Claude Discord Bot (Elixir)

A Discord bot powered by LLM AI, ported from Python to Elixir using the Nostrum library.

Features

• AI Chat: Mention the bot in any message to get an AI-powered response
• Message Context: Bot reads channel history to maintain conversation context
• Image Support: Attach images to your messages for AI analysis
• Slash Commands: /help and /info commands for bot information
• Rate Limiting: Per-user rate limiting to prevent spam
• Typing Indicators: Shows typing while processing requests
• Extensible LLM Providers: Easy to add new LLM backends

Architecture

lib/
├── claude.ex                    # Main module with public API
├── claude/
│   ├── application.ex           # OTP Application & Supervisor
│   ├── config.ex                # Configuration (files & env vars)
│   ├── llm.ex                   # LLM facade module
│   ├── llm/
│   │   ├── provider.ex          # Provider behaviour
│   │   └── providers/
│   │       └── openai.ex        # OpenAI-compatible implementation
│   ├── message_consumer.ex      # Nostrum event handler
│   ├── message_handler.ex       # Message processing logic
│   ├── rate_limiter.ex          # Per-user rate limiting (non-blocking)
│   ├── user_cache.ex            # ETS-backed user cache
│   ├── member_cache.ex          # ETS-backed member cache
│   ├── utils.ex                 # Helper functions
│   └── commands/
│       ├── command.ex           # Command behaviour
│       ├── registry.ex          # Command registration & routing
│       ├── info.ex              # /info command
│       └── help.ex              # /help command

Configuration

Edit config/config.exs to configure the bot:

config :claude,
  discord: %{
    token: "YOUR_DISCORD_BOT_TOKEN"
  },
  llm: %{
    provider: Claude.LLM.Providers.OpenAI,
    model: "gpt-4",
    llm_api_key: "YOUR_LLM_API_KEY",
    llm_base_url: "https://api.openai.com/v1",
    max_tokens: 1024,
    max_context_messages: 50,
    rate_limit_ms: 2_000
  }

Alternatively, one can create config/dev.exs or config/production.exs to set configs specifically for those runtime environments.

Environment Variables

You can also configure the bot using environment variables (these take priority over config files):

export CLAUDE_DISCORD_TOKEN="your-discord-token"
export CLAUDE_LLM_API_KEY="your-llm-api-key"
export CLAUDE_LLM_BASE_URL="https://api.openai.com/v1"
export CLAUDE_MODEL="gpt-4"
export CLAUDE_MAX_TOKENS="1024"
export CLAUDE_MAX_CONTEXT_MESSAGES="50"
export CLAUDE_RATE_LIMIT_MS="2000"

This is useful for deployments where you don't want to commit secrets to config files.

Configuration Options

Option	Description	Default
provider	LLM provider module	Claude.LLM.Providers.OpenAI
model	Model name for chat completions	"doubao-seed-1-6-thinking-250615"
llm_api_key	API key for the LLM provider	Required
llm_base_url	Base URL for OpenAI-compatible API	"https://aihubmix.com/v1"
max_tokens	Maximum tokens in LLM response	1024
max_context_messages	Messages to include in context	50
rate_limit_ms	Milliseconds between messages per user	2000

Installation & Running

1. Install dependencies:

   mix deps.get

2. Configure the bot: Edit config/config.exs with your Discord token and LLM API key.

3. Run the bot:

   mix run --no-halt

   Or in IEx for development:

   iex -S mix

Usage

Chat with the Bot

Mention the bot in any message:

@Claude what's the weather like today?

The bot will:

1. Read recent channel history for context
2. Show a typing indicator
3. Call the LLM API
4. Reply with the AI response

Clear Conversation History

Send this message to reset context:

[DO NOT COUNT PAST THIS MESSAGE]

Slash Commands

• /help - Show usage information
• /info - Display bot configuration and status

Adding New LLM Providers

Implement the Claude.LLM.Provider behaviour:

defmodule Claude.LLM.Providers.MyProvider do
  @behaviour Claude.LLM.Provider

  @impl true
  def name, do: "My Provider"

  @impl true
  def validate_config do
    # Return :ok or {:error, reason}
    :ok
  end

  @impl true
  def chat_completion(messages, options) do
    # Return {:ok, response_text} or {:error, reason}
    {:ok, "Hello!"}
  end

  @impl true
  def supports_images?, do: true
end

Then configure it:

config :claude, :llm, %{
  provider: Claude.LLM.Providers.MyProvider,
  # ... other options
}

Supervision Tree

Claude.Supervisor
├── Claude.RateLimiter (GenServer)
├── Claude.UserCache (ETS)
├── Claude.MemberCache (ETS)
└── Nostrum.Bot
    └── Claude.MessageConsumer

Development

Run Tests

mix test

Generate Documentation

mix docs

Todo

• /image command with fal.ai integration
• File-based logging with rotation
• Guild-specific configuration
• Conversation memory persistence

Credits

• Original Python bot by fizz
• Elixir port uses Nostrum for Discord API
• HTTP requests via Req

License

AGPLv3