Uniting text-generation-webui and Stable Diffusion (A1111 / Forge / ReForge) for casual fun, creativity & professional use.
- Both text-generation-webui and Stable Diffusion are optional.
- The features of both can be independently enabled/disabled in the main config file.
For support / discussion, visit the dedicated ad_discordbot channel on the TextGenWebUI Discord server.
What's new | Features | Installation | Usage | Updating
click to expand
04/04/2025: Major rewrite of core install logic (can now be installed as a Standalone!) Update Wizards!
This bot has always required text-generation-webui to function. Now, the bot may installed as a standalone, or optionally with TGWUI integration.
This update also replaces the Update scripts with Update Wizards. From the wizard it is possible to change between Standalone and TGWUI integration.
08/19/2024: Major User Settings Enhancement: Per-Server Settings!
There is now an option to enable "Per-Server" settings:
- Settings for new servers will be initialized as a copy of the "main" settings
- All settings are managed separately (EXCLUDING
config.yaml) - There are sub-options for "per-character settings" and "per-server ImdModels"
- Note that
basesettings.yamlapplies to ALL settings.
08/12/2024: New /prompt command! Enhanced Post Active Settings feature!
There is now a /prompt command to add some advanced options to your message request.
Also, the Post Active Settings feature received a nice overhaul and is definitely worth trying out!
Requires using yet another new command, /set_server_settings_channel, in each server that you would like to share settings in.
This feature will automatically post a copy of your settings to the channel (EXCLUDING config.yaml)
08/02/2024: New Behavior - Streaming Responses!
-
Base Behavior / Character Behavior now have option to stream responses!
-
You no longer have to wait for the entire message to generate!
-
Actually works with all the other crazy features!
07/24/2024: New Behaviors to bring your characters to life!
Some new settings appeared recently... which were slightly buggy. Today, I announce that they are working very well!
These new behaviors can make your character more humanistic (or just be like a computer program, by default)
- Character can go idle, taking more time to respond
- Pause to "read messages" before writing
- Simulate typing at different speeds (does not actually generate text slower)
- *DOES NOT* throttle the bot's performance! All tasks running full speed on the backend!
- These behaviors will continue to improve in the coming days/weeks
07/02/2024: New Behaviors "Spontaneous Messaging"
- Character can now be silently prompted to say something after inactivity
- A few settings control the behavior
- It can also serve as an "Auto-prompter" generating unlimited responses
06/18/2024: Per-Server Voice Channels (TTS)
- New command to set voice channels: /set_server_voice_channel
- The context menu commands (regen / continue / edit history) vastly improved
- The bot now reacts to messages to identify their status (hidden, etc)
06/14/2024: Direct Messages Support. Image model changes.
- The bot now supports Direct Messages, which have their own history!
- Most commands are disabled in DMs.
- Direct messaging can be disabled via config.yaml
- Img model "override settings" (checkpoint, vae, clip skip) are no longer saved.
06/11/2024: More Context Menu Commands!
06/06/2024: Context commands are now AMAZING
- "Continue" and "Regen" have been revised to work AMAZINGLY well!
- New "Edit History" command has been added, that also works AMAZINGLY well!
- You'll have to try these for yourself to see just how perfect they work.
Right click any message to invoke these context commands
06/04/2024: Launchers! Updaters! New History Manager!
Shoutout to @Artificiangel who coded an amazing new History Manager. Great new features are coming!
- We now have launcher and updater scripts (...look familiar?)
- New CMD_FLAGS.txt to add your custom launch flags (...look familiar?)
- bot.py no longer has to be moved! Correct location is now in main 'ad_discordbot' folder
- New history manager is much more flexible and unlocks new possibilities
05/28/2024: Per-Channel History!
- New setting 'per_channel_history' enables all channels to have their own chat history.
- Custom logging format puts chat history, server name, and channel name under each channel.id key.
- A utility .bat file is now included to split these custom logs into normal logs, if needed.
- New '/announce/ command allows channels to be assigned as Announcement channels.
