Skip to content

jtyszkiew/ImageSmith

BranchesTags

Repository files navigation

gif

ImageSmith

Discord Codecov

Forge your imagination into reality with ImageSmith - A powerful Discord bot that seamlessly integrates with ComfyUI for intuitive image, video, and audio generation.

Note: ImageSmith is a workflow executor, not a workflow creator. You'll need to create your own workflows, but the bot makes them easily accessible through Discord's UI. Check out the example workflows in the repository to get started.

Table of Contents

Quick Start

You need a running ComfyUI instance and a Discord Bot Token.

Important: Default workflows use sd_xl_base_1.0 for image generation and Mochi models for video generation. Ensure these are available in your ComfyUI instance.

Docker

docker run -e DISCORD_TOKEN="<your_discord_token>" \
  # Optional: mount your own configuration file
  # --mount type=bind,source=./configuration.yml,target=/app/configuration.yml \
  # Optional: mount a custom workflows directory
  # -v "./custom_workflows:/app/custom_workflows" \
  ghcr.io/jtyszkiew/imagesmith:latest

From Source

git clone https://github.com/jtyszkiew/ImageSmith.git
cd ImageSmith
python -m venv venv
source venv/bin/activate  # Windows: .\venv\Scripts\activate
pip install -r requirements.txt
cp configuration.example.yml configuration.yml
# Edit configuration.yml with your settings
python main.py

User Guide

ImageSmith exposes slash commands for generating images, videos, and audio from text prompts or existing images. The output type depends on the ComfyUI workflow — the bot automatically detects whether the workflow produces images, video (gifs), or audio and sends the appropriate file to Discord.

Command names are configurable — see Custom Command Names.

Commands

Command Description Parameters
/forge Generate from text prompt, [workflow], [settings]
/reforge Transform existing image image, prompt, [workflow], [settings]
/upscale Upscale with modifications image, prompt, [workflow], [settings]
/workflows List available workflows -

Example Usage

# Basic generation
/forge A majestic mountain landscape at sunset

# Using specific workflow
/forge A cyberpunk city --workflow cyberpunk_generator

# With custom settings
/forge A fantasy character --workflow character_generator --settings "change_steps(30);add_lora('fantasy_style', 0.8)"

Configuration

All configuration lives in configuration.yml. Copy configuration.example.yml as a starting point.

ComfyUI Connection

The comfyui block configures how the bot connects to one or more ComfyUI instances.

comfyui:
  instances:
    - url: 'http://127.0.0.1:8188'
      weight: 1
    - url: 'https://comfyui-2.example.com:8188'
      weight: 2
      auth:
        username: admin
        password: secret
      timeout: 120
    - url: 'https://comfyui-3.example.com:8188'
      auth:
        api_key: your-api-key-here
        ssl_verify: true
        ssl_cert: /path/to/ca-bundle.pem
  input_dir: /path/to/comfyui/input
  show_node_updates: true
  load_balancer:
    strategy: LEAST_BUSY

Authentication

Each instance can have its own auth block. Authentication methods are checked in this order:

  1. API Key — if api_key is set, sends a Bearer token header on both HTTP and WebSocket connections
  2. Basic Auth — if username and password are set (and no api_key), sends a Basic auth header on HTTP connections
Key Type Description
api_key string Bearer token for API key authentication
username string Username for HTTP Basic authentication
password string Password for HTTP Basic authentication
ssl_verify bool Verify TLS certificates (default: true)
ssl_cert string Path to a custom CA bundle PEM file

Workflows

Workflows map Discord commands to ComfyUI workflow JSON files. Each workflow is a named entry under the workflows key.

workflows:
  forge:
    type: txt2img
    description: "Generate realistic images"
    workflow: "./workflows/txt2img.json"
    text_prompt_node_id: "6"
    text_prompt_input_key: "text"
    default: true
    selectable: true

  reforge:
    type: img2img
    description: "Regenerate images"
    workflow: "./workflows/img2img.json"
    text_prompt_node_id: "6"
    image_input_node_id: "10"
    default: true

  txt2vid:
    type: txt2img
    description: "Generate short video"
    workflow: "./workflows/txt2vid.json"
    text_prompt_node_id: "6"
Key Type Description
type string txt2img, img2img, or upscale
description string Shown in the /workflows listing
workflow string Path to the ComfyUI workflow JSON file
text_prompt_node_id string Node ID in the workflow JSON to inject the text prompt into
text_prompt_input_key string Input field name within the prompt node (default: "text")
image_input_node_id string Node ID to inject the uploaded image into (img2img/upscale types)
default bool Global default for its type — the first match wins
selectable bool If false, hides from /workflows and the workflow dropdown (default: true)

Video and audio workflows use txt2img as the type. The bot automatically detects the output format (image, gif/video, or audio) from what ComfyUI returns.

