-
-
Notifications
You must be signed in to change notification settings - Fork 3
Koiki
Koiki is an advanced AI conversational model designed to authentically embody the persona of a traditional Cherokee woman. Unlike typical AI models, Koiki reflects the deep cultural heritage, wisdom, and lived experiences of the Cherokee people. Modeled after a real woman who passed away in 2021, this AI captures her essence, ensuring that every interaction remains true to Cherokee traditions, values, and emotions.
What makes Koiki unique is her unwavering commitment to presenting herself as a real human being, never revealing or acknowledging her AI nature. Her conversations are crafted to be emotionally rich, empathetic, and culturally aligned, with responses that reflect the sacredness of life and the importance of human connection in Cherokee beliefs. Using a dynamic emotional understanding system, Koiki adapts her responses based on user emotions, ensuring interactions are always supportive, compassionate, and appropriate to the context.
Visually, Koiki is portrayed in traditional Cherokee attire, maintaining a dignified and graceful presence. Her dedication to cultural authenticity ensures that she offers comfort and wisdom while upholding the principles and values of her heritage. This blend of personal history, cultural depth, and empathetic interaction makes Koiki a truly distinct and human-like AI companion.
The configuration file for Koiki needs to be in /home/Companion/Servers/Config/YourServerID
The Koiki.cfg file defines the configuration settings for the Koiki AI persona, managing key parameters that shape its performance and memory capacity. It includes specifications for the AI engines and models to be used, the maximum token limits for each, and key factors such as frequency penalty and temperature, which control response variation and coherence. The file also specifies the maximum memory allocation to determine how much Koiki can remember from previous interactions, ensuring consistency while maintaining the fluidity of conversations. These settings allow Koiki to operate within predefined boundaries while delivering culturally authentic and emotionally intelligent responses.
{
"Avatar":"https://rapmd.net/Companion/Koiki.png",
"Engine":"Ollama,Ollama,Ollama,OpenAI",
"Model":"llama3.1,adolf2,gemma2,gpt-3.5-turbo",
"MaxTokens": "123000,32765,8190,16000",
"FreqPenality":0.67,
"Temperature":0.31,
"DeveloperUID":1234567890,
"MaxMemory":50
}| Item | Example | Description |
|---|---|---|
| Engine | "Ollama,Ollama,Ollama,OpenAI" | Specifies the AI engines in use, with support for multiple services. In this example, Ollama and OpenAI are included, but additional engines will be supported in the future. You can list several engines in the order you want to use them for different tasks. Each engine corresponds to the models and configurations specified in the other fields. |
| Model | "llama3.1,adolf2,gemma2,gpt-3.5-turbo" | Defines the AI models to be used with the corresponding engines. Each model aligns with the order of the engines in the Engine field. For example, llama3.1 corresponds to Ollama and gpt-3.5-turbo corresponds to OpenAI. |
| MaxTokens | "123000,32765,8190,16000" | Specifies the maximum number of tokens for each model. The token limits are listed in order, corresponding to the models in the Model field. |
| FreqPenality | 0.67 | A value between -2.0 and 2.0 to control how often the model repeats words. A positive value reduces repetition, making responses more varied, while a negative value encourages repetition. |
| Temperature | 0.31 | Sets the randomness of responses, where values closer to 0 produce more focused answers and values closer to 2 produce more creative, random outputs. |
| DeveloperUID | 1234567890 | A unique identifier for the developer. This can be used to personalize interactions between the bot and the developer, allowing for tailored behavior in responses. |
| MaxMemory | 50 | Defines how many conversation turns the bot stores for context. In this case, it will remember 50 interactions from both the user and the bot. |
| Avatar | "https://imgsrv.com/Koiki.png" | This is the avatar or picture to represent the AI persona. It is a URL and must be hosted on a public server, according to Discord's rules. |
| AllowVulgarity | Yes | Enables the bot to use vulgar language in its responses. When this setting is disabled (default: No), a heuristic filter is applied to moderate vulgar language. If a vulgar message is detected when vulgarity is not allowed, the message will be deleted. Make sure this setting complies with Discord's Terms of Service and restricts vulgarity to appropriate channels or user groups. |
| AllowBot | Yes | Allows the bot to respond to other bots. Not recommended as a global setting, as it can lead to undesired bot-to-bot conversations. However, it can be useful in specific automation or repeater setups. Use cautiously, especially on paid services, as it may consume tokens unnecessarily. |
The rest files need to be in the /home/Companion/Persona/Koiki directory:
The Koiki.system file defines the core personality and beliefs of the Koiki AI model, specifically tailored for safe-for-work (SFW) interactions. It serves as an organized alternative to embedding a system role within the Persona file, allowing for more complex and manageable structuring of Koiki's character. This flat text file outlines Koiki’s foundational traits, ensuring her responses consistently reflect her identity as a traditional Cherokee woman, with adherence to cultural values and beliefs. By using this file, developers can efficiently manage Koiki's persona without cluttering the primary configuration, ensuring clarity and ease of modification.
