AI Discord Bot

A feature-rich Discord bot powered by OpenAI (ChatGPT), Google (Gemini), or Anthropic (Claude) models, designed to provide intelligent conversational capabilities with image analysis support right within your Discord server.

🚀 Quick Start

Prerequisites

• Discord Bot Token - Create a new application and bot
• OpenAI, Gemini, or Anthropic API key - Depending on which provider you use
• Doppler - For secure secret management at runtime
• Docker and Docker Compose

Docker Deployment

1. Set up Doppler:

   • Create a Doppler project and config (e.g. dev, prd)
   • Add the following secrets to your Doppler config:
     • Required: DISCORD_BOT_TOKEN, DISCORD_CLIENT_ID, and one of: OPENAI_API_KEY, GEMINI_API_KEY, or ANTHROPIC_API_KEY
     • Optional: AI_PROVIDER, OPENAI_MODEL_NAME, GEMINI_MODEL_NAME, CLAUDE_MODEL_NAME, LOG_LEVEL, MAX_HISTORY_LENGTH, MAX_HISTORY_TOKENS, REASONING_EFFORT, RESPONSES_VERBOSITY, ENABLE_WEB_SEARCH, ENABLE_GOOGLE_MAPS, ENABLE_CONTEXT_CACHE, USER_COOLDOWN_MS, CHANNEL_COOLDOWN_MS, MAX_PENDING_PER_CHANNEL, IMAGE_DOWNLOAD_TIMEOUT_MS, MAX_IMAGE_BYTES, and others listed in Configuration below
   • Create a service token for the config and copy it (you will pass it as DOPPLER_TOKEN)

2. 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

3. Set the Doppler token and deploy:

export DOPPLER_TOKEN=your_doppler_service_token_here
docker compose up -d

⚙️ Configuration

Doppler Setup

This bot uses Doppler to inject secrets as environment variables at runtime. The container runs doppler run -- ... so all keys in your Doppler config are available to the app.

Required:

1. Create a Doppler project and config (e.g. prd).
2. Add your secrets in the Doppler dashboard (or CLI). At minimum: DISCORD_BOT_TOKEN, DISCORD_CLIENT_ID, and one of OPENAI_API_KEY, GEMINI_API_KEY, or ANTHROPIC_API_KEY.
3. Generate a service token for that config and pass it when running the container as DOPPLER_TOKEN (e.g. in docker-compose.yml or your orchestration).

Environment Variables

Variable	Description	Required	Default	Example
DOPPLER_TOKEN	Doppler service token for the project/config	✅	-	-

All other variables below can be stored in Doppler (recommended) or set in environment in your docker-compose.yml. The app reads them after Doppler injects them at startup.

AI provider and models

Variable	Description	Default
AI_PROVIDER	Backend to use: openai, gemini, or claude.	openai
OPENAI_MODEL_NAME	OpenAI model when AI_PROVIDER=openai.	gpt-5-nano
GEMINI_MODEL_NAME	Gemini model when AI_PROVIDER=gemini. Falls back to OPENAI_MODEL_NAME if unset.	gemini-2.5-flash
CLAUDE_MODEL_NAME	Claude model when AI_PROVIDER=claude. Falls back to OPENAI_MODEL_NAME if unset.	claude-haiku-4-5-20251001

OpenAI (Responses API, text + image, reasoning, verbosity, optional web search):
gpt-5.2, gpt-5.1, gpt-5, gpt-5-mini, gpt-5-nano, gpt-5.2-pro, gpt-5-pro, gpt-4.1, gpt-4.1-mini, gpt-4.1-nano, o3, o4-mini, o3-pro, o3-mini

Gemini (text + image, search grounding, thinking):
gemini-3-pro-preview, gemini-3-flash-preview, gemini-2.5-flash, gemini-2.5-flash-lite, gemini-2.5-pro, gemini-2.0-flash, gemini-2.0-flash-lite

Claude (vision, extended thinking on 4.5):
claude-sonnet-4-5-20250929, claude-sonnet-4-5, claude-haiku-4-5-20251001, claude-haiku-4-5, claude-opus-4-5-20251101, claude-opus-4-5, claude-3-5-sonnet-20241022, claude-3-5-haiku-20241022, claude-3-opus-20240229, claude-3-sonnet-20240229, claude-3-haiku-20240307, claude-sonnet-4-20250514, claude-3-5-sonnet-20240620

Set OPENAI_API_KEY for OpenAI; GEMINI_API_KEY for Gemini (Google AI Studio); ANTHROPIC_API_KEY for Claude (Anthropic Console).

