https://github.com/victordibia/blenderlm
BlenderLM is a Python package that enables LLMs (Large Language Models) to control and interact with Blender, the open-source 3D creation suite.
https://github.com/victordibia/blenderlm
Last synced: 11 months ago
JSON representation
BlenderLM is a Python package that enables LLMs (Large Language Models) to control and interact with Blender, the open-source 3D creation suite.
- Host: GitHub
- URL: https://github.com/victordibia/blenderlm
- Owner: victordibia
- License: mit
- Created: 2025-03-11T22:55:09.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2025-06-05T05:49:15.000Z (about 1 year ago)
- Last Synced: 2025-06-05T07:44:03.733Z (about 1 year ago)
- Language: TypeScript
- Size: 463 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# BlenderLM
[](https://opensource.org/licenses/MIT)
> Note: This is a sample project I built as part of a talk (it may not be maintained going forward).
BlenderLM is a Python sample app that enables LLMs (Large Language Models) to control and interact with Blender, the open-source 3D creation suite. It provides a clean API and tools designed specifically for use with LLM agent frameworks like Autogen/GoogleADK/CrewAI.
## Features
- REST API for controlling Blender programmatically
- Ready-to-use tools for LLM agents
- Support for creating, modifying, and manipulating 3D objects
- Material and scene management
- Web UI for intuitive interaction (optional)
## Installation
```bash
# Install the base package
pip install blenderlm
```
## Architecture
BlenderLM consists of three main components:
1. **Blender Addon**: A Blender addon that exposes Blender functionality via a socket server
2. **API Server**: A FastAPI server that communicates with the Blender addon and provides a REST API
3. **Client Library**: Python client and tools for interacting with the API server (with Autogen integration)
## Blender Addon Setup
1. Install the BlenderLM addon in Blender:
- Download the Blender addon file from the `addon` directory
- In Blender, go to Edit > Preferences > Add-ons > Install
- Select the downloaded file and click "Install Add-on"
- Enable the addon by checking the box next to "Interface: BlenderLM"
2. Start the Blender addon server:
- In Blender, go to the sidebar (N key)
- Find the "BlenderLM" tab
- Set the port (default: 9876)
- Click "Start Server"
## Usage
### Starting the BlenderLM Server
```bash
# Start the server
blenderlm serve --port 8000 --blender-port 9876
```
### Interacting with BlenderLM
You can interact with BlenderLM using the provided client library. Below is an example of how to create a simple agent that uses BlenderLM tools to create a 3D object in Blender.
```python
import asyncio
from blenderlm.client.agents import OpenAIAgent
from blenderlm.client import get_blender_tools
from blenderlm.client.agents import AgentMessage
async def main():
blender_tool_functions = await get_blender_tools()
# Instantiate the agent, passing the tool functions
agent = OpenAIAgent(
tools=blender_tool_functions,
model_name="gpt-4.1-mini",
)
updates = agent.run_stream(
task="Create a low poly well with the right materials and colors.",
)
async for update in updates:
if isinstance(update, AgentMessage):
print(update.role,":", update.content, str(update.metadata))
else:
print(update.role,":", update.content)
if __name__ == "__main__":
asyncio.run(main())
```
## Available Tools
BlenderLM provides the following tools:
- **Object Creation**: Create 3D primitives (cube, sphere, cylinder, etc.)
- **Object Manipulation**: Move, rotate, and scale objects
- **Material Management**: Create and apply materials
- **Scene Management**: Manage the 3D scene
- **Rendering**: Render the scene to an image
## Command Line Interface
```bash
# Show help
blenderlm --help
# Show version
blenderlm version
# Start the server
blenderlm serve --port 8000 --blender-port 9876
# Test connection
blenderlm test
```
## Security Note
The BlenderLM addon opens a socket server that accepts and executes commands. Only use it on trusted networks as it has no authentication mechanism by default.
## License
MIT
## Acknowledgement
Inspired by [blender-mcp](https://github.com/ahujasid/blender-mcp)