Turn any field’s evolution history into an interactive 3D spatiotemporal map: world-map base, vertical time axis, nodes floating by real lat/long + year, hover-to-front detail cards. Use for “development history / evolution map / timeline map” requests. Input JSON to generate self-contained HTML.
3D Industry-Evolution Spatiotemporal Graph Generator
Turn "the key milestones of a domain (AI, automotive, semiconductors, pharma, internet…)" into a rotatable, zoomable, hover-for-details 3D web page:
world map as the base + a vertical time axis (year → height) + nodes placed by real lat/lon + person avatars / brand logos / event tetrahedra + hover-to-front with detail cards.
The full logic was battle-tested in an "AI Industry History" project (hover cards hugging the node, oldest node raised off the ground, hover-to-front without being occluded). This skill parameterizes and generalizes that work.
When to use
The user wants "an evolution graph / timeline map / industry-layout visualization of domain X".
The user provides a list with year / location (lat/lon or city) / type and wants an interactive web page.
Unlike wb-finance-skill or neodata-financial-search: this skill focuses on geo + time dual-axis visualization, not data fetching.
Workflow
Collect data: have the user provide a milestone list (or you research/complete it). Each item needs at least: year, type, title, loc, lat, lon, desc.
type is one of three: person (people/entrepreneurs, shows avatar), product (product/tech, shows logo or brand-colored abbreviation plate), event (event/conference, shows tetrahedron).
(Optional) Images: for people, portraits: { pid: "image URL or local path" }; for brands, logos: { key: "image URL or path" } and map title to { logo: key, color:"#hex", abbr:"abbr" } in brandMap.
Optional: people degrade to an "initials disc", products degrade to a "brand color + abbreviation plate".
(Optional) Map: map takes a path or URL to an equirectangular/Mercator world-map PNG (inlined at generation). Leave empty for a solid-color base (still works, just no geographic basemap).
Generate:
bash
# generate.py auto-locates template.html in its own directory via __file__, no need to cd
# Replace SKILL_DIR with the actual install path: user-level ~/.workbuddy/skills/industry-evolution-3d, or project-level <workspace>/.workbuddy/skills/industry-evolution-3d
SKILL_DIR=~/.workbuddy/skills/industry-evolution-3d
python3 "$SKILL_DIR/generate.py" "$SKILL_DIR/examples/ai_history_sample.json" /tmp/demo.html
Hover-to-front: setNodeActive sets the hovered node's renderOrder=1000 + depthTest=false, so it stays clickable and visible even when occluded by collapsed nodes.
Hover card: placeCard uses nodeScreenCenter(node) to anchor the card +14px down-right of the current node, and only does viewport clamping — this is the user-approved baseline; do NOT add complex "avoid other nodes / keep fully visible" logic that pushes the card away (we hit that pitfall, see lessons).
Height above ground: the oldest year must have AXIS_BASE>0, otherwise the tetrahedron/sprite sinks into the ground plane (a point the user explicitly wanted fixed).
Image inlining: at generation, avatars/logos/map are base64-inlined, so the page opens offline; SVG is inlined as base64 directly, bitmaps are cropped to a circle (avatar) / shrunk (logo/map) via PIL.
Boundaries / notes
The geographic basemap defaults to a world map; for "non-geographic" domains (e.g. a pure timeline), leave map empty or swap in a custom basemap (flowchart/topology needs a different template).
Too many nodes (>60) get crowded; consider splitting by sub-period or tuning MAP_W/MAP_D.
Deploying to CloudStudio with the same sandbox id keeps the URL stable (see memU/deploy notes).
See references/lessons.md for the pitfalls we hit and the final solutions while getting this graph right.