AI Selfbot

Discord selfbot that uses AI to respond to pings and DMs. Learns your communication style from chat history to maintain consistent personality per relationship.

Warning: Discord selfbots violate Discord's Terms of Service. Use at your own risk. This project is for educational purposes only.

Features

• AI-powered responses via OpenRouter or local Ollama
• Per-user profiles - learns how you talk to each person from DM history
• Memory system - remembers facts about people and topics discussed
• Server awareness - tracks users and context in servers
• Dead chat revival - optionally starts conversations in quiet channels
• GIF responses - can respond with configured GIFs when appropriate
• Configurable delays - humanizes response timing
• Blacklists - ignore specific users or servers

Setup

1. Clone and install dependencies:

git clone itzdaimy/AI-Selfbot
cd ai-selfbot
npm install

2. Copy .env.example to .env and fill in:

discord_token=discord account token here
openrouter_key=openrouter api key here

3. Edit config.json to your liking

4. Run:

npm run start

Configuration

config.json

Field	Description
provider	"openrouter" or "ollama"
model	Model ID for OpenRouter
ollama.baseUrl	Ollama API URL (default: http://localhost:11434)
ollama.model	Model name for Ollama
contextMessages	How many recent messages to include as context
profileHistoryLimit	Messages to analyze when learning a new profile
learningInterval	Interactions before updating a profile
memoryExtraction	Enable/disable memory extraction from conversations
defaultPrompt	System prompt for AI responses
serverPrompts	Per-server system prompts (key = server ID)
blacklist.users	Array of user IDs to ignore
blacklist.servers	Array of server IDs to ignore
delay.min / delay.max	Response delay range in ms

Server Chat Settings

"serverChat": {
  "readingPeriod": {
    "min": 3000,
    "max": 15000,
    "extendOnActivity": true,
    "maxExtensions": 3
  },
  "engagement": {
    "idleTimeout": 60000,
    "maxReplies": 10
  },
  "deadChat": {
    "enabled": true,
    "minBotMessages": 15,
    "idleMinutes": 30,
    "checkInterval": 300000,
    "chance": 0.3
  }
}

GIF Responses

Create gifs.json in the root directory:

{
  "enabled": true,
  "chance": 0.3,
  "gifs": [
    {
      "url": "https://tenor.com/...",
      "context": "when someone says something funny",
      "triggers": ["lmao", "lol", "haha"]
    }
  ]
}

How It Works

1. DMs: When someone DMs you, the bot checks if it has a profile for them. If not, it analyzes your chat history to learn your communication style with that person. Then it responds using AI with that context.

2. Server pings: When mentioned in a server, the bot enters a "reading period" to collect context from the conversation before responding. It can continue engaging naturally for multiple replies.

3. Profiles: Stored in data/profiles/. Each profile contains relationship summary, communication style, example exchanges, and common phrases.

4. Memory: Stored in data/memory/. Extracts and remembers facts about users and topics for future context.

File Structure

src/
├── index.js        # entry point
├── client.js       # discord connection and event routing
├── handler.js      # builds prompts and sends responses
├── profiles.js     # per-user profile learning and storage
├── memory.js       # fact and topic memory system
├── context.js      # time and conversation context
├── conversations.js # server conversation state management
├── servers.js      # server and user tracking
├── gifs.js         # gif response handling
└── ai.js           # openrouter/ollama api calls

data/
├── profiles/       # per-user profile JSON files
├── memory/         # memory storage
│   ├── users/      # per-user memories
│   └── self.json   # facts about yourself
└── servers/        # server tracking data

License

Check LICENSE file