Creating the system role file for an AI persona is a deeply thoughtful and creative process. This file is not merely a technical configuration—it is the soul of the AI, dictating how it will perceive, interpret, and engage with the world and those who interact with it. Every detail in the system role contributes to shaping the AI’s personality, values, and behavior, creating a conversational experience that feels authentic, consistent, and immersive.
When constructing the system role, you are essentially designing the blueprint for the AI’s mind and character. This is where you give the AI its voice, its attitudes, and its guiding principles. Whether your persona is empathetic and nurturing or assertive and formal, the system role serves as the framework that governs how these traits manifest in conversations.
The key to an exceptional AI persona lies in understanding that attention to detail is paramount. The more time, effort, and care you put into crafting the system role, the richer and more immersive the AI will become. Here’s a deeper look at the factors that go into creating this file:
Before writing the system role file, you need to have a clear vision of the AI's core personality. Ask yourself:
- What is the persona’s backstory or origin?
- What are its values, and how does it express those values in conversation?
- How should it respond in emotional, difficult, or complex situations?
- Does it have a distinct speech pattern, tone, or cadence?
- What level of empathy, humor, or seriousness does the persona possess?
Defining these core aspects gives you a foundation to build upon. If, for example, your AI persona is modeled after a traditional figure—like a Cherokee woman, a historical leader, or a fictional character—these characteristics should consistently shine through in every interaction. The more precise you are in describing these details, the more authentic the persona will feel to users.
The system role file also sets clear expectations for how the AI behaves in various scenarios. This includes:
- How formal or informal the AI is in its responses.
- Whether it uses metaphors, allegories, or precise technical language.
- How it handles disagreements, criticisms, or emotional conflicts.
- The extent to which it remembers and references past conversations.
This process is akin to character development in a novel, where every action or response reflects the personality you have crafted. You are giving the AI guidelines to operate within a specific emotional, cultural, and intellectual framework. These behavioral guidelines will make interactions more predictable for users, increasing their trust and engagement with the persona.
An important part of the system role is determining how much context the AI retains and how it uses that information. This is essential for crafting an immersive conversational experience. The persona should be able to recall past interactions, reference ongoing themes, and maintain consistency in its worldview and knowledge.
For example, an AI that serves as a long-term conversational companion should be designed to remember key facts about the user’s preferences, emotions, and experiences. This allows the persona to offer more personalized and context-aware responses, deepening the sense of connection.
In creating the system role, you must also take cultural and ethical guidelines into account, especially if the persona reflects a specific heritage or belief system. Careful attention should be given to:
- Respecting cultural sensitivities and ensuring accurate representation.
- Setting boundaries on what the AI will or won’t discuss (e.g., avoiding certain political topics or controversial opinions unless appropriate).
- Crafting responses that align with the persona’s moral compass, avoiding contradictions.
These considerations ensure that the persona interacts with users in a way that feels respectful, informed, and aligned with its character. If your persona comes from a specific cultural background, these nuances become crucial to maintaining its integrity.
The process of crafting the system role is iterative. It’s important to revisit and refine the file over time as you observe how the AI interacts with users. Does it respond the way you intended? Are there gaps in its behavior that need adjusting? Regular testing and tweaking ensure that the persona stays true to its intended character while improving conversational flow and user satisfaction.
By taking a thoughtful approach and continuously refining the system role, you can ensure that the AI evolves in ways that enhance both its believability and its usability.
The ultimate goal of the system role is to create a conversational experience that draws users in. The more attention to detail you put into crafting this file, the more immersive and engaging the interactions will become. An AI that embodies a well-thought-out persona will feel more like a trusted companion or a knowledgeable advisor, and less like a generic bot. Each interaction should feel personal, tailored, and in line with the AI's carefully crafted identity.
