AI Driven Parametric CAD Automation

Architecture

A Large Language Model receives a structured part specification and emits Python or OpenSCAD source. The script is evaluated headlessly inside a sandboxed FreeCAD or OpenSCAD runtime, producing STEP and STL outputs ready for downstream CAM, BIM, or rendering pipelines.

Each generated part is fully parametric. Length, hole pitch, gear modulus, and material thickness are exposed as typed inputs, so the same generator can produce a four hole bracket or a twelve hole bracket without a single mouse click.

The result is a deterministic, version controlled, reviewable pipeline. Every fixture, mount, and gear lives in Git as source code. Pull requests review geometry the same way they review software.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
# freecad_bracket_generator.py
"""LLM-driven parametric bracket generator for FreeCAD.
Generates load-bearing mounting brackets from natural language specs.
"""
from __future__ import annotations

import FreeCAD as App
import Part
from dataclasses import dataclass

@dataclass(frozen=True)
class BracketSpec:
    length_mm: float
    height_mm: float
    thickness_mm: float
    hole_diameter_mm: float
    hole_count: int

def build_bracket(spec: BracketSpec) -> Part.Shape:
    base = Part.makeBox(spec.length_mm, spec.thickness_mm, spec.height_mm)
    pitch = spec.length_mm / (spec.hole_count + 1)

    for i in range(1, spec.hole_count + 1):
        cyl = Part.makeCylinder(spec.hole_diameter_mm / 2, spec.thickness_mm)
        cyl.translate(App.Vector(pitch * i, 0, spec.height_mm / 2))
        base = base.cut(cyl)

    return base

def export(shape: Part.Shape, path: str) -> None:
    shape.exportStep(path)

if __name__ == "__main__":
    spec = BracketSpec(120.0, 40.0, 6.0, 5.0, 4)
    export(build_bracket(spec), "./out/bracket.step")

Loading model...

● connected . pipeline://pointwake/aec