This repo has initialized a fastapi server with multiple ComfyUI workflow tasks.

1. Navigate to your ComfyUI/custom_nodes directory

2. Clone this repo

   git clone https://github.com/Thatmakes11/ComfyUI-to-PyTorch-functions.git

3. Move .py files in examples to the ComfyUI directory

   After cloning and moving, your ComfyUI directory should look like this:

   /comfy /comfy_extras /custom_nodes --/ComfyUI-to-Python-Extension /input /models /output /script_examples /web .gitignore LICENSE README.md comfyui_screenshot.png clothes_swap.py # file from examples cuda_mollac.py execution.py extra_model_paths.yaml.example fastapi_server.py # file from examples folder_paths.py latent_preview.py main.py nodes.py node_manager.py # file from examples path_manager.py # file from examples remove_background.py # file from examples requirements.txt server.py 

1. Navigate to the ComfyUI-to-Python-Extension folder and install requirements

   pip install -r requirements.txt

2. Load up your favorite workflows in ComfyUI, then click the newly enabled Save (API Format) button under Queue Prompt

3. Run the script with optional arguments:

   python comfyui_to_pytorch_function.py --input_file "workflow_api.json" --output_file my_workflow.py

4. After running comfyui_to_pytorch_function.py, a new .py file will be created in the current working directory. If you made no changes, look for workflow_api.py.

5. Now you have a function that execute PyTorch process.

To start the FastAPI server, run:

uvicorn fastapi_server:app --host 0.0.0.0 --port 8080

or

python fastapi_server.py

Endpoint: /predict/image_generation

Method: POST

Request Body:

{ "task_type": "string", "task_id": "string", "params": "dict" }

Current supported tasks are ClothesSwap and Rmbg, if you wish to add your own tasks (Workflow to Pytorch), check for line 15 and line 52-58 in fastapi_server.py.

Also, your customed torch functions are required to return a dict:

{ "status": "success or failed", "output_image": "Tensor or None", "message": "error or completed" }

Response:

{ "status": "success or failed", "output_path": "/path/to/generated/image_{timestamp}.png or None", "message": "error or completed" }

The result image will be saved in the "output_path".