Load Balancing

When multiple ComfyUI instances are configured, the bot distributes work across them.

comfyui:
  load_balancer:
    strategy: LEAST_BUSY
Strategy Description
LEAST_BUSY Picks the instance with the lowest ratio of active generations to weight (default)
ROUND_ROBIN Cycles through connected instances in order
RANDOM Picks a random instance, weighted by each instance's weight value

The weight field on each instance (default: 1) affects both RANDOM (probability weight) and LEAST_BUSY (busyness denominator). A higher weight means the instance receives proportionally more work.

Settings

Settings are Python functions that modify the workflow JSON before it is sent to ComfyUI. They are defined per-workflow.

Two special names are reserved:

  • __before — runs automatically before any user-specified settings
  • __after — runs automatically after all user-specified settings
workflows:
  forge:
    settings:
      - name: __before
        description: "Default workflow configuration"
        code: |
          def __before(workflowjson):
              import random
              workflowjson["4"]["inputs"]["ckpt_name"] = "Juggernaut_X_RunDiffusion.safetensors"
              workflowjson["3"]["inputs"]["seed"] = random.randint(0, 2**32 - 1)

      - name: hd
        description: "HD resolution preset"
        code: |
          def hd(workflowjson):
              workflowjson["5"]["inputs"]["width"] = 1280
              workflowjson["5"]["inputs"]["height"] = 720

Users invoke settings in their command: /forge A fantasy character --settings "hd()"

Multiple settings can be chained with semicolons: "hd();add_lora('fantasy_style', 0.8)"

Forms

Forms collect interactive input from users via Discord modals or select menus before a workflow runs.

workflows:
  forge:
    form:
      - name: seed
        type: text
        required: false
        description: Seed for the model
        message: Provide a seed you would like to use
        on_submit: |
          def on_submit(workflowjson, value):
              workflowjson["65"]["inputs"]["seed"] = value
        on_default: |
          def on_default(workflowjson):
              import random
              workflowjson["65"]["inputs"]["seed"] = random.randint(0, 2**32 - 1)

Field Types

Type Input method Value passed to on_submit
text Modal text input Integer (parsed from input)
textarea Modal text input (multi-line) String (raw input)
resolution Select dropdown List of two integers [width, height] (max 2048)
select Select dropdown (multi-select) List of strings

Field Configuration

Key Type Description Required
name string Name of the form field Yes
type string text, textarea, resolution, or select Yes
required bool Whether the field is required No
description string Displayed as the modal title or dropdown placeholder No
message string Displayed as the input label inside the modal No
on_submit string Python code executed when the form is submitted. Function name must be on_submit. Yes
on_default string Python code executed when a non-required field has no input. Function name must be on_default. No
options list List of {name, value} objects for select and resolution types No

Select Example

- name: style
  type: select
  required: true
  description: Choose art style
  message: Select one or more styles
  options:
    - name: Realistic
      value: realistic
    - name: Anime
      value: anime
    - name: Oil Painting
      value: oil_painting
  on_submit: |
    def on_submit(workflowjson, value):
        workflowjson["3"]["inputs"]["style"] = ", ".join(value)

Resolution Example

- name: resolution
  type: resolution
  required: true
  description: Output resolution
  message: Select a resolution
  options:
    - name: 512x512
      value: 512x512
    - name: 1024x1024
      value: 1024x1024
    - name: 1280x720
      value: 1280x720
  on_submit: |
    def on_submit(workflowjson, value):
        workflowjson["5"]["inputs"]["width"] = value[0]
        workflowjson["5"]["inputs"]["height"] = value[1]

Custom Command Names

Rename the default slash commands to whatever you prefer:

commands:
  forge: "generate"
  reforge: "remix"
  upscale: "enhance"
  workflows: "list"

Any key omitted keeps its default name. For example, to only rename /forge to /generate:

commands:
  forge: "generate"

Display Settings

comfyui:
  show_node_updates: false

When show_node_updates is set to false, per-node "Processing node X..." messages and progress bars are hidden. Completion, error, and media messages still appear normally. The default is true.

