
pptify is the plugin root. The manifest is pptify/.github/plugin/plugin.json, and all exposed agents, skills, scripts, and resources are bundled there.
| Package | Purpose |
|---|---|
| pptify/.github/plugin/plugin.json | VS Code/Copilot plugin metadata declaring exposed component folders |
| pptify/agents | Custom agents discovered by the plugin |
| pptify/skills | Agent Skills discovered by the plugin |
| pptify/skills/pptify-tooling/scripts | Bundled support scripts called by skills and agents |
| pptify/skills/pptify-tooling/resources/design | Predefined design profiles and template context |
The plugin manifest intentionally declares the supported Copilot component folders, pptify/skills/ and pptify/agents/. The end-to-end deck-generation workflow is consolidated into the custom agent; scripts and resources are bundled inside pptify/skills/pptify-tooling/ as support assets referenced by declared components.
See ARCHITECTURE.md for details.
pptify-kit-stress-demo.pptx (densed and overcomplicated layout for stress testing)
uv sync # base dependencies
uv sync --extra plugins # add source conversion, image search, vector tracingWhen OpenAI or Azure OpenAI image generation is needed, create a local .env from pptify/skills/pptify-tooling/resources/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 pptify/skills/pptify-tooling/resources/env.template .envuv run python pptify/skills/pptify-tooling/scripts/design/design_context_catalog.py --list --pretty
uv run python pptify/skills/pptify-tooling/scripts/audit/audit.py deck-spec.json --json
uv run python pptify/skills/pptify-tooling/scripts/images/iconfy_search.py --query governance --collection fluent --color 0078D4 --pretty
uv run python pptify/skills/pptify-tooling/scripts/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\skills\pptify-tooling\scripts\download-external-assets.ps1