NekoPaw Render

Use this skill when the user wants to turn content into display-ready bitmap files for a NekoPaw device.

Environment and setup:

• Install runtime packages with pip install -r skill/render/requirements.txt
• Install the browser binary with python -m playwright install chromium
• Install flowText / pretext JavaScript dependencies with npm install

Agent workflow:

1. Decide whether the input is Markdown or scene json.
2. Render a preview PNG first so the layout can be inspected.
3. Convert that preview into a raw 1bpp bitmap file.
4. If the user wants to show it on a device, hand the bitmap file to python skill/bridge_cli.py display bitmap --input ...

Examples:

python skill/render_cli.py markdown --input note.md --preview out/note.png --bitmap out/note.bin

python skill/render_cli.py scene --input scene.json --preview out/scene.png --bitmap out/scene.bin

python skill/render_cli.py scene --input skill/render/examples/scene_news_card.json --preview out/news_card.png --bitmap out/news_card.bin

python skill/render_cli.py scene --input skill/render/examples/scene_poster_card.json --preview out/poster_card.png --bitmap out/poster_card.bin

python skill/render_cli.py scene --input skill/render/examples/scene_pet_companion.json --preview out/pet_companion.png --bitmap out/pet_companion.bin --bw-preview out/pet_companion_bw.png

python skill/render_cli.py bitmap --input out/scene.png --output out/scene.bin --dither floyd-steinberg

python skill/render_cli.py confirm-assets \
  --title "Smart Home" \
  --body "Turn on the fan?" \
  --output-dir out/confirm

Notes:

• markdown supports regular text flow and the first image block as a figure sidebar.
• scene is the free-layout path for absolute-positioned text and image blocks.
• scene supports flowText for boxed key text that uses pretext, avoids wrapping image rectangles, and reports overflow through layoutReport.
• scene now supports stable z stacking within each visual layer, plus image anchor placement for cover / contain / fill.
• On the current 296x128 low-res black-and-white target, scene text blocks now render into the final bitmap on the target pixel grid instead of only relying on browser text downscaling.
• On that same low-res target, text blocks are rasterized in their own block canvas before being composited back, so block clipping, centering, and invert badges stay more stable than the old whole-page white overlay path.
• The low-res text path still accepts NEKOPAW_RENDER_FONT_REGULAR and NEKOPAW_RENDER_FONT_BOLD as font overrides; without them, ASCII-heavy blocks prefer Verdana / Tahoma, while blocks with Chinese or other non-ASCII text prefer Microsoft YaHei / Yu Gothic.
• See docs/SCENE_JSON.md for the current scene schema and skill/render/examples/ for ready-to-render examples.
• confirm-assets emits pending / confirmed / cancelled / timeout preview and bitmap files together.
• Those confirm-assets outputs can be sent to the device with python skill/bridge_cli.py display confirm create --assets-dir out/confirm.
• Successful commands print JSON to stdout and exit 0. Local validation, dependency, file, or rendering errors print local JSON and exit 2.