🤖 AI Discord Bot

A highly advanced, multi-provider AI Discord assistant powered by OpenAI, Google Gemini, and Anthropic Claude.

---

✨ Key Features

• 🌐 Multi-Model Flexibility: Seamlessly switch between OpenAI, Google (Gemini), and Anthropic (Claude). Always use the best model for the job.
• 👁️ Vision & Multi-Modal: Send images as attachments. The bot can describe them, answer visual Q&A, and analyze charts effortlessly.
• 💬 Shared Conversation History: Maintains contextual memory per Discord channel, allowing multiple users to collaboratively converse in the same thread.
• 🧵 Smart Context Tracing: Understands the Discord reply chain. Reply to old messages and the bot dynamically traces the correct context.
• 🔎 Real-Time Web Search & Maps: Ground your prompts in reality with live web search (OpenAI & Gemini) and Google Maps API integration (Gemini).
• ⚡ Prompt Caching: Drastically reduces costs and API latency for long conversations via built-in context caching.
• 🛡️ Safety & Anti-Spam: Features robust per-user and per-channel cooldowns, queue limits (backpressure), and safe mention defaults (never accidentally pings @everyone!).
• 📊 Production-Ready Observability: Deep integration with Sentry (traces, profiling, metrics) and structured Pino logging.
• 🔐 Secure Secrets Management: Powered by Doppler for effortless, secure environment variable injection at runtime.

---

🚀 Quick Start

1. Prerequisites

• Discord Developer Portal: Create a new app, add a bot, and secure the Token and Client ID.
• API Keys: Get an API key from OpenAI, Gemini, or Anthropic.
• Doppler: Create a free account for secure runtime secrets.
• Docker & Docker Compose installed on your host machine.

2. Configure Doppler

Create a Doppler project and add the following required secrets to your config (e.g., prd or dev):

• DISCORD_BOT_TOKEN
• DISCORD_CLIENT_ID
• OPENAI_API_KEY (or GEMINI_API_KEY / ANTHROPIC_API_KEY depending on your provider)

Generate a Service Token for your Doppler config. You will use this token to securely boot the bot.

3. Deploy with Docker Compose

Create a docker-compose.yml file:

services:
  ai:
    image: ghcr.io/doubleangels/ai:latest
    container_name: ai-discord-bot
    restart: unless-stopped
    cap_drop:
      - ALL
    cap_add:
      - CHOWN
      - SETGID
      - SETUID
    security_opt:
      - no-new-privileges:true
    read_only: true
    environment:
      - DOPPLER_TOKEN=${DOPPLER_TOKEN}
    tmpfs:
      - /tmp

Start the bot:

export DOPPLER_TOKEN="dp.st.config.your_token_here"
docker compose up -d

---

⚙️ Configuration Guide

The bot is heavily customizable. Add these optional keys to your Doppler environment to tweak its behavior.

🧠 AI Provider & Models

Variable	Description	Default
AI_PROVIDER	Choose your backend: openai, gemini, or claude.	openai
OPENAI_MODEL_NAME	Model to use when provider is OpenAI.	gpt-5.4-nano
GEMINI_MODEL_NAME	Model to use when provider is Gemini.	gemini-3-flash-preview
CLAUDE_MODEL_NAME	Model to use when provider is Claude.	claude-sonnet-4-6
ALLOWED_GUILD_IDS	Comma-separated list of allowed Discord Server IDs. Limits where the bot works.	All Servers

🛠️ Advanced AI Settings

Variable	Description	Default
ENABLE_WEB_SEARCH	true or false. Enables live internet search for OpenAI/Gemini.	false
ENABLE_GOOGLE_MAPS	true or false. Enables Google Maps grounding for Gemini.	false
ENABLE_CONTEXT_CACHE	true or false. Caches long prompt contexts to drastically save costs.	false
REASONING_EFFORT	low, medium, high. Configures OpenAI reasoning models (e.g. o1/o3/o4).	none
RESPONSES_VERBOSITY	low, medium, high. Dictates verbosity for supported models.	low
CLAUDE_THINKING_BUDGET_TOKENS	Token budget for extended reasoning on Claude 4.5+ models (0 = disabled).	0

🛑 Conversation & Cost Limits

Variable	Description	Default
MAX_OUTPUT_TOKENS	Hard cap on response tokens. Clamped between 256–65536.	1024
MAX_HISTORY_TOKENS	Token limit for channel history. Older messages drop off when this is reached.	0 (Off)
USER_COOLDOWN_MS	Delay between requests per user (anti-spam feature).	4000
CHANNEL_COOLDOWN_MS	Delay between requests per channel.	1500
MAX_PENDING_PER_CHANNEL	Max requests queued before bot replies with "I'm busy".	3

📈 Observability (Sentry)

If you wish to use Sentry for error tracking, add your SENTRY_DSN to Doppler.

Variable	Description	Default
SENTRY_DSN	Your Sentry DSN endpoint.	None
SENTRY_TRACES_SAMPLE_RATE	0.0 to 1.0 (Performance tracing capture rate).	1.0
SENTRY_PROFILE_SESSION_SAMPLE_RATE	0.0 to 1.0 (Code profiling capture rate).	1.0

---

🎮 Usage & Commands

Interacting with the Bot

• Mention the bot: Start a prompt by pinging it: @AI What is the capital of France?
• Reply to the bot: Right-click a previous bot message and use Discord's native "Reply" feature to continue a thread seamlessly.
• Note: The bot purposefully ignores @here and @everyone to prevent accidental spam.

Multi-Modal Image Analysis

Drag and drop an image into Discord, add a caption like @AI describe this chart, and hit send! The bot will automatically analyze and respond contextually.

Admin Commands

• /reset: Clears the AI's short-term memory (conversation history). Can be scoped to the current channel or the entire server. (Note: Only users with the Administrator permission can see and use this command).

_{Built with ❤️ using Node.js and Discord.js}