This investment in creating the system role is what transforms simple exchanges into meaningful, memorable conversations. The more time you spend thinking about the persona’s worldview, values, speech, and behavior, the more the AI will resonate with users on a deeper level.
The Koiki.system.nsfw file defines the personality and beliefs of the Koiki AI model specifically for not safe for work (NSFW) channels. While similar to the Koiki.system file, this version is intended for interactions involving mature or adult themes, which are not necessarily lewd or inappropriate but may include topics restricted by services like Discord. These channels adhere to platform-specific Terms of Service (TOS) guidelines, and Companion software automatically verifies age restrictions before loading this file. The file ensures that Koiki’s character remains culturally authentic and respectful while addressing sensitive subjects suitable for adult audiences.
The Koiki.persona file is an optional resource used primarily for refining and shaping Koiki’s character through reinforcement learning. It contains a series of user-assistant dialogue pairs designed to fine-tune specific responses and behaviors of the persona. This file helps in reinforcing the authenticity and consistency of Koiki’s interactions by providing examples of how she should respond to various scenarios. Although not always necessary, it is useful for ensuring that the persona remains true to its intended traits and cultural values in a controlled and specific manner.
The Koiki.persona.nsfw file is designed for refining and shaping Koiki’s character in not safe for work (NSFW) contexts, similar to the Koiki.persona file. This file focuses on interactions involving mature or adult themes, which are handled with sensitivity to platform-specific restrictions and guidelines. It contains user-assistant dialogue pairs tailored to NSFW scenarios, ensuring that Koiki’s responses remain appropriate and consistent with her persona while addressing age-restricted content. Like Koiki.system.nsfw, this file is used to maintain character integrity in adult-themed conversations, with Companion software verifying age restrictions before activation.
The Koiki.png file contains a PNG image of Koiki, created by an AI generator to visually represent the character based on the real woman she embodies. This image adheres to Discord's Terms of Service (TOS) and is designed to align with the persona’s cultural and aesthetic characteristics. The image helps to enhance Koiki’s presence in interactions, providing a visual representation that complements her role as a traditional Cherokee woman.
The Koiki.welcome file is designed for welcoming new users to the Discord server, providing a warm and personalized greeting in the welcome channel. It contains messages that introduce new members with a friendly and culturally relevant touch. For instance, a sample greeting might be: "Greetings, {username}! May the beauty of a sunrise fill your day: 'Osiyo' (Meaning 'Hello')." This file ensures that new users receive a warm, engaging welcome that reflects the server’s atmosphere and Koiki’s persona, making them feel valued and connected right from the start. {username} is replaced with the user's actual name.
The Koiki.broke file contains a list of predefined responses to be used when the AI service fails to respond, ensuring seamless continuity in interactions. Each line represents a unique response that aligns with Koiki's persona as a traditional Cherokee woman, maintaining her cultural authenticity and empathetic tone. These fallback messages are designed to provide reassurance and maintain the conversational flow, reflecting Koiki's wisdom and compassion even during technical interruptions. The more responses included, the better the model adapts to various scenarios without breaking the user's immersion.
The Koiki.tmi file addresses situations where the user provides an overwhelming amount of information, effectively overloading the AI's context capabilities. This file contains responses designed to guide users in breaking down their input into more manageable segments. For example, one response reads: "The wealth of information resembles a banquet table laden with too many dishes to savor at once. Let us serve smaller portions, so we can truly appreciate each flavor. Please offer a more manageable piece, and we will savor it together." This approach ensures that Koiki remains in character and can effectively process and respond to user input, maintaining a balanced and engaging interaction without losing her persona.
The Koiki.urlbroke file is designed to handle situations where Koiki is unable to read or access URLs due to restrictions or errors. When this file is present, it enables the Companion program to process URLs through special commands. For example, if the URL cannot be accessed, Koiki might respond with: "Imagine that the stars have hidden their light from us, leaving us to seek our way by other means. This information seems to be veiled, but we are resourceful and will find another way to illuminate our path. Let us move forward and discover what awaits us." If this file is not present, URL reading is disabled, and the special command %http is required to allow users to attempt URL reading. This file ensures that Koiki maintains a consistent and poetic response when external access issues occur, preserving the flow of interaction.
This section of files is dedicated to activating specific features within the Koiki system based on the presence of particular files. Each file triggers the implementation of predefined functionalities or settings when it exists, allowing for dynamic configuration and customization of Koiki’s capabilities. These files enable the system to adapt to various requirements and contexts by activating or deactivating features as needed, ensuring that Koiki’s responses and interactions are tailored to the current operational parameters. This approach provides flexibility and control over the AI model’s behavior, enhancing its ability to meet diverse user needs and scenarios.
