sixtysevenlf/dsh-blender-plugin

What it solves

This project provides a direct, high-speed communication bridge between an AI agent and Blender. It eliminates the need for manual screenshots or cumbersome UI interactions, allowing an AI to programmatically control the 3D viewport, perform scene edits, and verify results in real-time.

How it works

It uses a TCP-based channel to connect an AI agent (via the DeepSeek Harness framework) to a Blender addon. The system exposes 16 specialized tools that allow the AI to perform tasks like capturing viewport frames, executing Python scripts directly within Blender, running optimization loops, and managing headless render jobs. The plugin includes a "contract layer" that provides a catalog of over 180 operations, ensuring the AI can discover and use Blender's capabilities reliably.

Who it’s for

This tool is designed for developers and AI researchers building agentic workflows that require precise, automated 3D modeling, scene generation, or automated rendering tasks within Blender.

Highlights

• Direct Control: Enables the AI to "see" the viewport and execute commands without human intervention.
• High Performance: Optimized for low-latency feedback, with typical viewport capture times between 50–160ms.
• Inner Loop Optimization: Supports running thousands of iterations inside Blender's main thread for tasks like procedural shape generation or parameter tuning.
• Robust Integration: Includes a "doctor" diagnostic tool to verify installation and connectivity, ensuring the agent and Blender are properly synchronized.
• Cross-Platform: Supports Windows, WSL, and macOS with automated path mapping and environment detection.