1-comfyui

ComfyUI API Wrapper

This project provides a FastAPI-based wrapper around ComfyUI, making it easy to run Stable Diffusion workflows programmatically through a RESTful API.

Overview

The ComfyUI API Wrapper starts a ComfyUI server in the background and exposes endpoints to:

• Run Stable Diffusion workflows with custom parameters
• Check server health status
• Dynamically load and configure models

The API automatically handles model downloading, workflow template substitution, and converting the generated images to base64 for easy integration with other services.

Installation

1. Clone this repository:

git clone https://github.com/yourusername/comfyui-api-wrapper.git
cd comfyui-api-wrapper

2. Install dependencies:

pip install -r requirements.txt

3. Clone ComfyUI repository (if not already included):

git clone https://github.com/comfyanonymous/ComfyUI
pip install -r ComfyUI/requirements.txt

4. Update your workflow template in workflow_api.json and models in model.json if needed.

5. Deploy to Cerebrium:

cerebrium deploy

Usage

API Endpoints

Run a Workflow

POST /run

This endpoint accepts a JSON payload with parameters that will be substituted into the workflow template.

Example Request:

curl -X POST http://localhost:8765/run \
  -H "Content-Type: application/json" \
  -d '{
    "seed": 12345,
    "positive_prompt": "aerial view, a futuristic research complex in a bright foggy jungle, hard lighting",
    "negative_prompt": "worst quality, low quality, blurry, cropped, lowres",
    "controlnet_image": "https://example.com/image.png"
  }'

Example Response:

{
  "result": [
    {
      "node_id": "18",
      "data": "iVBORw0KGgoAAAANSUhEUgAA...",
      "format": "png"
    },
    {
      "node_id": "15",
      "data": "R0lGODlhAQABAIAA...",
      "format": "png"
    }
  ],
  "status": "success"
}

The data field contains the base64-encoded image data that can be directly embedded in HTML or saved to a file.

Customizing Workflows

The API uses workflow_api.json as a template for image generation. You can customize this file to use different models, parameters, or node configurations.

The template uses placeholders in the format {{parameter_name}} which are replaced with values from your API request.

Default Placeholders:

• {{seed}} - Random seed for generation stability
• {{positive_prompt}} - Text prompt describing what you want to generate
• {{negative_prompt}} - Text prompt describing what you want to avoid
• {{controlnet_image}} - URL or base64 data of an image for ControlNet

Advanced Usage

Custom Models

To use custom models, modify the model.json file to include URLs and paths for additional models:

[
  {
    "url": "https://huggingface.co/your-model-url/resolve/main/model.safetensors",
    "path": "models/checkpoints/your_model.safetensors"
  },
  {
    "url": "https://github.com/username/custom-node-repo.git",
    "path": "custom_nodes"
  }
]