ComfyUI utils

This package provides simple utils for:

1. Parsing out prompt arguments, e.g. "a beautiful forest $num_steps=12"
2. Running a workflow in parsed API format against a ComfyUI endpoint, with callbacks for specified events.

It's designed primarily for developing casual chatbots (e.g. a Discord bot) where users can adjust certain parameters and receive live progress updates.

Limitations:

• Only integer arguments are currently supported in addition to the prompt itself. The plan is to add at least floats and strings.
• Only one output from the workflow is supported.

Supports:

• Arbitrary number of integer args embedded in the main string prompt.
• Queuing with a callback when the queue position changes.
• Fetching cached results.
• Reporting intermediate progress of nodes like KSampler.

Install

pip install comfyui_utils

Usage

(better docs are coming, for now please look at the source code / sample script)

from comfyui_utils import gen_prompts, comfy
gen_prompts.make_config("GenVid", [gen_prompts.IntArg("num_steps", default_value=12, min_value=1, max_value=80)])
...
try:
    parsed = gen_prompts.parse_args(raw_prompt, prompt_config)
except ValueError as e:
    print(f"Invalid prompt {e.args[0]}")
prompt_data = ...
class Callbacks(comfy.Callbacks):
    ...
await comfyui.submit(prompt_data, Callbacks())
def on_load(data_buffer):
    ...
await comfyui.fetch(backend_filepath, on_load)

Example

To test the library with a sample SDXL workflow, run the following after installing (replace the address with your ComfyUI endpoint). Make sure your ComfyUI has sd_xl_base_1.0.safetensors and sd_xl_refiner_1.0.safetensors installed (or replace the workflow).

comfy_ui_example_e2e\
  --address='192.168.0.10:11010'\
  --prompt='a smiling potato $base_steps=8$refiner_steps=3'\
  --output='./potato.png'

The single quotes are important so your shell doesn't try to parse the $'s. Expected output:

Queuing workflow.
Queue position: #0
Base...
Base: 1/8
Base: 2/8
Base: 3/8
Base: 4/8
Base: 5/8
Base: 6/8
Base: 7/8
Base: 8/8
Refiner...
Refiner: 1/3
Refiner: 2/3
Refiner: 3/3
Decoding...
Saving image on backend...
Result (cached: no):
{'images': [{'filename': 'ComfyUI_00101_.png', 'subfolder': '', 'type': 'output'}]}

The file will be saved in the root directory.

Use your own workflow

After finalizing the workflow, use the "Save (API format)" button to store the workflow. Then, edit the PromptConfig in the script to reflect the arguments you wish to make available, and ensure the prompt has them replaced after parsing.