Skip to content

debim v1.1.0 — Urban Scale GIS Ingestion & ISO 19650 IFC4.3 Civil Infrastructure

Choose a tag to compare

@PRIDA-TAKON PRIDA-TAKON released this 21 Oct 22:22
· 27 commits to main since this release

debim v1.1.0 — Urban Scale GIS Ingestion & ISO 19650 IFC4.3 Civil Infrastructure

We are proud to release debim v1.1.0, expanding the declarative Building-as-Code engine from standalone buildings to macro-urban scale digital twins and civil infrastructure compliant with buildingSMART ISO 19650 IFC4.3!


🏙️ What's New in v1.1.0

1. Open GIS to .debim Ingestion Pipeline (debim urban ingest)

  • OSM Overpass Ingestion: Ingest entire city bounding boxes directly from OpenStreetMap Overpass API into native .debim AST.
  • GeoJSON Feature Collection: Automatic translation of 2D polygon footprints into 3D extruded IfcBuildingElementProxy and 2D linestrings into swept IfcRoad entities.
  • Pure Python WGS84-to-UTM Projector: Lossless translation between spherical GPS coordinates (WGS84) and metric Cartesian coordinates (EPSG:32647) with zero external C-dependencies.

2. ISO 19650 & buildingSMART IFC4.3 Infrastructure Entities

  • Real-World Geo-Referencing: Automatic compilation of IfcProjectedCRS and IfcMapConversion anchors ensuring geospatial BIM alignment in GIS viewers, BlenderBIM, and Civil 3D.
  • Linear Infrastructure Spatial Hierarchy: Support for IfcFacility, IfcRoad, and IfcFacilityPart (ROADWAY, SIDEWALK, SHOULDER).
  • Corridor Sweeper: Geometry generation using IfcSectionedSolidHorizontal swept solids along road alignments.
  • Civil & Utility Entities: Native support for IfcPavement (layered asphalt & sub-base), IfcKerb (precast concrete curbs), and IfcPipeSegment (drainage culverts).

3. High-Performance Kaggle Urban Benchmark Lab

  • Stress Tested at 10,000 Elements: Proven end-to-end ingestion, resolving, and compilation throughput on Kaggle (pridatakon/debim-urban-gis-benchmark).
  • Storage Efficiency: Proven 1.7x - 1.8x storage compactness of .debim compared to verbose ISO 10303-21 STEP IFC4.3 files.
  • Fast Iterations: Ingestion & resolving in seconds, enabling fast design iterations before emitting final IFC artifacts.

4. Autonomous Agent Workflow (Jules Watchdog)

  • Coordinated multi-stage feature delivery with Jules AI and jules_watchdog orchestrator, passing 100% of unit tests and regression suites across all milestones.

📦 Installation & Upgrade

pip install -U debim

Verify version:

debim --version
# debim, version 1.1.0