Summary

Articraft addresses the complexity of generating articulated 3D assets by transforming the process into a programmatic, LLM-driven code-generation workflow. It targets engineers and researchers needing scalable, controllable 3D asset creation, offering a local-first architecture for simplified management and a small repository footprint.

How It Works

This system leverages Large Language Models (LLMs) to generate Python code that defines articulated 3D assets. Its core approach is an agentic workflow where LLMs produce executable scripts, enabling programmatic asset creation. The architecture is designed as local-first, storing records in a plain data folder and maintaining a minimal code repository, with a simple manifest for browsing complete local assets. This design prioritizes user control and simplifies deployment.

Quick Start & Requirements

• Installation: Run just setup from the repository root.
• Prerequisites: Python 3.12 (3.11 supported; 3.13+ not supported), uv for package management, and npm (optional, for the local viewer frontend).
• Configuration: Set provider API keys (e.g., OPENAI_API_KEY) in .env, or integrate external agents like Codex.
• Asset Generation: Use uv run articraft generate "Your asset description.".
• Viewer: Start the local viewer with just viewer.
• Documentation: Paper, Project Page, Codex Plugin Setup, Architecture & Project Structure, Documentation & Advanced Usage.

Highlighted Details

• Enables scalable generation of articulated 3D assets via LLM-driven code.
• Features a local-first architecture with data stored in a plain, gitignored folder.
• Includes a local viewer with a React frontend for browsing generated assets.
• Supports image-conditioned generation and integration with external AI agents like Codex.
• Security Note: Executes generated model.py files as Python code, requiring trust in the source.

Maintenance & Community

The provided README does not detail specific community channels (like Discord/Slack), active maintainers beyond the paper's authors, or a public roadmap.

Licensing & Compatibility

The project is licensed under the Apache-2.0 License, which is generally permissive for commercial use and integration into closed-source projects. However, data contributed to Articraft is explicitly released under the Creative Commons Attribution 4.0 International (CC-BY 4.0) license.

Limitations & Caveats

Python versions 3.13 and later are not currently supported. Users must exercise caution and only run generated records and model scripts from trusted sources due to the execution of Python code. Contributions of data are subject to public release under CC-BY 4.0.