GENESISCORE ADDON - DEEP DIVE

Overview

Repo: https://github.com/AIGODLIKE/GenesisCore Version: 0.0.1 | Blender: 4.0+ | License: Open Source Stars: 134 | Language: Python

GenesisCore is a full MCP Client + Server built directly into Blender. Unlike blender-mcp (server only), GenesisCore includes the LLM client, so you chat with AI inside Blender via N-panel.

Architecture

┌────────────────────────────────────────────────────────────┐
│                    GENESISCORE ADDON                        │
├────────────────────────────────────────────────────────────┤
│  src/                                                       │
│  ├── client/           # MCP Clients (LLM providers)       │
│  │   ├── base.py       # MCPClientBase - core logic        │
│  │   ├── openai.py     # OpenAI compatible                │
│  │   ├── openrouter.py # OpenRouter (extends OpenAI)      │
│  │   ├── deepseek.py   # DeepSeek API                     │
│  │   ├── claude.py     # Anthropic Claude                 │
│  │   ├── ollama.py     # Local Ollama                     │
│  │   └── siliconflow.py# SiliconFlow                      │
│  ├── server/           # MCP Server (FastMCP)              │
│  │   ├── server.py     # BlenderMCPServer on :45677       │
│  │   └── tools/        # Tool modules                      │
│  │       ├── common_tools.py    # Scene info, code exec    │
│  │       ├── object_tools.py    # Create/modify/delete     │
│  │       ├── material_tools.py  # PBR materials            │
│  │       ├── polyhaven_tools.py # Poly Haven assets         │
│  │       └── asset_tools.py     # Local assets             │
│  ├── preference.py     # Settings UI + config persistence  │
│  ├── ui.py             # N-panel interface                  │
│  ├── executor.py       # Thread-safe bpy execution         │
│  └── logger.py         # Logging                           │
└────────────────────────────────────────────────────────────┘

Key Features

1. Built-in MCP Client (No External Host Needed)

# In base.py - connects to local MCP server on port 45677
async def connect_to_server(self):
    headers = {"Content-Type": "application/json", "Accept": "application/json"}
    stdio_transport = await self.exit_stack.enter_async_context(
        sse_client(url="http://localhost:45677/sse", headers=headers)
    )

2. Multi-Provider Support

Provider	Base URL	Default Model
OpenRouter	`https://openrouter.ai/api`	`anthropic/claude-3.5-haiku`
DeepSeek	`https://api.deepseek.com`	`deepseek-chat`
OpenAI	`https://api.openai.com/v1`	`gpt-4o`
Anthropic	`https://api.anthropic.com`	`claude-3-5-sonnet`
Ollama	`http://localhost:11434/v1`	`llama3.2`
SiliconFlow	`https://api.siliconflow.cn/v1`	`Qwen/Qwen2.5-Coder-32B`

3. Tool Modules (Extensible)

Each module registers tools with the MCP server:

# common_tools.py - Most powerful tool
def execute_blender_code(code: str) -> dict:
    """Execute arbitrary Blender Python API code"""
    namespace = {"bpy": bpy}
    exec(code, namespace)
    return {"executed": True}

Available Tools:

Module	Tools
`common_tools`	`get_scene_info`, `get_simple_info`, `execute_blender_code`
`object_tools`	`create_object`, `modify_object`, `delete_object`, `get_object_info`
`material_tools`	`set_material` (Principled BSDF)
`polyhaven_tools`	`search_models`, `fetch_model_info`, `use_model_asset`, `search_hdris`, `use_hdri`

4. Poly Haven Integration

Caches to %TEMP%\polyhaven_{models|hdris}
Auto-loads cached assets
Supports 1k/2k/4k/8k/16k resolutions

5. External MCP Host Connection

{
  "mcpServers": {
    "BlenderGenesis": {
      "url": "http://localhost:45677"
    }
  }
}

Allows Claude Desktop / Cursor / Codex to connect to GenesisCore's server.

