ComfyUI Bridge for AI Power Grid

Connect your local ComfyUI installation to the AI Power Grid network and run it as a distributed image generation worker.

---

🚀 Overview

• Bridge: Receives image-generation jobs from AI Power Grid.
• Worker: Executes jobs via your local ComfyUI instance.
• Return: Uploads generated images back to the network.

This allows you to contribute GPU cycles to a decentralized AI rendering network while leveraging your local ComfyUI setup.

---

🎯 Features

• Auto-detects installed ComfyUI model checkpoints and maps them to AI Power Grid model names.
• Customizable: override advertised models via GRID_MODEL (supports comma-separated lists).
• Workflow templating: use your own ComfyUI .json workflow files.
• Async, multi-threaded job polling and processing.

---

🛠 Prerequisites

1. Python 3.9+
2. ComfyUI running locally (default: http://127.0.0.1:8000).
3. AI Power Grid account + API key: https://dashboard.aipowergrid.io

---

📦 Installation

# 1. Clone the repo
git clone https://github.com/youruser/comfy-bridge.git
cd comfy-bridge

# 2. Create & activate a virtual environment
python -m venv venv
# macOS/Linux
source venv/bin/activate
# Windows
venv\Scripts\activate

# 3. Install dependencies
pip install -e .

---

⚙️ Configuration

Copy the example .env and adjust values:

# .env
GRID_API_KEY=your_powergrid_api_key          # required
GRID_WORKER_NAME=MyComfyWorker.APIG_Wallet   # optional
COMFYUI_URL=http://127.0.0.1:8000            # optional
GRID_API_URL=https://api.aipowergrid.io/api  # optional
GRID_NSFW=false                              # allow NSFW? true/false
GRID_THREADS=2                               # concurrent jobs
GRID_MAX_PIXELS=1048576                      # max output resolution (pixels)
GRID_MODEL=stable_diffusion, Flux.1-Krea-dev Uncensored (fp8+CLIP+VAE)  # comma-separated model names
WORKFLOW_FILE=my_workflow.json               # ComfyUI JSON export template

• GRID_MODEL supports one or more model keys (comma-separated). If unset, the bridge auto-detects from your ComfyUI checkpoints.
• WORKFLOW_FILE points to a JSON workflow in your workflows/ directory.

---

▶️ Running the Bridge

Start your ComfyUI web server, then:

# Via CLI module
python -m comfy_bridge.cli

Or directly (legacy):

start_bridge.py

The bridge will:

1. Register as a worker with AI Power Grid.
2. Poll for jobs every few seconds.
3. Render in ComfyUI.
4. Submit results back to the network.

---

🐳 Docker

Build & Run the Container

1. Build the Docker image:

   docker build -t comfy-bridge .

2. Run the container:

   • Linux (host networking):

     docker run --rm --network host --env-file .env comfy-bridge

   • macOS/Windows (using host.docker.internal):

     docker run --rm \
       -v "$(pwd)/workflows:/app/workflows" \
       --env-file .env \
       -e COMFYUI_URL=http://host.docker.internal:8000 \
       comfy-bridge

---

🐳 Docker Compose

If you prefer using Docker Compose to run your Container

Linux

Build & run:

docker-compose -f docker-compose.linux.yml up --build

macOS/Windows

Build & run:

docker-compose -f docker-compose.win-macos.yml up --build

---

✅ Testing

All core modules include unit and async tests. To run them:

pytest

Tests use pytest-asyncio for async routines and respx for HTTP mocking.

---

🐞 Troubleshooting

• No jobs found? Check Advertising models: log; ensure GRID_MODEL is set or your checkpoints match default mappings.
• 400 Bad Request: unrecognized models—verify model key names or adjust GRID_MODEL.
• ComfyUI unreachable: confirm COMFYUI_URL and that the server is running.
• API auth errors: verify GRID_API_KEY and network access.

Logs are printed at INFO (bridge flow) and DEBUG (detailed payloads) levels. Adjust via:

import logging
logging.basicConfig(level=logging.DEBUG)

---

📄 License

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

---

🙏 Acknowledgements

• AI Power Grid (https://aipowergrid.io) - For the API
• ComfyUI (https://github.com/comfyanonymous/ComfyUI) - For the local image generation backend
• httpx, aiohttp, pytest, pytest-asyncio ❤️