Reasoning, verbosity, and search

Variable	Description	Default
REASONING_EFFORT	OpenAI only. Reasoning effort: none, low, medium, high, xhigh. GPT-5.2 also supports none; GPT-5.2-pro supports medium, high, xhigh.	none
RESPONSES_VERBOSITY	OpenAI GPT-5 only. Response verbosity: low, medium, high.	low
ENABLE_WEB_SEARCH	OpenAI and Gemini. Set to true or 1 to enable web search (OpenAI built-in tool; Gemini Google Search grounding).	off
ENABLE_GOOGLE_MAPS	Gemini only. Set to true or 1 to enable grounding with Google Maps (places, area summaries, location-aware answers).	off

Optional tuning variables

Variable	Description	Default
MAX_OUTPUT_TOKENS	Max tokens per reply (all providers). Clamped 256–65536. Lower values reduce cost.	1024
MAX_HISTORY_TOKENS	Rough token-estimated cap for stored per-channel conversation history (in addition to MAX_HISTORY_LENGTH). 0 disables token trimming.	0
USER_COOLDOWN_MS	Minimum time between requests per user (basic anti-spam/cost control).	4000
CHANNEL_COOLDOWN_MS	Minimum time between requests per channel (reduces pile-ups).	1500
MAX_PENDING_PER_CHANNEL	Max queued requests per channel before the bot responds “busy”.	3
IMAGE_DOWNLOAD_TIMEOUT_MS	Timeout for downloading image attachments.	8000
MAX_IMAGE_BYTES	Max bytes downloaded per image attachment.	6000000
OPENAI_TIMEOUT_MS	OpenAI only. Request timeout in milliseconds (5000–300000).	60000
OPENAI_MAX_RETRIES	OpenAI only. Max retries for transient failures (0–5).	2

Context caching

Reduces cost and latency by caching static prompt content. Single switch for all providers. Can be stored in Doppler or set in your environment.

Variable	Description	Default
ENABLE_CONTEXT_CACHE	Set to true or 1 to enable context/prompt caching for all providers.	false
GEMINI_CACHE_TTL_SECONDS	TTL for Gemini context cache (60–2073600).	3600

Claude extended thinking

Variable	Description	Default
CLAUDE_THINKING_BUDGET_TOKENS	Claude only. Token budget for extended thinking on supported 4.5 models (0 = disabled, max 32000).	0

Gemini safety

Variable	Description	Default
GEMINI_SAFETY_SETTINGS	Gemini only. JSON array of {"category":"...","threshold":"..."} to tune safety filters. Example: [{"category":"HARM_CATEGORY_HARASSMENT","threshold":"BLOCK_MEDIUM_AND_ABOVE"}]. Unset = API defaults.	-

🖼️ Image Analysis

The bot supports comprehensive image analysis when using vision-capable models:

• Image Descriptions: Get detailed descriptions of image content
• Visual Q&A: Ask questions about images and receive contextual answers
• Multi-Modal Input: Combine text and images in the same message
• Automatic Detection: Automatically processes image attachments

Usage Examples:

• Send an image with text: "What's in this image?"
• Ask follow-up questions about previously shared images
• Get analysis of charts, diagrams, or screenshots

💬 Conversation Features

Multi-Channel Support

• Shared conversation history per channel, allowing multiple users to participate
• Context preservation across message exchanges from all users
• Automatic history management and cleanup
• Basic backpressure (per-channel queue limit) and cooldowns to reduce spam/cost

Interaction Methods

• Mentions: @AI What's the weather like?
• Replies: Reply to any bot message to continue the conversation

Safety Defaults

• The bot sends messages with mentions disabled (allowedMentions: { parse: [] }) to avoid accidental @everyone / role pings.
• Logs avoid printing full message contents and full model replies by default (metadata only).

🔧 Commands

/reset (admin only)

Reset conversation history for a specific channel or all channels. Only users with the Administrator permission can see and use this command.

• No channel specified: Resets conversation history for all channels
• Channel specified: Resets conversation history for the selected channel only