Repository navigation
Releases
v0.5.0 — Bring-your-own-CAD import
import-cad: use the plugin on an EXISTING CAD file
Until now the plugin assumed geometry-as-code. v0.5.0 adds the other door: hand it a CAD file and verify/simulate it.
- import_cad.py loads STEP/.stp (exact BREP via build123d/OCCT) or STL/OBJ/GLB/PLY meshes (trimesh), converts to canonical mm, and writes
out/imported.json(bbox, volume, centroid, mass, watertight, warnings) + a mesh — the artifacts the audits already read. - Then add-scored-audit / fea_continuum (gmsh→CalculiX) / viz / add-material all run on the imported geometry.
- Handles the traps (GUARDRAILS §9): units (STL is unitless, STEP often wrong — a 25.4x/1000x error silently corrupts everything), mass needs a cited density (materials.json), watertightness gates volume/FEA, and native formats (.sldprt/Fusion/CATIA) must be exported to STEP/STL first.
New import-cad skill; scaffold copies import_cad.py; README documents the two entry points.
Install: /plugin marketplace add JMMonte/agentic-digital-twin
v0.4.0 — Cited materials + realistic hero
Materials source of truth (no hallucinated properties)
There is no free machine-readable database of engineering design allowables (MMPDS/CMH-17/ASM/Granta are paid; Materials Project/PAULING are computed crystal data). So the plugin applies the same discipline as geometry:
- design/materials.json — every property carries a citation +
verifiedflag; audits read from it, never inline a number. - add-material skill + GUARDRAILS §8 — never hallucinate; composites need the specific prepreg + fiber-volume-fraction + layup (no 'generic CFRP'); use A/B-basis allowables not typicals; fluids via CoolProp.
Also in this release
- Realistic 3D hero (real NASA 777 + MRO satellite + Artec truck, rendered in PyVista with real scalar-field colormaps) — replaces the earlier primitive/illustrated banner.
- Badges at top of README; MIT license + model attribution.
Install: /plugin marketplace add JMMonte/agentic-digital-twin
v0.3.0 — Agentic Digital Twins, as Code
First public release under the new name (was sim-hardware-as-code).
Build engineering digital twins as code with AI agents: design-of-record JSON + scored audits (PASS/WARN/FAIL) + rerun visualization + scored tradeoff studies with sensitivity + tiered FEA (analytic → beam → gmsh+CalculiX) + the automation⇄agent manufacturability loop + a living guardrails checklist.
- MIT licensed
- Hero + Built-with badges
- Install:
/plugin marketplace add JMMonte/agentic-digital-twin