Announcement channels will receive Model / Character change announcements
instead of interaction channels.
05/22/2024: Big Update - Easier to Update Moving Forward
Quick shoutout to @Artificiangel who has recently joined development and made stunning contributions.
- The directory '/internal/' which contains persistent settings (not intended to be modified by users)
is no longer part of the bot package. Instead, '/internal/' and its contents are created dynamically if missing.
- User settings are now present in a '/settings_templates/' which will be automatically copied into the root directory,
if not done manually by users. This allows the bot to be easily updated without conflicts due to modified files.
05/16/2024: Significant Changes to File Structure
- The main bot script has grown massive, so it is now split to modules (new '/modules/' subdirectory)
- activesettings.yaml is now in an '/internal/' subdirectory. Your settings will migrate automatically.
- 'bot.db' has been superceded by a 'database.yaml' file. Your settings will migrate automatically.
- Note: Changes to 'dict_base_settings.yaml' and 'activesettings.yaml' are just comment updates
04/29/2024: Huge Improvement to ControlNet in /image Command
- Options are now dynamically filtered and populated using the '/controlnet/control_type' API endpoint
- Sets almost identical default options as in SD WebUI interface
- This took a lot of time and effort, please try it out!
04/21/2024: Revamped Main Config. Made textgenwebui and SD WebUI Optional!
- replaced config.py with config.yaml
- Config.py will still work, but is now unsupported and will receive no updates.
- Textgenwebui and SD WebUI are now optional elements of the bot that can be disabled in config.yaml
04/19/2024: Overhauled Img Model handling. Now "API" Method only.
At first, there was only the '.YAML method' - each model required its own definition.
Later, fetching models via API became a secondary option.
Now, I noticed that all the improvements to API method have made the 'YAML method' obsolete:
- Filter / Exclusion settings to control models that get loaded
- Sophisticated calculations for 'Sizes' menu in '/image' command
- Apply model settings and 'Tags' based on intelligent filter matching
- Now, additional check for 'exact_match' if necessary.
Please take care migrating to this change:
- Fetch the new version of 'dict_imgmodels.yaml' which is now the settings panel for the API method.
- Migrate your 'imgmodel' settings from config.yaml
- Delete the whole 'imgmodels' block in config.yaml!
04/18/2024: New Feature: Dynamic Prompting.
Works exactly very similarly to the SD WebUI extension sd-dynamic-prompts
04/16/2024: Enhanced Flows and "Instant Tags". Many other improvements.
- Changed the 'Logging Level' from DEBUG to INFO - LESS SPAM!
- Performance may be more optimized... all settings were being stored in the discord client object.
Now, they are stored in a dedicated class object.
- Characters can now be omitted from /character command with new parameter (see M1nty example char)
- The feature to create tags instantly from your text has been upgraded.
ANY tag values can be created including dictionaries, lists, sublists... anything.
- The SD API "Guess imgmodel params" feature has much better success rate now.
- "Flows" feature can now use variables for tag values.
- Added a new "Flows" example in 'dict_tags.yaml'
- Added new forge-couple param.
- Better error handling when failure to change Img model
- Fixed img prompt insertions from Tags sometimes creating a line break.
04/12/2024: Changed user images dir. New Tags. Enhanced Image Selection.
- All images now go into a root 'user_images' folder.
There are no longer separate root folders for ReActor, ControlNet, Img2Img, Inpainting masks, etc.
Users can organize their images in 'user_images' however they wish - just include path in Tags values.
- New tags:
- 'img2img'. Previous commit added img2img to /images cmd - now, it's also a Tag.
- 'img2img_mask' (inpainting)... and now also added to /image command!
- 'send_user_image' - can send a local, non-AI generated image! Can be triggered to send an image after LLM Gen and/or after Img Gen.
- Enhanced image selection:
When processing images from tags (ReActor, ControlNet, etc etc), if the value is a directory (does not include .jpg/.png/.txt),
the function will recursively attempt to select a random image - if no images are in the directory, it will try picking a random directory, and so on,
until an image is found or reaches an empty directory (error). So rather than just picking a random image from a folder, it can now pick from a random folder.
