Hi Graphisoft — a custom rundown ahead of our chat next week.
We ported the engines to WebGPU.
Vienna, 800m. Wind, solar hours, thermal comfort — recomputed as buildings move. First results.
- 7 fps
- Wind (CFD)
- 5 fps
- Solar hours
- 1 fps
- UTCI
- 128 MB
- Model bundle
Up to ~1 GB VRAM peak during compute. Numbers from a Vienna 800m bench on a recent integrated GPU. The bundle is curated — wind retrained on a smaller architecture, physics ported to WGSL. Heavier models and regulatory-grade pipelines stay cloud-side, same wire format.
Ways forward.
TypeScript SDK
The swap path. Same wire format as the Python SDK — realtime and cloud share one client surface.
Live socket session
Warm GPU container per session, 1–3 clients per instance. For scenes that don't fit on a laptop. Serverless, scales to zero.
Cloud-side full fidelity
Current cloud API. Heavier physics and regulatory-grade pipelines, async route.
Today's SDK — try it.
Live SDK endpoints, eight engines, automatic tiling. Python today, TypeScript shipping shortly.
In the existing SDK.
Automatic tiling
Run on any city polygon. Tile splitting, parallel jobs, result merging — one call.
Weather built-in
Nearest weather station lookup is part of the SDK. You bring geometry, we bring climate.
Eight analyses, one client
Wind, PWC, daylight, sun, solar, sky-view, UTCI, thermal stats. Async-orchestrated, all typed.
Recap — things in the pipeline
- TypeScript SDK — same surface, same wire format. Browser, Node 18+, Cloudflare Workers.
- Energy & embodied carbon — building on the indoor analysis layer (below).
Build with your AI agents.
Drop-in skills for Claude, Cursor, and Codex, plus a Jupyter cookbook for every analysis. Same recipes for humans and agents.
- Agent-friendly SDK with typed responses
- Notebooks for every analysis
- Integrations that took a sprint now ship in an afternoon
Actionable insights, not raw simulations.
Curated workflows take a design question and return a decision-ready answer. Skip the simulation setup, weather lookup, and synthesis — climate experts have wired the pipeline.
- Composite maps, hotspot zones, written recommendations
- End-to-end curated workflow — not a black box
- Customisable per client and per topic
Indoor — daylight.
Drop in your building model, get a daylight heatmap per floor. Same SDK as the rest.
- Per-floor daylight maps from your building model
- Schema-flexible — bring your own geometry format
- Same SDK — drops into the same workflow as the rest
Pluvial flood risk.
Fast surface-water flood analysis for any site. Street-level risk maps in minutes, not weeks.
- Decision-ready outputs — depth maps, flow paths, building exposure
- Aligned with German KAnG climate-adaptation reporting
- Same SDK — drops into the same workflow as the rest
