tags - altoiddealer/ad_discordbot GitHub Wiki
The 'tags' feature is a versatile method to tailor the bot's behavior for your use case.
Tags may be defined in dict_tags.yaml for global effect, as well as in character cards and imgmodel definitions.
Notes:
- 'trigger' key is recommended for most tags (except 'global_tag_keys' and 'tag_preset_name').
- Multiple comma-separated trigger phrases can be used per each tag (trigger: 'a boy,a dog' can match either 'a boy' or 'a dog')
- You may simplify triggers using this syntax: 'a {boy|dog}' --- the script will read this as 'a boy,a dog'
- You may use multiple trigger keys in a tag definition such as 'trigger1', 'trigger2' - in which case, a trigger phrase must match the search text from each one
- The behavior of tags listed higher may override ones listed lower. Sources are also prioritized: character > imgmodel > base_tags
- Valid tag keys are detailed at the end of this file and on GitHub
CLICK TO EXPAND
You can create/activate one time use Tags on-the-fly by using the syntax [[ key1 : value1 | key2 : value2 | etc... ]] or [[ key: value ]]
EXPAND FOR SCREENSHOT
- Very flexible with syntax:
- keys / values can be wrapped in 'quotes' or "double quotes" or neither, it does not matter
- White space can be included or not, it does not matter
- The matched syntax is removed from your text
- These tags are immediately activated
- Works for ANY tag value, including strings (text), booleans (True/False), {Dictionaries}, [Lists], and even [[sub][lists]]
| tag name | value type | description | default |
|---|---|---|---|
| search_mode | string | Search method for trigger matching. 'user' = user prompt only / 'llm' = bot reply only / 'userllm' = search both for matches | userllm |
| trigger | string | Recommended for most tags (except 'global_tag_keys' and 'tag_preset_name'). | None |
| Additional triggers (trigger1, trigger2, etc) | string | if using multiple trigger keys, a trigger phrase must match the search text from each one. The key names just need to start with 'trigger' and be unique, the order is unimportant | None |
| only_with_tags | [list of strings] | A list of tag names (see name param). This condition will be True if any named tag was matched. (ex, only_with_tags: ['imgmodel', 'another tag name']) |
None |
| random | float | Chance for tag to process. 0.0 = 0% chance and 1.0 = 100% chance. | 1.0 |
| case_sensitive | True/False | If the triggers are case sensitive | False |
| on_prefix_only | True/False | If triggers must be matched at beginning of the searched text | False |
To access the "id" needed for the following conditions, you must enable Developer Mode (warning: Ads) for your account (Account Settings > Advanced > Developer Mode). |
These can be either one id, example: int / or a comma separated list, example: [int, int, int] | ||
| for_guild_ids_only | int | Value is a 'server_id' number. Get this by right-clicking a Server and selecting "Copy Server ID". | None |
| for_channel_ids_only | int | Value is a 'channel_id' number. Get this by right-clicking a channel (or DM) and selecting "Copy Channel ID" | None |
| persist | int | Tags matched with persist will automatically re-apply for that many future requests, in the same channel, during the same "match phase", before the tag is checked normally again. Ignores triggers, but NOT random. persist: -1 will make triggered tag persist forever. |
None |
| tag name | value type | description | default |
|---|---|---|---|
| name | string | Used in print statements and some tag interactions. Required when using a 'persist' tag. | None |
| send_user_image | string | Sends a local image to the channel. Value values: Image in '/user_images/' (png/jpg/txt in base64) -OR- Directory name in '/user_images/' (picks random img) | None |
| call_api | dictionary | Refer to 'dict_api_settings.yaml'. Required: {endpoint_name: The Endpoint Name}. Recommended to include: {api_type: (imggen/ttsgen/textgen)} for 'main' APIs, or {api_name: The Client Name} for all others. Any endpoint params can be overridden in this dict. Payload can be updated with {input_data: {key: value, ...}}. | None |
| run_workflow | dictionary | Refer to 'dict_api_settings.yaml'. Runs a predefined workflow. Valid usage: {name: Name of the Workflow, (optional)input_data: {key1: value1, key2: value2, etc}}. For 'call_api' and 'run_workflow': Any string values can contain bot formtting syntax, ex: "{prompt}" will be replaced with the actual user's text. | None |
| flow | dictionary | A dict of 'flow_steps'. When a 'flow' tag is triggered, it creates a message loop for each defined 'loop_step' | None |
| flow_step | list | The tags you want to apply in each flow step. Without a trigger, they will auto-apply. 'format_prompt' is very useful in each flow_step | None |
| flow_step_loops | integer | The number of times you want to loop a particular flow_step. | 0 |