04/10/2024: Upgraded '/image' cmd. Added Tags. Added sd-forge-couple extension support.
- Upgraded the /image command:
- ControlNet and ReActor now only appear in the select options if enabled in config.yaml
- 'img2img' has been added. If an image is attached, it will prompt for the Denoise Strength.
- ControlNet now follows up asking for model/map if an image is attached, to simplify the main menu.
- Added 'sd_output_dir' tag, so now you can control the image save location in your tag definitions.
- Added extension support for SD Forge Couple. Currently only useful for '/image' command unless you can get the LLM to reply with the correct format (I'm working on that!)
04/01/2024: Pretty massive update. Be sure to update textgen-webui, and take care updating settings files.
- Overhauled SD WebUI extension support (ControlNet, layerdiffuse, ReActor) to be much more powerful and manageable.
- ControlNet support received a massive update in particular... multi-ControlNet is even supported!
- These extensions each have a simple primary Tag to activate and apply.
- **ALL** of their parameters are now easily controlled by the Tags system.
- Added a new method to create Tags on-the-fly using this syntax: [[key:value]] or [[key1:value1 | key2:value2]]. These go into immediate effect, particularly useful for controlling the extension settings.
- 4 older parameters were recently removed from textgen-webui - mirrored in this bot update.
03/07/2024: layerdiffuse support added to Tags feature!
02/13/2024: Major update introducing the Tags feature. Take care migrating your existing settings
12/11/2023: New "/speak" command! Silero and ElevenLabs TTS extensions now supported!
12/8/2023: TTS Support, and Character Specific Extension Settings now added!
-
Chat History for each channel!
- Each channel has its chat history maintained seperately and cleanly.
- History is gracefully loaded for all channels on startup.
/reset_conversationwill reset history in interaction channel only.
-
Robust "Tags" system to manipulate bot behavior persistently or via trigger phrases, including:
- Trigger Text and/or Image response
- Image censoring settings (None / Spoiler / Block)
- Powerful ControlNet, ReActor, Forge Couple, and layerdiffuse integration
- Automatically apply loractl scaling (Currently ReForge/A1111 only)
- Swapping / Changing LLM characters (ei: "draw... " can trigger character tailored for image prompting)
- Swapping / Changing LLM models and Img models
- Modifying LLM state (temperature, repetition penalty, etc) and image gen settings (width, height, cfg scale, etc)
- Keep things spicy by factoring random variations to LLM state / img gen settings
- Modifying the user's prompt and LLM's reply with delete/insert/replace text, img LORAs, etc
- Manipulate LLM history (Suppress history, limit history, save or do-no-save reply, etc)
- New Feature: Flexible, Instant-Tags by using syntax [[key:value]] or [[key1:value1 | key2:value2 ]]
- New features being added frequently and easily due to the framework of the "Tags" system
-
TTS Support!
- alltalk_tts, coqui_tts, silero_tts, edge_tts and elevenlabs_tts
- Bot can speak on Voice channel, upload copy of audio file, or both!
- Per-character TTS settings! Give each character a unique voice!
-
Sophisticated function to send text responses over Discord's 2,000 character limit
- "chunks" messages by looking back to nearest line break or sentence completion.
- Preserves syntax between chunks such as bold, italic, and even
code formatting
-
- /character - Change character
- /reset_conversation - Starts a new conversation in the current channel.
- /image - Allows more controlled image prompting (positive prompt, neg prompt, Size settings, ControlNet, ReActor)
- /speak - Bot can speak any text, using any voices (including user attach .mp3 or .wav for alltalk_tts)!
- /imgmodel - Change image model along with any custom settings for it
- /llmmodel - Change LLM model
- /set_X_for_server - A number of commands to assign a specific channel in your servers (Voice channel for TTS, Announcement channel, Settings channel, Starboard channel, etc).
- Plus many more!
