Profile My skills Saved Collections Edit profile Submit a skill

Geometry Generator

Generate parametric bioinspired ribbed membrane STL geometry via LLM-guided design. Takes a spec JSON (from StructureAnalyst/PropertyPredictor upstream artifacts), calls the LLM with a structured CAD prompt to produce design parameters, then builds a triangulated STL mesh in Python. Returns artifact JSON with stl_path, mesh stats, and the prompt used.

lamm-mit d127c12 4 files · 40.0 KB Updated
AI & ML

File contents

Geometry Generator

Generates parametric bioinspired hierarchical ribbed membrane STL geometry.

All design parameters flow from upstream artifacts (StructureAnalyst motifs + PropertyPredictor targets) — no hardcoded values.

Usage

# From upstream artifact spec file
python3 {baseDir}/scripts/stl_generator.py \
  --spec '{"rib_spacing_mm":2.5,"thickness_mm":0.4,"aspect_ratio":3.0,"num_scales":2}' \
  --output /tmp/membrane.stl

# From upstream artifact file
python3 {baseDir}/scripts/stl_generator.py \
  --spec-file /path/to/structural_motifs.json \
  --output /tmp/membrane.stl

Output JSON

{
  "stl_path": "/path/to/membrane.stl",
  "num_vertices": 1234,
  "num_faces": 2468,
  "bounding_box_mm": {"x": 20.0, "y": 20.0, "z": 1.2},
  "primary_rib_count": 8,
  "secondary_rib_count": 16,
  "prompt_used": "...",
  "design_params": {...}
}

STL Prompt

The LLM is called with the canonical bioinspired ribbed membrane prompt (see PROMPT.md). It returns structured design parameters as JSON. Python then constructs the mesh from those parameters.

lamm-mit/scienceclaw/tree/main/skills/geometry-generator commit d127c1229a

Browse category

Frequently asked questions

npx skillmds@latest add lamm-mit/geometry-generator