ComfyBridge

ComfyBridge is a Python-based service that acts as a bridge to the ComfyUI API, facilitating image generation requests. It manages the lifecycle of image generation requests, polls for their completion, and returns the final image as a base64-encoded string. This wrapper is designed to simplify integration with ComfyUI for developers needing a synchronous-like behavior over an inherently asynchronous operation.

Features

• Asynchronous Handling: Manages asynchronous image generation tasks and provides a synchronous HTTP interface to clients.
• Base64 Encoding: Returns images in a base64-encoded format, ready for use in web applications or for further processing.
• Polling with Timeout: Implements efficient polling with customizable timeouts to check for image generation completion.

Getting Started

Prerequisites

• Python 3.6+
• Flask
• Requests
• Pillow

Installation and Running

Clone the repository:

git clone https://github.com/RayKitajima/ComfyBridge.git
cd ComfyBridge

Install the required packages and run the Flask application:

$ python -m venv sandbox
$ cd sandbox/
$ source bin/activate
(sandbox) $ pip3 install flask requests pillow
(sandbox) $ cd ..
(sandbox) $ sandbox/bin/python app.py

To run the Flask application with a custom timeout, use the --timeout option:

(sandbox) $ sandbox/bin/python app.py --timeout 30

This will set the polling timeout to 30 seconds.

Running the ComfyUI API

You also need to have the ComfyUI API running on your local machine:

$ python main.py --listen

Usage

Send a POST request to /generate with the appropriate JSON payload:

(Replace <seed>, <positive prompt>, and <negative prompt> with the desired values.)

curl --location --request POST 'http://localhost:8189/generate' \
--header 'Content-Type: application/json' \
--data '{
    "prompt": {
        "3": {
            "inputs": {
                "seed": <seed>,
                "steps": 20,
                "cfg": 8,
                "sampler_name": "euler",
                "scheduler": "normal",
                "denoise": 1,
                "model": [
                    "4",
                    0
                ],
                "positive": [
                    "6",
                    0
                ],
                "negative": [
                    "7",
                    0
                ],
                "latent_image": [
                    "5",
                    0
                ]
            },
            "class_type": "KSampler",
            "_meta": {
                "title": "KSampler"
            }
        },
        "4": {
            "inputs": {
                "ckpt_name": "<model>"
            },
            "class_type": "CheckpointLoaderSimple",
            "_meta": {
                "title": "Load Checkpoint"
            }
        },
        "5": {
            "inputs": {
                "width": 512,
                "height": 512,
                "batch_size": 1
            },
            "class_type": "EmptyLatentImage",
            "_meta": {
                "title": "Empty Latent Image"
            }
        },
        "6": {
            "inputs": {
                "text": "<positive prompt>",
                "clip": [
                    "4",
                    1
                ]
            },
            "class_type": "CLIPTextEncode",
            "_meta": {
                "title": "CLIP Text Encode (Prompt)"
            }
        },
        "7": {
            "inputs": {
                "text": "<negative prompt>",
                "clip": [
                    "4",
                    1
                ]
            },
            "class_type": "CLIPTextEncode",
            "_meta": {
                "title": "CLIP Text Encode (Prompt)"
            }
        },
        "8": {
            "inputs": {
                "samples": [
                    "3",
                    0
                ],
                "vae": [
                    "4",
                    2
                ]
            },
            "class_type": "VAEDecode",
            "_meta": {
                "title": "VAE Decode"
            }
        },
        "9": {
            "inputs": {
                "filename_prefix": "ComfyUI",
                "images": [
                    "8",
                    0
                ]
            },
            "class_type": "SaveImage",
            "_meta": {
                "title": "Save Image"
            }
        }
    }
}'

API Reference

POST /generate

Accepts a JSON payload with the generation request details and returns a base64-encoded image.

Input

• prompt: JSON object representing the image generation prompt and configuration.

Output

• image: Base64-encoded string of the generated image.