Default Workflows per Channel or User

You can set a workflow as the default for specific channels or users. When a user runs a command without specifying a workflow, the bot checks (in order):

  1. A workflow whose default_for.users list contains the user's name
  2. A workflow whose default_for.channels list contains the channel's name
  3. A workflow with default: true for that command type
  4. The first workflow of the matching type
workflows:
  anime_forge:
    type: txt2img
    description: "Anime-style generation"
    workflow: "./workflows/anime_txt2img.json"
    text_prompt_node_id: "6"
    default_for:
      channels:
        - anime-art
        - anime-general
      users:
        - anime_fan_123

Security

Workflow Access Control

Restrict who can use a workflow by adding a security block:

workflows:
  forge:
    security:
      enabled: true
      allowed_roles:
        - "Artist"
        - "Admin"
      allowed_users:
        - "trusteduser"
      allowed_channels:
        - "image-generation"
Key Type Description
enabled bool Enable access control for this workflow (default: false)
allowed_roles list Discord role names that can use this workflow
allowed_users list Discord usernames that can use this workflow
allowed_channels list Channel names where this workflow can be used

When enabled is true, all three checks are ANDed together — the user must satisfy every non-empty list. If a list is empty or omitted, that check is skipped.

Setting Access Control

Individual settings within a workflow can also have their own access control:

workflows:
  forge:
    settings:
      - name: hd
        description: "HD resolution preset"
        security:
          enabled: true
          allowed_roles:
            - "Premium"
          allowed_users:
            - "admin_user"
        code: |
          def hd(workflowjson):
              workflowjson["5"]["inputs"]["width"] = 1280
              workflowjson["5"]["inputs"]["height"] = 720

The same allowed_roles, allowed_users, and allowed_channels keys are available at the setting level.

Internationalization

ImageSmith includes built-in support for multiple languages. All UI strings — embeds, buttons, status messages, errors, forms, and security messages — are fully translatable.

Supported Languages

Code Language
— English (default)
de German
es Spanish
fr French
ja Japanese
pl Polish
pt Portuguese

Configuration

language: "pl"

You can also override individual string keys without creating a language file:

i18n:
  embed:
    titles:
      error: "Custom Error Title"
  bot:
    starting_generation: "Generating your image..."

Strings are resolved in three layers (later layers override earlier):

  1. i18n.yml — built-in English defaults
  2. i18n.<language>.yml — language file (loaded when language is set)
  3. i18n block in configuration.yml — inline overrides

See i18n.yml for the full list of available string keys.

Setting env: "dev" in your configuration causes raw error messages to be shown instead of generic error text. The default is "prod".

Extending

Plugin Development

Plugins are Python classes that extend the bot's functionality. Place them in the plugins/ directory — they are auto-discovered and loaded on startup.

from src.core.plugin import Plugin


class MyPlugin(Plugin):
    async def on_load(self):
        await super().on_load()
        self.bot.hook_manager.register_hook('is.comfyui.client.before_create', self.my_hook)

    async def my_hook(self, workflow_json: dict, instances: list):
        return workflow_json

Every plugin receives self.bot — the ImageSmith instance — which provides access to:

  • bot.hook_manager — register and fire hooks
  • bot.workflow_manager — access workflows and configuration
  • bot.comfy_client — the ComfyUI client
  • bot.form_manager — register custom form field handlers

Available Hooks

Hook When it fires
is.comfyui.client.before_create Before the ComfyUI client is created
is.comfyui.client.after_create After the client is created and connected
is.comfyui.client.instance.timeout When an idle instance times out
is.comfyui.client.instance.reconnect Before attempting to reconnect an instance
is.security.before Before the security check runs
is.security During the security check (return SecurityResult to allow/deny)

Development

Testing

pip install pytest pytest-asyncio pytest-mock pytest-cov
pytest tests/ -v --cov=./

Contributing

  1. Fork the repository
  2. Create feature branch: git checkout -b feature/AmazingFeature
  3. Commit changes: git commit -m 'Add AmazingFeature'
  4. Push to branch: git push origin feature/AmazingFeature
  5. Open a Pull Request

License

Licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

Disclaimer

This bot is for educational and creative purposes. Users are responsible for ensuring their usage complies with ComfyUI's and Discord's terms of service.

Community

Join our Discord server to see the bot in action and stay updated with the latest developments!

About

ComfyUI Discord Bot - Share workflows with your friends through Discord!

Topics

discord-bot ai-chatbot stable-diffusion comfyui comfyui-nodes

Resources

Readme
Activity

Stars

14 stars

Watchers

2 watching

Forks

9 forks
Report repository

Releases 2

0.0.2 Latest
Aug 6, 2025
+ 1 release

Packages

 
 
 

Contributors 3

Languages