Discord Ollama Bot

A Discord chatbot that uses Ollama to run AI models locally. Built with .NET 8 and Discord.NET.

Features

• @Mention responses - Mention the bot in any channel to chat
• Direct message support - DM the bot without needing to @mention
• Per-user conversation memory - Each user has their own conversation context
• Customizable personality - Configure the system prompt to change bot behavior
• Fully local - No API costs, runs entirely on your machine

Prerequisites

• .NET 8 SDK
• Ollama installed and running
• A Discord bot token (Discord Developer Portal)

Quick Start

1. Clone the repository

   git clone https://github.com/yourusername/DiscordOllamaBot.git
   cd DiscordOllamaBot

2. Pull the Ollama model

   ollama pull qwen3:4b

3. Configure the bot

   Edit appsettings.json with your Discord bot token:

   {
     "Discord": {
       "Token": "YOUR_BOT_TOKEN"
     }
   }

4. Enable Message Content Intent

   In the Discord Developer Portal:

   • Select your application → Bot → Privileged Gateway Intents
   • Enable Message Content Intent

5. Run the bot

   dotnet run

Configuration

Setting	Default	Description
Discord:Token	-	Your Discord bot token (required)
Ollama:Endpoint	http://localhost:11434	Ollama API endpoint
Ollama:Model	qwen3:4b	Ollama model to use
Ollama:SystemPrompt	You are a helpful Discord assistant...	Bot personality/instructions
Conversation:MaxHistory	20	Max messages to remember per user

Project Structure

DiscordOllamaBot/
├── Program.cs                 # Entry point & DI setup
├── appsettings.json           # Configuration
├── Models/
│   ├── BotConfiguration.cs    # Config POCOs
│   └── ConversationContext.cs # User conversation state
├── Services/
│   ├── OllamaService.cs       # Ollama API integration
│   └── ConversationService.cs # Per-user history management
└── Handlers/
    └── MessageHandler.cs      # Discord message processing

Usage

In a server channel:

@BotName What is the capital of France?

In direct messages:

What is the capital of France?

The bot remembers your conversation history, so you can have multi-turn conversations.

Using Different Models

You can use any model available in Ollama. Update appsettings.json:

{
  "Ollama": {
    "Model": "llama3:8b"
  }
}

Popular options:

• qwen3:4b - Fast, good for general chat
• llama3:8b - Balanced performance
• mistral:7b - Great reasoning capabilities
• codellama:7b - Optimized for code

License

MIT