Relevant to text generation (LLM):
| tag name | value type | description | default |
|---|---|---|---|
| should_gen_text | True/False | If LLM should generate text. If False and 'should_gen_image' is True, passes the user's text as prompt for image gen. | True |
| should_send_text | True/False | If LLM should send generated text to channel. Mainly used for 'flows' tag. | True |
| llm_censoring | True/False | True = blocks text from being generated | False |
| begin_reply_with | string | The LLM will continue their reply from this. | None |
| change_llmmodel | string | For (change/swap)_llmmodel: 'change' will trigger a persistent change. 'swap' will only change the LLM model for the current interaction. | None |
| swap_llmmodel | string | Name of an LLM model (ei: 'Mistral-7B-OpenOrca-GPTQ'). Value 'None' will unload the LLM model! | None |
| time_offset | integer/float | 0 = today's date (system time). -0.5 shifts the current date to be 12 hours ago. (UOM = days) | 0.0 |
| time_format | string | Examples: '%H:%M:%S' (HH:MM:SS), '%I:%M:%S %p' (12HR:MM:SS AM/PM) | '%H:%M:%S' |
| date_format | string | Examples: '%Y-%m-%d' (YYYY-MM-DD), '%A, %B %d, %Y' (DayName, MonthName DD, YYYY) | '%Y-%m-%d' |
| format_prompt | string | Useful for adding text before and/or after your prompt to the LLM. If you do not include {prompt}, your prompt will be replaced entirely with the value of 'format_prompt'. Can include any variables, see '/tips_and_info/Message Variables.txt' | None |
| insert_text | string | Text you may want to insert relative to the matched search text. | '' |
| insert_text_method | string | Relevant in tags with key 'insert_text'. 'after' = insert value after matched text / 'before' = before text / 'replace' = replace text | after |
| text_joining | string | Relevant to all insertion methods except "replace" | ' ' |
| change_character | string | Filename of character in /characters (do not include extension). That character's name, context, and state parameters will be used in LLM payload. | None |
| swap_character | string | For (change/swap)_character: 'change' will trigger a persistent change. 'swap' will only change the character for the current interaction. | None |
| filter_history_for | string | Applies before other "history" tags. Filters the history based on the name of a user or a (character/swap character). Includes the user/assistant exchange for matched items. | None |
| load_history | integer | 0 = default (all history included) / -1 = prompt excludes chat history / > 1 = llm only sees this many recent exchanges. | 0 |
| save_history | True/False | Whether to save this interaction to history or not. | True |
| include_hidden_history | true/false | If true, the history for this interaction will include ones previously hidden via 'save_history' tag, or toggled off via 'toggle as hidden' App Command | False |
| state | {dictionary} | Update values for textgen-webui state parameters. Must be formatted as a dictionary. | None |
| prefix_context | string | Inserts text immediately before the 'context', joined by \n (new line). Multiple can accumulate. | None |
| suffix_context | string | Inserts text immediately after the 'context', joined by \n (new line). Multiple can accumulate. | None |
| llm_param_variances | {dictionary} | Randomization ranges for textgen-webui state parameters. Must be formatted as a dictionary. | None |
Relevant to Img generation:
| tag name | value type | description | default |
|---|---|---|---|
| sd_output_dir | string | Location to save any generated images. | 'ad_discordbot/sd_outputs/' |
| should_gen_image | True/False | If bot should generate an image using Stable Diffusion. The LLM's reply will be used as the img prompt, unless no LLM Model is loaded or 'should_gen_text' is False - in which case user message is prompt. | False |
| should_send_image | True/False | If bot should send generated image to channel. Mainly used for 'flows' tag. | False |
| change_imgmodel | string | For (change/swap)_imgmodel: 'change' will trigger a persistent change. 'swap' will only change the Img model for the current interaction | None |
| swap_imgmodel | string | A valid model name/value | '' |
| img_censoring | integer | 0 = disabled / 1 = blur image / 2 = block image from being generated | 0 |
| positive_prompt | string | Text you may want to insert relative to the matched text in the image prompt. | None |
| positive_prompt_method | string | Relevant to 'positive_prompt'. 'after' = insert value after matched text / 'before' = before text / 'replace' = replace text | after |
| positive_prompt_prefix | string | Insert text at the beginning of the image positive prompt | None |
| positive_prompt_suffix | string | Insert text at the end of the image positive prompt | None |
| negative_prompt_prefix | string | Insert text at the beginning of the image negative prompt | None |
