CADomatic 🛠️

From prompt to CAD

Preview CADomatic on Hugging face

CADomatic is a Python-based tool that generates editable parametric CAD scripts for FreeCAD. Instead of creating static 3D models, CADomatic produces fully customizable CAD models with Python scripts that build CAD geometry — allowing engineers to programmatically define parts, reuse templates, and iterate fast.

---

🔍 What It Does

CADomatic Demo

Fully editable parametric designs

• ✅ Generates editable FreeCAD Python scripts for parts like screws, nuts, fasteners, and more
• ✅ Each script can be modified for custom parameters (length, diameter, features, etc.)
• ✅ Outputs native .py scripts which use FreeCAD’s API to build geometry
• ✅ Enables version-controlled, reusable, parametric CAD pipelines
• ✅ Eliminates the need for manual modeling in the FreeCAD GUI

---

Workflow

Current workflow of CADomatic v 1.0

- Current Workflow (v1.0)

The current version of CADomatic follows a simple pipeline to generate CAD models from natural language descriptions:

1. User Input – Provide a design description as text.
2. LLM Processing – The LLM generates a FreeCAD Python script.
3. Model Creation – The script runs in FreeCAD (via CADomatic) to create a 3D model and render an image.
4. Output – The final CAD file and preview image are returned to the user.

Proposed workflow of CADomatic v 2.0

- Proposed Workflow (v2.0)

I am currently working on an enhanced workflow for CADomatic v2.0, which introduces image-based validation, feedback loops, and optional reference images for higher accuracy and user control:

1. User Input – Provide a text description and optionally a reference keyword.
2. Reference Image Input – Provide an optional reference image or a drafting of the desired CAD model.
3. LLM Processing – The LLM generates a FreeCAD Python script.
4. Model Creation & Rendering – CADomatic creates the 3D model and renders an image.
5. Similarity Scoring – The rendered image is compared against reference images to compute a similarity score.
6. Validation & Feedback Loop
   • If the score meets the threshold, the final CAD model is returned.
   • If not, a VLM compares the images, explains differences, and feeds the desired changes into the LLM for regeneration for furtehr refinement.

---

💡 Why Use CADomatic?

• 🔁 Automate repetitive CAD tasks
• 🧱 Build part libraries as code
• 🧪 Integrate CAD into testing or CI workflows
• 🔧 Customize geometry by changing script parameters
• 📐 Keep models lightweight and editable at the code level

---

💬 Example Prompts

Here are some example natural language prompts you can use to generate CAD scripts with CADomatic:

• "Build a flange with a 100mm outer diameter, 10mm thickness, and 6 bolt holes evenly spaced."
• "Make a cylindrical spacer, 20mm diameter and 30mm height, with a 5mm through hole."
• "Produce a washer with an outer diameter of 25mm and an inner diameter of 10mm."
• "Design a toy car with a rectangular box as the body and 4 circular wheels attached to the sides of the box."

These prompts will be converted into editable Python scripts that you can modify and reuse.

---

⚠️ This is the first version of CADomatic — a flash of what's possible.
Future versions under development will include:

• Improved LLM-driven script generation
• A dedicated user interface for part selection and parameter tuning
• More robust template and geometry libraries

---

🚀 How to Use CADomatic

✅ Prerequisites

• Python 3.11+
• FreeCAD (must be installed and added to PATH)
• uv (install via pip install uv)

⚙️ Setup

git clone https://github.com/yas1nsyed/CADomatic.git
cd CADomatic

# Create and activate virtual environment
python -m venv .venv
.venv\Scripts\activate  

# Install dependencies
uv pip install -r requirements.txt

• 🔐 Set Up Gemini API Key

• Create a .env file in the project root:

• ▶️ Run CADomatic

# Run the program
uv run main.py

• Enter your prompt (e.g., "Create a 10mm cube with 2mm hole").
• FreeCAD will auto-launch with your generated model.

---