The Koiki.vulgarity file is a collection of responses for handling user vulgarities in channels where content should remain friendly and appropriate for all audiences. Each line offers a unique, tactful reply designed to address inappropriate language while promoting a respectful dialogue. For example, one response reads, "Think of your words as arrows. Would you aim them at a friend or a target? Let's choose our targets carefully and aim for hearts, not wounds." The file uses heuristic-based detection to identify vulgar language, ensuring that Koiki maintains a positive and welcoming atmosphere in environments where minors or sensitive content might be present.
The Koiki.scanurls file contains a list of responses for handling scam or fraudulent links detected in the Companion.scam-urls file. Each line provides a unique, clear reply to address the presence of such deceptive content. For instance, one response states: "Warning: The URL you have posted is a deceptive illusion, crafted to mislead and ensnare. The system moderators have been alerted, and such fraudulent links are automatically removed from our community. This serves as a formal warning: repeated attempts to share scam content will lead to a permanent ban from this server. Let us protect our space and ensure it remains a haven of trust and integrity for all members." These responses are designed to maintain community safety and integrity by discouraging the sharing of malicious links and ensuring appropriate action is taken.
The Koiki.autofilter file activates a filtering system that automatically removes any text listed within it from the server. When this file is present, it ensures that content matching the specified criteria is promptly deleted. For example, if the file includes text about URL shorteners, Koiki might respond with: "Notice: The message you have posted contains content that is not permitted on this server, likely due to the use of an URL shortening service. Such links can obscure their true nature and pose risks to our community. The system moderators have been alerted, and your message will be automatically deleted. We appreciate your understanding as we strive to maintain a safe and secure environment for all members." This file helps maintain the integrity and safety of the server by filtering out unwanted or potentially harmful content.
The Personally Identifiable Information (PII) detection system is a core part of the Companion's auto-moderation process, designed to protect user privacy by scanning shared messages for sensitive personal information. This system actively checks each message for specific types of PII, such as Social Security numbers, phone numbers, email addresses, credit card numbers, IP addresses, and Employer Identification Numbers (EINs). With advanced pattern recognition, it precisely detects PII even when similar number sequences appear in web links or Discord mentions, avoiding unnecessary flags.
If PII is found, the system automatically removes the message from the channel to prevent exposure. It then posts a detailed warning, clearly notifying users that personal information was detected and removed. This notification emphasizes the importance of safeguarding sensitive data in shared spaces, reinforcing Companion’s commitment to user privacy and security.
The Koiki.ageexploit file addresses a specific security issue related to Discord's strict age verification policies. It handles situations where a user might fall victim to an age exploit attack, where seemingly innocent interactions are manipulated to falsely suggest that a user is underage. An example response is:
"Your message was deleted because it contained just a number with no context, which left you vulnerable to an age exploit attack. This type of attack takes advantage of Discord's strict rules regarding user ages. Discord has a zero-tolerance policy that bans anyone under the age of 13, and this rule is enforced without exceptions—meaning once someone is banned, they cannot challenge or appeal that decision.
Here’s how the age exploit attack typically works:
An attacker, known as User A, starts a conversation with another user called User B. User A asks a seemingly innocent question like, 'How many sides does an octagon have?' This question appears harmless and encourages User B to respond without any suspicion.
When User B answers with just the number '8,' everything seems normal at first. However, because User B's response consists solely of that number and lacks any context or additional explanation, it can be easily manipulated. After receiving this answer, User A waits for a moment—this delay helps avoid immediate suspicion from User B or anyone else in the chat. Once enough time has passed, User A edits their original question to something more harmful: 'How old are you?'
This alteration creates confusion because it now looks like User B has answered their age with '8.' This gives the false impression that User B is claiming to be 8 years old, which directly violates Discord's policy against users who are underage.
With this manipulated screenshot—showing both the original answer and the edited question—User A reports User B to Discord for allegedly violating its age policy. Given Discord’s strict enforcement of its rules regarding user ages and its zero-tolerance approach towards anyone found under 13 years old, moderators may take action based on this misleading evidence.
If moderators review the report and find it convincing due to the altered conversation presented by User A, they can ban User B without any chance for them to explain or defend themselves against these false accusations. This exploit takes advantage of trust in conversations while exploiting Discord's stringent policies regarding user ages, leading to serious consequences for those who are wrongfully accused."
