AI Character System

A flexible system for creating and managing AI characters with platform-specific response styles and personality traits.

Features

• Character creation and management
• Platform-specific response formatting (Twitter, Discord)
• Tool integration for real-time data
• Multiple interaction modes
• Personality and style injection
• Memory management
• Event tracking

Installation

npm install
# or
bun install

Create a .env file with:

DATABASE_URL=your_database_url
OPENAI_API_KEY=your_openai_api_key

Quick Start

import { ai } from "../core/ai";

const assistant = new ai({
  databaseUrl: process.env.DATABASE_URL!,
  llmConfig: {
    apiKey: process.env.OPENAI_API_KEY!,
    llm: { model: "gpt-4-turbo-preview", temperature: 0.7 },
    analyzer: { model: "gpt-3.5-turbo", temperature: 0.3 },
  },
});

Example Usage

Character Management

// Find or create character
const characters = await assistant.db
  .select()
  .from(schema.characters)
  .where(eq(schema.characters.name, "Ducky"));

let character;
if (characters.length > 0) {
  character = characters[0];
} else {
  const duckyConfig = await createDuckyCharacter();
  character = await assistant.createCharacter(duckyConfig);
}

Platform-Specific Responses

Twitter Interactions

// Basic Tweet
const twitterResponse = await assistant.interact(
  {
    system: "Create a thread analyzing BTC market conditions",
    user: "Give me a spicy take on BTC price action",
  },
  {
    characterId: character.id,
    mode: "enhanced",
    responseType: "tweet_thread",
    tools: ["btc-price"],
  }
);

// Reply to Tweet
const replyResponse = await assistant.interact(
  {
    system: "Reply with signature sass",
    user: "Responding to: 'Just bought the dip! 🚀'",
  },
  {
    characterId: character.id,
    mode: "enhanced",
    responseType: "reply",
    platform: "twitter",
  }
);

Discord Interactions

// Technical Analysis
const technicalAnalysis = await assistant.interact(
  {
    system: "Provide detailed technical analysis",
    user: "Deep dive on BTC market structure",
  },
  {
    characterId: character.id,
    mode: "enhanced",
    responseType: "technical_analysis",
    platform: "discord",
    tools: ["btc-price"],
  }
);

// Alpha Calls
const alphaCall = await assistant.interact(
  {
    system: "Share market alpha",
    user: "What's your latest BTC alpha?",
  },
  {
    characterId: character.id,
    mode: "enhanced",
    responseType: "alpha_calls",
    platform: "discord",
  }
);

// Meme Response
const memeResponse = await assistant.interact(
  {
    system: "Create meme-worthy market commentary",
    user: "Thoughts on leverage traders getting rekt?",
  },
  {
    characterId: character.id,
    responseType: "meme_response",
    platform: "discord",
  }
);

Different Interaction Modes

// Raw Mode
const rawResponse = await assistant.interact(
  {
    system: "You are a market analyst",
    user: "Current BTC state?",
  },
  {
    mode: "raw",
    tools: ["btc-price"],
  }
);

// Enhanced Mode
const enhancedResponse = await assistant.interact(
  {
    system: "Analyze market conditions",
    user: "Technical analysis please",
  },
  {
    characterId: character.id,
    mode: "enhanced",
    responseType: "technical_analysis",
  }
);

// Mixed Mode with Custom Injection
const mixedResponse = await assistant.interact(
  {
    system: "Analyze market sentiment",
    user: "Unfiltered market thoughts?",
  },
  {
    characterId: character.id,
    mode: "mixed",
    platform: "discord",
    responseType: "general_chat",
    injections: {
      injectPersonality: true,
      injectStyle: false,
      customInjections: [
        {
          name: "market_sentiment",
          content: "You're feeling bearish but hiding it",
          position: "before",
        },
      ],
    },
    tools: ["btc-price"],
  }
);

Response Types

Twitter

• tweet_thread: Multi-tweet analysis (max 280 chars each)
• reply: Single tweet responses

Discord

• general_chat: Casual conversation (max 1000 chars)
• technical_analysis: Detailed market analysis
• alpha_calls: Trading insights and calls
• meme_response: Meme-worthy commentary

Character Configuration

Characters can be configured with:

• Base traits and personality
• Platform-specific response styles
• Formatting rules
• Interaction guidelines

See createDuckyCharacter() for a complete example of character configuration.

Tools Integration

Tools can be used to fetch real-time data:

const response = await assistant.interact(input, {
  tools: ["btc-price"],
  toolContext: {
    /* optional context */
  },
});

Development

# Run tests
bun test

# Run example
bun run example/index.ts

License

MIT