Installation

Method 1: ZIP Install (Recommended)

Download: https://github.com/AIGODLIKE/GenesisCore/archive/refs/heads/main.zip
Blender → Edit → Preferences → Add-ons → Install from ZIP
Select the ZIP file
Enable GenesisCore (search "GenesisCore" or "AI")

Method 2: Git Clone (Auto-updates)

cd %APPDATA%\Blender Foundation\Blender\4.2\scripts\addons
git clone https://github.com/AIGODLIKE/GenesisCore.git

Then enable in Preferences.

Configuration

OpenRouter Setup (Recommended)

Get API key: https://openrouter.ai/keys
In Blender N-panel (press N) → GenesisCore tab
Provider: OpenRouter
API Key: Paste your key
Fetch Models → Select: deepseek/deepseek-chat or qwen/qwen-2.5-coder-32b
Tool Modules: Enable all (hold Shift to multi-select)
Save Config

Model Recommendations (2026)

Task	Model	Cost
Complex reasoning	`anthropic/claude-3.5-sonnet`	$$$
Coding/Blender API	`deepseek/deepseek-chat`	$
Fast/cheap	`qwen/qwen-2.5-coder-32b`	$
Local (free)	`ollama:codellama:13b`	Free

Usage Workflow

1. Open N-panel (N) → GenesisCore
2. Select Provider + Model
3. Enable Tool Modules (Common, Object, Material, Polyhaven)
4. Type prompt in chat box
5. Press Enter / Click Execute
6. AI plans → calls tools → executes in Blender
7. Result appears in chat + viewport updates

Example Prompts

Basic Geometry

"Create a low-poly castle with 4 towers, a central keep, and a moat. Use simple primitives."

Material + Lighting

"Create a studio lighting setup: 3-point lighting with key, fill, and rim lights. Make a chrome sphere in center."

Poly Haven Assets

"Search Poly Haven for a Victorian chair and an HDRI for indoor lighting. Import both and set up the scene."

Complex Scene

"Build a cyberpunk alley scene: wet pavement (use HDRI), neon signs (emission materials), 
steam particles, and a hovering vehicle. Use Poly Haven assets where possible."

Code Execution (Direct)

"Execute this code to create a procedural city grid with random building heights"

→ AI calls execute_blender_code with full bpy script

Advanced: Custom Tool Development

1. Create Tool Module

# src/tools/my_tools.py
from .common import ToolsPackageBase

class MyTools(ToolsPackageBase):
    def my_custom_tool(param: str) -> dict:
        """Description for LLM"""
        # bpy operations here
        return {"result": "success"}

2. Register in `init.py`

# src/tools/__init__.py
from .my_tools import MyTools
# Order matters for UI display

3. Custom LLM Client

# src/client/custom.py
from .base import MCPClientBase

class MCPClientCustom(MCPClientBase):
    @classmethod
    def info(cls):
        return {"name": "Custom", "description": "My Provider", "version": "0.1"}
    
    @classmethod
    def default_config(cls):
        return {"base_url": "https://api.custom.com", "model": "default"}

Troubleshooting

Issue	Fix
"Connecting..." hangs	Check port 45677 not blocked; restart Blender
Model list empty	Verify API key; check network/proxy
Tools not showing	Enable modules in N-panel (Shift+click for multiple)
Poly Haven fails	Check internet; cache in `%TEMP%\polyhaven_*`
Code execution errors	Open System Console (Window → Toggle System Console)

File Locations

File	Path
Addon root	`%APPDATA%\Blender Foundation\Blender\4.2\scripts\addons\GenesisCore\`
Config	`%APPDATA%\Blender Foundation\Blender\4.2\config\genesiscore_config.json`
Poly Haven cache	`%TEMP%\polyhaven_models\` / `polyhaven_hdris\`
Logs	Blender System Console

🔗 СВЯЗИ

Прочие в категории «Референсы»: BLENDER-MCP ADDON - DEEP DIVE, INSTALLATION GUIDE - COMPLETE SETUP, TROUBLESHOOTING GUIDE, BLENDER-MCP - FULL DOCUMENTATION, DREAM TEXTURES ADDON - DEEP DIVE
По темам: MCP ARCHITECTURE - MODEL CONTEXT PROTOCOL FOR BLENDER, MCP CLIENT CONFIGURATIONS, MCP SERVER API REFERENCE
Теги: #addon, #mcp, #reference

🎯 ЦЕЛЬ

Референс категории «Референсы», сложность —.