English | 中文

🎯 ComfyUI-SymbOmni: Evolving Agentic Omni Models via Symbolic Concept Learning

🌟 Introduction

ComfyUI-SymbOmni is a fully open-source, locally-deployable AIGC intelligent assistant built on ComfyUI. Unlike its predecessor, all features run entirely on your local machine using your own LLM API (OpenAI-compatible endpoints, LM Studio, etc.), ensuring complete privacy and control over your data.

✨ Key Highlights

• 🔓 100% Open Source - All code is open-source, no cloud dependencies
• 🏠 Local First - Runs entirely on your machine with your own LLM
• 🔌 OpenAI-Compatible - Works with any OpenAI-compatible API (OpenAI, Claude, LM Studio, Ollama, etc.)
• 📚 Workflow Library - Built-in workflow management and library system
• 🧠 Memory System - Mem0-based planning memory for improved task understanding
• 🎨 Parameter Lab - Visual parameter tuning with batch execution

---

🔥 Core Features

1. 💎 Workflow Generation & Library

Generate workflows from natural language descriptions. The system searches your local workflow library and provides AI-generated suggestions.

• Type in the input box: Generate a cat and then xxx
• Browse and manage workflows in the Library tab

2. 💎 Workflow Library Management

The new Library tab provides:

• 📁 Browse and search your workflow collection
• 🏷️ Organize workflows with tags and categories
• ⬇️ One-click import workflows to canvas
• 📤 Export and share workflows

3. 💎 Planning Memory (Mem0 Integration)

The system learns from your planning experiences:

• 🧠 Stores successful planning patterns locally
• 📖 Retrieves relevant experiences for similar tasks
• 🔄 Continuously improves task understanding

4. 💎 Intelligent Workflow Debug

Automatically analyze and fix workflow errors, including parameter issues, connection problems, and missing models.

• Click the Debug button in the input box to debug the current canvas
• Auto-detect and prompt for missing model downloads

5. 💎 Workflow Rewriting & Modification

Describe what you want to change, and the AI will help modify your workflow - add nodes, adjust parameters, and optimize structure.

• Type: Help me add xxx to the current canvas
• Add expert knowledge to help with newer models and techniques

6. 💎 GenLab - Parameter Tuning Laboratory

Visual parameter tuning tool for batch experimentation. Set parameter ranges and the system automatically executes different combinations with visual comparison results.

• Switch to the GenLab tab
• Define parameter ranges and run batch experiments
• Compare results visually to find optimal configurations

---

⚙️ Configuration

LLM Setup

ComfyUI-SymbOmni works with any OpenAI-compatible API. Configure in the settings panel:

1. API Base URL: Your LLM endpoint (e.g., https://api.openai.com/v1, http://localhost:1234/v1 for LM Studio)
2. API Key: Your API key (optional for local LLMs like LM Studio)
3. Model: Select your preferred model

Recommended Models:

• gemini-2.5-flash (Default, great balance of speed and quality)
• gpt-4o / gpt-4o-mini
• claude-4.5-sonnet
• Local models via LM Studio or Ollama

Memory Backend

For enhanced planning capabilities, install Mem0:

pip install mem0ai

Configure embedding model in settings for better experience retrieval.

---

🚀 Getting Started

Installation

Method 1: Git Clone (Recommended)

cd ComfyUI/custom_nodes
git clone --depth 1 https://github.com/Snowflake-Pink/ComfyUI-SymbOmni.git
cd ComfyUI-SymbOmni
pip install -r requirements.txt

For Windows portable version:

python_embeded\python.exe -m pip install -r ComfyUI\custom_nodes\ComfyUI-SymbOmni\requirements.txt

Method 2: ComfyUI Manager

Search for "ComfyUI-SymbOmni" in Custom Nodes Manager and install.

Quick Start

1. Start ComfyUI
2. Click the SymbOmni icon in the sidebar
3. Configure your LLM API in settings (gear icon)
4. Start chatting! Try: "I want a workflow for text to image generation"

---

📁 Project Structure

ComfyUI-SymbOmni/
├── backend/
│   ├── service/
│   │   ├── symbomni/          # Core AI services
│   │   │   ├── llm_client.py  # LLM integration (OpenAI-compatible)
│   │   │   ├── memory.py      # Mem0-based planning memory
│   │   │   ├── planning.py    # Workflow planning engine
│   │   │   └── workflow_runner.py  # Workflow execution
│   │   ├── workflow_manager.py     # Workflow library management
│   │   └── workflow_rewrite_agent.py  # Workflow modification
│   └── data/
│       └── atomic_workflow/   # Workflow library storage
├── ui/                        # React frontend
└── locales/                   # i18n translations

---

🤝 Contributing

We welcome contributions! Feel free to:

• 🐛 Report bugs via Issues
• 💡 Suggest new features
• 🔧 Submit Pull Requests
• 📖 Improve documentation

---

🙏 Acknowledgments

This project is forked from ComfyUI-Copilot. We sincerely thank the AIDC-AI team for their excellent foundational work that made this open-source evolution possible.

Special thanks to:

• The ComfyUI community
• Mem0 for the memory system
• All contributors and users

---

📚 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Star History