-
Dynamic settings handling:
- Core bot settings managed in
config.yaml(bot behavior, discord features, extensions, etc.) - The "Tags" system is configured in
dict_tags.yaml(global Tags, default Tag params, Tag presets, etc.) - Foundational layer of user settings configured in
base_settings.yaml. - Character files can include custom Tags, TTS settings, LLM state parameters, and special behaviors, which prioritize over basesettings.
- Custom Image models settings defined in
dict_imgmodels.yaml(Tags, payload params) which prioritize over basesettings. - All user settings commit to
internal/activesettings.yaml, which serves as a dashboard to view the bot's current state.
- Core bot settings managed in
-
Automatic Img model changing:
- Adjustable duration and mode (random / cycle)
- Smart filters and settings to auto-update relavent settings (SD1.5, SDXL, Turbo, etc)
-
Powerful Context Menu commands (right-click on a message):
- Can target any message in chat history (not only the most recent one!), making Continue & Regenerate more powerful than native TGWUI.
- Custom Regenerate methods:
- regenerate replace works as everyone is familiar with (generates a new reply and replaces the original).
- regenerate create makes a new generation while "hiding" the previous response. These can be easily toggled with toggle as hidden command, so you can choose your favorite reply!
- edit history allows you to edit any message in history
- toggle as hidden will hide or reveal the user/bot reply exchange in history.
- The above are restricted to your own / the bot's messages (can't target other users' messages).
-
All tasks queue up and process elagently - go ahead, spam it with requests!
-
Built in Starboard feature
-
Feature to post current settings in a dedicated channel
-
ALWAYS MORE TO COME
-
OPTIONAL FOR TGWUI INTEGRATION: Install oobabooga's text-generation-webui
-
Create a Discord bot account, invite it to your server, and note its authentication token.
!!! IMPORTANT: You must allow your bot to have "Privileged Intents" for "
MESSAGE_CONTENT" !!!!!! Enabling Privileged Intents !!!
-
Clone this repository anywhere.
FOR TGWUI INTEGRATION: clone into /text-generation-webui/
git clone https://github.com/altoiddealer/ad_discordbot -
Run the launcher for your OS (ex: start_windows.bat)
-
Enter your bot token (from Step 2) into the CMD window
-
The bot should now be up and running!
If a Welcome message does not appear in your main channel, you can use
/helpmenuto see it.A number of user settings files will appear (copied in from
/user_settings/) where you can customize the bot.config.yaml,dict_base_settings.yaml,dict_cmdoptions.yaml,dict_imgmodels.yaml,dict_tags.yaml
- Run the launcher for your OS (ex: start_windows.bat)
- Optionally add launch flags to CMD_FLAGS.txt
EXAMPLE CMD Flags:
--loader exllama --model airoboros-l2-13b-gpt4-2.0-GPTQ
- In Discord UI, use command /character to choose a character.
-
@ mention the bot
-
Use command /main to set a main channel. The bot won't need to be @ mentioned in main channels.
-
If you enclose your text in parenthesis (like this), the bot will not respond.
(A1111 or sd-webui-forge must be running!)
-
By default, starting your request with "draw " or "generate " will trigger an image response via the Tags system (see
dict_tags.yaml) -
Use /image command to use your own prompt with advanced options
-
Install your TTS extension.
Follow the specific instructions for your TTS extension!!
Example instructions for coqui_tts:
Run the .cmd file in text-generation-webui directory (ex: cmd_windows.bat), and performing the following commands:
Linux / Mac:
pip install -r extensions/coqui_tts/requirements.txtWindows:
pip install -r extensions\coqui_tts\requirements.txt -
Ensure that your bot has sufficient permissions to access the Voice channel and/or upload files (From your bot invite/Discord Developer portal, and your Discord server/channel settings)
-
Configure config.yaml in the section discord > tts_settings
-
If necessary, model file(s) should download on first launch of the bot. If not, then first launch textgen-webui normally and enable the extension.
-
Your characters can have their own settings including voices! See example character M1nty for usage
- Run the update-wizard for your OS (ex: update_wizard_windows.bat)
