CICADA: Collaborative Intelligent CAD Automation Design Agent

中文 | English

Welcome to CICADA, the Collaborative Intelligent CAD Automation Design Agent. CICADA is a cutting-edge framework designed to streamline and enhance the CAD design process through intelligent automation and collaboration. This repository contains the core modules and utilities that power CICADA, enabling seamless integration with CAD workflows.

📖 Documentation: For detailed documentation and tutorials, visit CICADA Documentation.

---

Repository Structure

The repository is organized into the following main modules:

• common: Core utilities and shared functionalities across the framework.
• geometry_pipeline: Tools for processing and converting 3D models, including point cloud generation and snapshots.
• describe: Components for generating and managing descriptive metadata for 3D models.
• coding: Code generation, execution, and debugging tools for CAD automation.
• feedbacks: Modules for analyzing and providing feedback on design iterations.
• retrieval: Tools for retrieving and management of documentation, model data, and design resources.
• workflow: Orchestration of CICADA's automation workflows and agent management.

---

Setting Up the Environment

Prerequisites

Before setting up CICADA, ensure you have the following installed:

• Python 3.9+
• Conda or pip (for dependency management)

Installation Steps (Quick Start)

# activate your venv or conda env first
pip install cicada-agent

Installation Steps (For Developers)

1. Clone the Repository

git clone https://github.com/Oaklight/cicada.git
cd cicada

2. Install Dependencies

Option 1: Using Conda (Recommended)

conda env create -f environment.yml
conda activate cicada

Option 2: Using pip

python -m venv cicada
source cicada/bin/activate
pip install -r requirements.txt

3. Update API Keys

The provided API keys in the config files are deprecated. Update the api_key and api_base_url in config.yaml or config/*.yaml in each module:

---

Key Modules and Usage

geometry_pipeline

• convert.py: Converts 3D models (STEP, OBJ, STL) to point cloud data (PLY) or other formats.

  python geometry_pipeline/convert.py --step_file <path_to_step_file> --convert_step2obj

  Options:
  --convert_step2obj, --convert_obj2pc, --convert_step2stl, --convert_obj2stl, --convert_stl2obj, --convert_stl2pc, --reaxis_gravity

• snapshots.py: Generates preview snapshots of 3D models from multiple angles.

  python geometry_pipeline/snapshots.py --step_file <path_to_step_file> --snapshots

  Options:
  --obj_file, --step_file, --stl_file, -o OUTPUT_DIR, -r RESOLUTION, -d DIRECTION, -p, --reaxis_gravity

describe

• describer_v2.py: Generates descriptive metadata for 3D models using advanced language models.

  python describe/describer_v2.py "Describe the 3D model" --config <path_to_config> --prompts <path_to_prompts>

  Options:
  --config CONFIG, --prompts PROMPTS, -img REF_IMAGES, -o OUTPUT

coding

• coder.py: Generates CAD scripts based on design goals.

  python coding/coder.py "Design a mechanical part" --config <path_to_config> --prompts <path_to_prompts>

  Options:
  --config CONFIG, --master_config_path MASTER_CONFIG_PATH, --prompts PROMPTS, -o OUTPUT_DIR

feedbacks

• visual_feedback.py: Analyzes rendered images of a design against the design goal.

  python feedbacks/visual_feedback.py --design_goal "Design a mechanical part" --rendered_images <path_to_images>

  Options:
  --config CONFIG, --prompts PROMPTS, --reference_images REFERENCE_IMAGES, --rendered_images RENDERED_IMAGES

retrieval

• tools/build123d_retriever.py: Retrieves and manages documentation for CAD tools and libraries.

  python retrieval/tools/build123d_retriever.py [--force-rebuild] [--interactive] [--metric {l2,cosine}] [--query QUERY] [--debug]

  Options:
  --force-rebuild: Force rebuild the database.
  --interactive: Run in interactive mode to ask multiple questions.
  --metric {l2,cosine}: Distance metric to use for similarity search.
  --query QUERY: Query text to search in the database.
  --debug: Enable debug mode for detailed logging.

  Examples:
  Interactive mode:

  python retrieval/tools/build123d_retriever.py --interactive

  Single query:

  python retrieval/tools/build123d_retriever.py --query "How to extrude a shape?"

workflow

• codecad_agent.py: Orchestrates the automation workflows for CAD design.

  python workflow/codecad_agent.py "Design a mechanical part" --config <path_to_config> --prompts <path_to_prompts>

  Options:
  --config CONFIG: Path to the configuration file.
  --prompts PROMPTS: Path to the prompts file.
  -img REF_IMAGES: Path to reference images (optional).
  -o OUTPUT_DIR: Directory to save output files (optional).

  Example:

  python workflow/codecad_agent.py "Design a mechanical part" --config workflow/config/code-llm.yaml --prompts workflow/prompts/code-llm.yaml -o output/

---

Contributing

We welcome contributions from the community! If you'd like to contribute to CICADA, please follow these steps:

1. Fork the repository.
2. Create a new branch for your feature or bugfix.
3. Submit a pull request with a detailed description of your changes.

---

License

CICADA is licensed under the MIT License. For more details, see the LICENSE file.

---

Contact

For questions, feedback, or support, please post via GitHub Issues or contact us at [dingpeng]@@uchicago[dot]edu.

---

Citation

If you use Cicada in your research, please consider citing:

@software{Cicada,
  author = {Peng Ding},
  title = {Cicada: Collaborative Intelligent CAD Automation Design Agent},
  month = {January},
  year = {2025},
  url = {https://github.com/Oaklight/cicada}
}

---

CICADA — Revolutionizing CAD Design with Intelligent Automation. 🚀