About
How the Pipeline Works
Smart Building Design+ is a six-agent AI pipeline that turns a map coordinate into a complete Revit BIM model. Each agent owns one transformation step and communicates via a central orchestrator.
Map → Image
Google Maps APIEasyThe Google Street View Static API captures 3–4 frames of the building at optimal headings, plus a satellite overhead image for footprint reference.
Image → 3D Model
Meshy AIEasy–MediumMeshy's multi-view Image-to-3D API accepts up to four angles and produces a GLB mesh with PBR materials. Quality varies with occlusion and foliage.
3D Model → Semantic
Claude VisionEasy–MediumClaude analyses the Street View frames and extracts structured JSON: storey count, floor heights, window positions, facade material, and roof type.
3D Model → 2D CAD
trimesh + SVGHardThe mesh is sliced at inferred floor heights to produce orthographic plan and elevation SVGs plus a DXF for Revit import. Falls back to schematic if mesh slicing fails.
CAD → Revit
Claude + Revit MCPHardA Claude agent drives a C# MCP server inside Revit. It creates levels, floor plates, walls, and openings from the geometry. A self-correction loop handles Revit API errors.