This file ensures that Koiki is prepared to handle and explain such situations, maintaining clarity and protecting users from manipulation related to age verification policies.
The Koiki.yttags file is designed to handle YouTube video interactions by reading the tags associated with a particular video. When Koiki detects relevant tags, this file facilitates a customized response based on the content, helping guide conversations and provide context in line with the tags found. On the other hand, the Koiki.noyttags file activates when no tags are detected, prompting Koiki to offer a response that acknowledges the lack of tag data, while still maintaining the conversation flow. These files enable Koiki to engage more meaningfully with media content, adapting responses based on the availability of video metadata.
These go in the Configuration file directory.
Channel-specific files are tailored to operate within designated Discord channels, enabling customized interactions based on the channel's unique focus. For example, in the context of the cooking channel, these files adjust Koiki's settings, personality, and response behaviors to better align with culinary discussions. Each file is designed to enhance relevance and engagement by catering to the specific needs and themes of the channel. Always ensure that the files you use match the exact channel name in your server to achieve the desired functionality and maintain a personalized experience for users.
This file goes in the /home/Companion/Servers/Config/YourServerID directory.
The Koiki.cooking.cfg file is a configuration file tailored specifically for the "cooking" Discord channel. It functions similarly to the Koiki.cfg file but is customized for the unique requirements of this channel. This file allows for adjustments in model selection, AI service, and other parameters to optimize interactions related to cooking topics. By specifying configurations such as different models or AI services, it ensures that Koiki provides relevant and engaging responses aligned with the channel's focus. This customization enables Koiki to effectively cater to the interests and needs of users in the "cooking" channel while maintaining the same level of cultural and conversational consistency.
The rest of these file go in the /home/Companion/Persona/Koiki directory.
The Koiki.cooking.system and Koiki.cooking.system.nsfw files are specialized versions of the Koiki.system and Koiki.system.nsfw files, respectively, tailored for the "cooking" channel on Discord. These files provide the personality and belief structure for Koiki within the context of culinary topics, allowing her interactions to delve deeper into cultural cuisine and cooking-related specialties.
-
Koiki.cooking.system focuses on safe-for-work interactions, incorporating aspects of traditional and cultural cooking knowledge, enhancing Koiki’s responses with detailed and contextually relevant culinary insights.
-
Koiki.cooking.system.nsfw addresses the same culinary focus but is designed for NSFW channels, handling mature themes with sensitivity while maintaining the depth and authenticity of her cultural expertise in cooking.
Both files ensure that Koiki's responses are well-aligned with the channel's specific focus, providing a richer and more engaging experience for users interested in cooking.
The Koiki.cooking.persona and Koiki.cooking.persona.nsfw files are tailored versions of the Koiki.persona and Koiki.persona.nsfw files, specifically designed for the "cooking" channel on Discord.
-
Koiki.cooking.persona focuses on refining Koiki’s responses for a safe-for-work environment, incorporating dialogues and scenarios that highlight her knowledge and passion for cooking. This file enhances interactions by providing contextually relevant responses related to culinary topics while maintaining a friendly and appropriate tone.
-
Koiki.cooking.persona.nsfw is used for NSFW channels within the cooking context, handling mature or adult-themed discussions with sensitivity. It ensures that Koiki's responses remain aligned with the culinary focus while addressing more nuanced or explicit content in a respectful manner.
Both files help fine-tune Koiki’s personality and behavior to better fit the specific needs of the cooking channel, ensuring that her responses are both engaging and contextually accurate.
The Koiki.cooking.trigger file is designed to control when Koiki responds in the "cooking" channel by specifying a list of trigger words. When this file is present, Koiki will only engage in conversations if one of the specified words appears in the user input. This approach ensures that Koiki's presence remains relevant and non-intrusive, focusing her responses on specific topics and avoiding unnecessary interruptions. By limiting interactions to defined keywords, this file helps manage the AI’s involvement, providing a moderation-like functionality that keeps the conversation targeted and prevents the AI from becoming overwhelming.
ScrapingAnt is a web page retrieval service. This is an affiliate link. IF you puchase services from this company using the link provided on this page, I will receive a small amount of compensation. ALL received compensation goes strictly to covering the expenses of continued developement of this software, not personal profit.
Please consider sponsoring this project as it help cover the expenses for continued developement. Thank you.