| negative_prompt | string | Insert text at the end of the image negative prompt ('_suffix' also works) | None |
| img_text_joining | string | Relevant to all insertion methods except "replace" | ', ' |
| payload | {dictionary} | Replacements for active payload. Must be formatted as a dictionary. | None |
| aspect_ratio | string | This a shortcut to dynamically set the payload width/height dimensions based on the current average dimensions. Valid values examples: '4:3', '16:9', '2:3', etc | None |
| img_param_variances | {dictionary} | Randomization ranges for active payload parameters. Must be formatted as a dictionary. | None |
| img2img | string | Uses a local image as img2img input. Value values: Image in 'user_images' (png/jpg/txt in base64) -OR- Directory name in '/user_images/' (Picks a random image or random image from random subdirectory) | '' |
| img2img_mask | string | AKA Inpainting. Won't have any effect without img2img also triggered. Should be a greyscale image. Value values: Image in 'user_images' (png/jpg/txt in base64) -OR- Directory name in '/user_images/' (Picks a random image or random image from random subdirectory) | '' |
Relevant to Img generation (A1111/Forge/ReForge - not yet Supported for ComfyUI):
| tag name | value type | description | default |
|---|---|---|---|
| controlnet | string | Simplified tag to both apply ControlNet image/folder AND enable ControlNet. | None |
| controlnet# | string | Optional syntax for multi-ControlNet where '#' represents the ControlNet unit. ex: 'controlnet0: guy.png', 'controlnet1: architecture', etc. Valid values: Image in 'user_images' (png/jpg/txt in base64) -OR- Directory name in '/user_images/' (Picks a random image or random image from random subdirectory) | None |
| cnet_{key} | string | ANY key can be used (see dict_base_settings.yaml for complete list). A 'controlnet' tag must also be in effect when using these tags in order to properly assign the image. | None |
| cnet#_{key} | string | Optional syntax for multi-ControlNet parameters where '#' represents the ControlNet unit. ex: 'cnet0_weight: 0.5', 'cnet1_guidance_end: 0.8', etc. | None |
| cnet#_mask_image | string | The value for the 'cnet_mask_image' param will be handled exactly the same as the main controlnet (see writeup for 'controlnet' a few lines up) | None |
Relevant to Img generation (Forge ONLY):
| tag name | value type | description | default |
|---|---|---|---|
| forgecouple | string OR [list] | Simplified tag to both apply AND enable Forge Couple. For Basic mode, use string: 'Horizontal' / 'Vertical'. For Advanced mode, use a list of lists (refer to dict_base_settings.yaml) | None |
| couple_{key} | ANY key can be used (see dict_base_settings.yaml for complete list). Note: Forge Couple will not be enabled without a 'forgecouple' tag. | None |
Relevant to Img generation (Forge ONLY):
| tag name | value type | description | default |
|---|---|---|---|
| layerdiffuse | string | Simplified tag to both apply AND enable layerdiffuse. Note: 'laydiff_method' works exactly the same. | None |
| laydiff_method | string | Valid values for either 'layerdiffuse' or 'laydiff_method': "(SD1.5) Only Generate Transparent Image (Attention Injection)" "(SD1.5) From Foreground to Background (need batch size 2)" "(SD1.5) From Background to Foreground (need batch size 2)" "(SD1.5) Generate Everything Together (need batch size 3)" "(SDXL) Only Generate Transparent Image (Attention Injection)" "(SDXL) Only Generate Transparent Image (Conv Injection)" "(SDXL) From Foreground to Blending" "(SDXL) From Foreground and Blending to Background" "(SDXL) From Background to Blending" "(SDXL) From Background and Blending to Foreground" |
None |
| laydiff_{key} | string | ANY key can be used (see dict_base_settings.yaml for complete list). Note: layerdiffusion will not be enabled without either a 'layerdiffuse' or 'laydiff_method' tag. | None |
Relevant to Img generation (A1111/Forge/ReForge - not yet Supported for ComfyUI):
| tag name | value type | description | default |
|---|---|---|---|
| reactor | string | Simplified tag to both apply face image/folder/model AND enable ReActor. Valid values: Image in 'user_images' (png/jpg/txt in base64) -OR- Directory name in '/user_images/' (Picks a random image or random image from random subdirectory) -OR- Face model in '{SD_CLIENT}\models\reactor\faces' example 'M1nty.safetensors' | None |
| reactor_{key} | string | ANY key can be used (see dict_base_settings.yaml for complete list). Recommended to use in combination with 'reactor' tag - otherwise, you can't use "single face image method" (technically, you could send a base64 string... and you would need to include 'reactor_enabled: True') | None |
| reactor_mask | string | Value values: Image in 'user_images' (png/jpg/txt in base64), directory, etc. Should be a greyscale image, which will mask the face swap result. | None |