https://github.com/light-and-ray/image-edit-bot-open-webui
An closedwebui bot-wrapper around comfyui api, for simple image + prompt workflows
https://github.com/light-and-ray/image-edit-bot-open-webui
ai-image-editing bot comfyui flux-kontext-dev image-editing openwebui qwen-image-edit
Last synced: 18 days ago
JSON representation
An closedwebui bot-wrapper around comfyui api, for simple image + prompt workflows
- Host: GitHub
- URL: https://github.com/light-and-ray/image-edit-bot-open-webui
- Owner: light-and-ray
- Created: 2025-09-03T06:01:30.000Z (10 months ago)
- Default Branch: master
- Last Pushed: 2026-05-24T04:27:25.000Z (about 1 month ago)
- Last Synced: 2026-05-24T06:32:55.742Z (about 1 month ago)
- Topics: ai-image-editing, bot, comfyui, flux-kontext-dev, image-editing, openwebui, qwen-image-edit
- Language: Python
- Homepage:
- Size: 1.13 MB
- Stars: 7
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Qwen/Flux Kontext Image edit (and not only) bot for Openwebui
This is a simple bot for [openclosedwebui](https://github.com/open-webui/open-webui) that allows you to create a channel-wrapper around any [ComfyUI](https://github.com/comfyanonymous/ComfyUI) workflow that has one prompt or one image input, and one image output
This repo is based on two examples: [open-webui/bot](https://github.com/open-webui/bot) and [websockets_api_example.py](https://github.com/comfyanonymous/ComfyUI/blob/master/script_examples/websockets_api_example.py)
## Requirements
You can use the same environment that openwebui uses (it already has all dependencies), or create your own. Dependencies are `pip install dotenv pillow websocket-client python-socketio`
## How to set it up:
- enable channels in openwebui's admin setting
- create an account with admin rights for your bot
- create a channel with your desired name, e.g. `qwen-image-edit`. Make it public or private for the same group where both you and bot are
- have a working comfy ui with your workflow. I assume you already have it
- put text `prompt here` as a prompt (the bot will find the node by this text and replace it with proper prompt)
- delete all preview image nodes and similar, you need to have only one image out
- export it for API, and put the json file inside bot's `workflows/` directory. For example, `workflows/qwen_image_edit.json`
- set up `.env` file: the most important are bot's token and mapping *channel name* -> *workflow name*
- run the bot (`main.py`), openwebui and comfyui
## .env structure
```bash
WEBUI_URL="..." # default http://localhost:8080
TOKEN="..." # see below
COMFY_ADDRESS="..." # default localhost:8188
MAP_CHANNEL_NAME_WORKFLOW="..." # see below
LAZY_IMAGE_URLS=0 # default 0, see below
```
### TOKEN
To get it, log into bot's openwebui profile, go to `Settings` -> `Account` -> `API keys` -> `JWT Token`
Authentication and secure for ComfyUI connection is not supported. If you do it, please make a PR
### MAP_CHANNEL_NAME_WORKFLOW
This is a dictionary in json format that maps workflow and channel name. For example and by default `"{"qwen-image-edit": "qwen_image_edit.json"}"` means the bot will handle all messages from *"qwen-image-edit"* channel, send all user's requests there into *"workflows/qwen_image_edit.json"*. You can add channels and workflows how many you want
### LAZY_IMAGE_URLS
`0` means the bot will send images in base 64 format, `1` - will send url from Comfy's `/view?filename=...`. For the second choice comfy's address should be available from a client and Comfy server should be turned on