Text-to-CAD Python Requirements: Complete Dependency Guide for Skills-Based CAD Automation

The Text-to-CAD repository manages Python dependencies through isolated per-skill requirements.txt files plus a top-level requirements-dev.txt, with all skills sharing a core cadpy library installed in editable mode from scripts/packages/cadpy.

The text-to-CAD repository from earthtojake/text-to-cad implements CAD generation as a collection of modular agent skills. Understanding the Python requirements structure is essential for deploying individual skills or extending the framework. Each skill maintains its own dependency footprint while sharing a common geometry core.

How Python Dependencies Are Organized

The repository follows a skill-isolation pattern. Every capability—whether generating STEP files, building robot URDFs, or previewing models—lives in its own directory under skills/<skill-name>/ with a dedicated requirements.txt file.

Per-Skill Requirements Breakdown

Skill	Purpose	Requirements File	Key Packages
CAD	Solid model generation (STEP, STL, 3MF, GLB)	skills/cad/requirements.txt	playwright, editable cadpy
URDF	Robot description files	skills/urdf/requirements.txt	cadpy
SDF	Simulation worlds	skills/sdf/requirements.txt	cadpy
SRDF	MoveIt2 planning metadata	skills/srdf/requirements.txt	cadpy
DXF	2-D drawing export	skills/dxf/requirements.txt	cadpy
CAD-Viewer	Browser-based preview	skills/cad-viewer/requirements.txt	playwright, cadpy
Development	Linting, testing, type-checking	requirements-dev.txt	pytest, black, ruff, mypy

The Shared cadpy Core

All skills depend on a local Python package located at scripts/packages/cadpy. This is installed via an editable link in each skill's requirements.txt:

--editable ./scripts/packages/cadpy

The editable installation ensures that changes to the core geometry utilities propagate instantly across every skill without reinstallation.

Installing Text-to-CAD Python Requirements

Automated Installation via Skills CLI

The recommended path uses the Skills CLI, which handles dependency resolution automatically:

# Install the CAD skill and its Python dependencies

npx skills install earthtojake/text-to-cad#cad

This command:

1. Clones or downloads the repository artifact
2. Executes pip install -r skills/cad/requirements.txt
3. Installs cadpy in editable mode from the local path

Manual Installation for Development

For custom development or experimentation with the core library:

git clone https://github.com/earthtojake/text-to-cad.git
cd text-to-cad

# Install the core geometry library in editable mode

pip install -e ./scripts/packages/cadpy

# Install runtime dependencies for a specific skill

pip install -r skills/cad/requirements.txt

Development-Only Dependencies

To set up the full development environment—including linting, testing, and type-checking tools:

pip install -r requirements-dev.txt

These packages are never installed during skill-only deployments, preserving minimal runtime footprints.

Why Playwright Is Required

Two skills—CAD and CAD-Viewer—depend on playwright. This package provides a headless Chromium environment used to:

• Render implicit-CAD scenes for snapshot generation
• Capture preview images for the viewer UI
• Execute browser-based model conversion pipelines

The playwright dependency is isolated to these skills, so agents working purely with URDF or DXF generation avoid the overhead of browser automation.

Code Examples Using Installed Requirements

Generating Geometry with cadpy

Once cadpy is installed, create and export solid models:

# example_box.py

from cadpy import Box, Exporter

# Create a 100 × 50 × 20 mm box

box = Box(width=100, depth=50, height=20)

# Export as STEP for downstream CAD workflows

Exporter().write_step(box, "box.step")

Using the CAD Skill Programmatically

With the CAD skill requirements installed:

from skills.cad import CadSkill

cad = CadSkill()

# Generate parametric geometry from natural language

step_path = cad.generate(
    prompt="L-bracket 120 mm × 80 mm with 4 mm thickness and 6 mm radius fillet"
)
print(f"Model saved to: {step_path}")

The CadSkill class wraps the cadpy API and automatically routes outputs to the CAD Viewer for preview.

Launching the CAD Viewer

After installing the viewer skill:

npx skills install earthtojake/text-to-cad#cad-viewer

# Serve local models directory

cad-viewer serve --dir $(pwd)/models

The viewer starts at http://127.0.0.1:4178 by default.

Key Files for Python Requirements

• scripts/packages/cadpy/ — Core geometry library source
• skills/<skill>/requirements.txt — Per-skill dependency lists
• requirements-dev.txt — Repository-wide development tools

Summary

• Skill isolation: Each capability has its own requirements.txt at skills/<skill>/requirements.txt
• Shared core: All skills link to cadpy via --editable ./scripts/packages/cadpy
• Browser automation: playwright is required only for CAD generation and viewer skills
• Development tools: Separated into root-level requirements-dev.txt
• Installation paths: Use npx skills install for automation or manual pip commands for development

Frequently Asked Questions

What Python version does Text-to-CAD require?

The repository targets Python 3.10+ based on the cadpy package configuration and playwright compatibility requirements. Check scripts/packages/cadpy/pyproject.toml for the exact version spec.

Can I use a skill without installing Playwright?

Yes. Skills like URDF, SDF, SRDF, and DXF have no playwright dependency. Only CAD and CAD-Viewer require it for browser-based rendering and preview generation.

Why is cadpy installed with --editable?

The editable flag creates a symlink rather than copying files, so modifications to scripts/packages/cadpy/ are immediately available to all installed skills without reinstallation. This accelerates development across the multi-skill codebase.

How do I add a new dependency to a skill?

Edit the requirements.txt in the target skill directory. For dependencies used across multiple skills, consider adding them to cadpy itself so all skills inherit the capability through the shared editable install.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}