Template Overview

This template is basically just a character map showing you what kinds of details you should or could include, and how to do so without using an excessive number of tokens. See the Understanding Tokens section for information about tokens and why the count is important.

Each Chub.ai character relies on the use of several different sections to organize character data to be used as efficiently as possible. Knowing what kind of information and instructions to put in each section (and how they're used) is crucial to achieve the functionality you want.

Description

The most important content of this template is in the Description section. Put simply, the Description is the core of your character and can be written and formatted in countless possible ways, with at least a dozen commonly used methods. They can be very basic and simple, or incredibly long and complicated. Because of how LLM-based AI chatbots work, the Description is one of those "No wrong answers" kind of things - but some "answers" are definitely better than others.

Typically, the description is a combination of the following broad categories:

• Personality
• Appearance
• Background
• Scenario context
  

Understanding the template's format

The format I used is one of many commonly used formats when it comes to creating custom AI chatbots. I prefer this specific format because it's a nice balance between being concise and being descriptive. There are two sections which I'll be referring to as the Character Details and the Character History.

How to use Character Details

The character details use a format that is very loosely based on data file formats commonly used in programming, such as JSON and YAML. The entire section is contained within a single set of square brackets []. It's important that you don't use additional brackets ([] or {}) unless you know what you're doing because it could confuse the AI.

Categories and Data Lists

This section is broken into lists with comma-separated character traits. Each list starts with the list title followed by a colon and ends with a semicolon, like this:

Personality:
happy,
quirky,
inquisitive;

Multi-word List Items

You can also use small sentences in your lists, just make sure they don't include any additional punctuation such as: . , : ; ! ?

Here's an example of multi-word list items:

Likes:
reading,
fast food,
going to the park on the weekends,
dancing;

So many lines, so much ADHD...

You will also notice there are an unusual number of line breaks. Because of my ADHD, I find it easier to work with lists than paragraphs, so that's how I chose to format it. Some other AI services don't do well with excessive line breaks, but after months of testing, I haven't found Chub.ai to have any difficulty with them (using their Mercury and Venus models). You can leave the format as-is, or you can condense it. For example:

Original

Kinks:
item,
item
;

Condensed Example 1

Kinks:
item,
item;

Condensed Example 2

Kinks: item, item;

The Key Character Details

Keep in mind, the first few lines are formatted almost like they are single-item "lists". This was intentional, and I recommend you keep it that way, because these details are what you could call the "key details" for most characters.

Adding and Removing Categories

You can also add and remove lists/categories as needed. Here's an example of how someone might choose to change the key details:

First Name: Amy;
Last Name: Smith;
Age: 22;
Sex: Female;
Orientation: Heterosexual;

Here are some examples of other categories I've seen or used in the past as examples of what you could add:

• Mannerisms
• Fetishes (can be separate from kinks)
• Clothing/Outfits
  

How to use Character History

The area below the [Character Details] is much simpler, and while I decided to name it "Character History" for the sake of simplicity, it's not limited to just the character's past. Basically, this is where you write a summary of the character's history, explain how they might behave, their relationship to the {{user}}, context for the current/potential scenarios, etc.

This section should be written like you would write an actual summary for an essay. Typically, I usually write about 1-3 paragraphs here, but they can be much longer if you really want (though it's always better to keep it simple).

What are Tokens?

The subject of tokenization is extremely complicated as a whole, but here are the basics you need to know:

• Tokens are what all Large Language Model (LLM) AI models run on. It's how they process information, generate new data, and formulate responses.
• A token can be a whole word, part of a longer word, or even just a single character (often punctuation).
• All LLM AI models have a hard limit of how many tokens they can handle per request (message).
  

What is the Token Count?

The number of tokens per request is a combination of the following factors:

1. System and Character Prompts
   These are iinstructions that are always included with each message you send, but are never actually seen by you in the chat.
2. Conversation History
   This includes pieces of information collected from the chat history, often in the form of a "rolling memory" (the AI's short-term memory).
3. The content of your most recent message
4. The content of the AI's response
   

Why the Token Count is important

Put simply, every time you send a message, you're initiating a new request. The input of that request is the combination of steps 1 through 3. When you subtract that total number of tokens from the AI's token limit, that's how many tokens the AI will have to generate the output of that request. Among other things, too many tokens in your input can result in very short responses, or responses that suddenly get cut off.

Smaller models generally have limits ranging from around 2,000 to 5,000 (which is usually what you'll get from most free models and services). Larger, newer models can have limits of 8,000 and more (and they usually cost money).

Tokens TLDR

Just keep in mind, for most AI bots, you should do your best to follow the golden rule of "Keep it simple, stupid". As you explore more characters written by other people, you will see many with very long and detailed descriptions. This can be totally fine, and depending on what LLM you're using, those kinds of long descriptions can even be highly effective. But they always come at a cost, and that cost is usually the effective "short term memory" of the chatbot.

When making your own templates...

For anyone making templates of their own based on this template, I only ask that you credit me by leaving the Credits section at the very end of the Creator's Notes untouched.

When making your own characters...

For anyone using this template for it's primary purpose of creating your own custom AI chatbot characters, there is no need to credit me at all (though I would still appreciate if you did).