DenBot Discord Bot

A Discord bot that uses AI to respond to messages with multi-LLM provider support (Claude or OpenAI-compatible), analyze images, and provide tools like Wolfram Alpha, GPU performance lookups, and web search. Imitates an exaggerated version of DenTech's personality.

Features

• Multi-LLM Provider Support: Choose between Anthropic Claude or OpenAI-compatible local LLMs (LM Studio, Ollama, vLLM, etc.)
• AI-Powered Responses: Context-aware conversation using Discord reply chains for history
• Image Processing: Analyze and describe images via Claude API (when using Anthropic provider)
• Tool Integration:
  • Wolfram Alpha for computational queries
  • 3DMark GPU performance lookup with fuzzy matching
  • Web search capabilities
  • Website content summarization
  • YouTube video transcript context
• Auto-Reply Features:
  • Forum channel auto-replies for new posts
  • Regex-triggered automatic responses
• Permission System:
  • Role-based access control
  • Channel-specific permissions
  • User override capabilities
  • Forum channel whitelist
  • Server/guild-wide access control
• GitHub Integration: Automatic system prompt synchronization from GitHub repositories

Setup

1. Copy docker-compose.yml to a folder
2. Copy .env.example to .env in the same folder as the yml and configure:
   • Set LLM_PROVIDER to either anthropic (Claude) or openai (for local LLMs)
   • Fill in the appropriate API keys and settings for your chosen provider
   • Configure Discord permissions and channels
3. Run the bot: docker compose up -d

Configuration

Environment variables in v3/.env:

Required Variables

• BOT_API_KEY: Discord bot token
• LLM_PROVIDER: LLM provider to use (anthropic or openai)
• ALLOWED_CHANNELS: JSON array of channel IDs where bot responds to @mentions (e.g., ["123456789"])
• WOLFRAM_APPID: Wolfram Alpha API key for computational queries

When LLM_PROVIDER=anthropic:

• ANTHROPIC_API_KEY: Anthropic API key
• MODEL_NAME: Claude model to use (e.g., claude-haiku-4-5)
• SUBAGENT_MODEL_NAME: Model for subagent tasks (e.g., claude-haiku-4-5)

When LLM_PROVIDER=openai:

• OPENAI_BASE_URL: Base URL for OpenAI-compatible API (e.g., http://localhost:1234/v1)
• OPENAI_MODEL_NAME: Model identifier for your local LLM
• OPENAI_API_KEY: API key (often not-needed for local servers)

Optional Variables

Permissions:

• ALLOWED_ROLES: JSON array of role IDs that can use the bot anywhere
• OVERRIDE_USERS: JSON array of user IDs that bypass all restrictions
• ALLOWED_FORUM_CHANNELS: JSON array of forum channel IDs for auto-replies
• AUTHORIZED_SERVERS: JSON array of server/guild IDs where all members can use the bot

Feature Flags:

• FORUM_REPLIES_ENABLED: Enable auto-replies in forum channels (true/false, default: false)
• REGEX_REPLIES_ENABLED: Enable regex-triggered auto-replies (true/false, default: false)

GitHub Integration:

• GITHUB_TOKEN: GitHub personal access token for prompt syncing
• GITHUB_REPO: Repository in format owner/repo
• GITHUB_BRANCH: Branch to sync from (default: main)
• PROMPT_POLL_INTERVAL: Update check interval in seconds (default: 300)

Response Configuration:

• MAX_TOKENS: Maximum tokens for responses (default: 1024)
• WEB_SEARCH_MAX_TOKENS: Max tokens for web search results (Anthropic only, default: 1024)

Image Processing:

• IMAGE_MAX_DIMENSIONS: Maximum image width/height in pixels (default: 800)
• IMAGE_MAX_FILE_SIZE_MB: Maximum image file size in MB (default: 20)

Other:

• WOLFRAM_MAX_CHARS: Maximum characters from Wolfram Alpha (default: 1024)
• YOUTUBE_TRANSCRIPT_MAX_CHARS: Total transcript characters to include for YouTube links, split evenly between start and end (default: 4000)
• LOGGING_LEVEL: Logging verbosity (DEBUG, INFO, WARNING, ERROR, CRITICAL, default: INFO)

Usage

Interaction Methods

Context Menu:

• Right-click any message → Apps → Ask DenBot: Opens a modal to provide additional context, then the bot responds

@Mentions:

• Mention the bot in allowed channels to start a conversation
• The bot uses Discord reply chains to track conversation history

Forum Auto-Replies:

• When FORUM_REPLIES_ENABLED=true, the bot automatically replies to new posts in ALLOWED_FORUM_CHANNELS

Regex Auto-Replies:

• When REGEX_REPLIES_ENABLED=true, the bot responds to messages matching configured patterns in v3/regexreplies.json

Image Analysis:

• Attach images to your messages (when using LLM_PROVIDER=anthropic)
• The bot will analyze and describe images using Claude's vision capabilities

Requirements

Python Version:

• Python 3.12 (recommended and tested)

Dependencies:

• discord.py ~=2.4.0 - Discord API wrapper
• anthropic ~=0.40.0 - Anthropic Claude API client
• openai >=1.0.0 - OpenAI-compatible API client
• python-dotenv ~=1.0.0 - Environment variable management
• requests ~=2.32.0 - HTTP library
• thefuzz[speedup] ~=0.22.0 - Fuzzy string matching (for GPU lookup)
• aiohttp ~=3.11.0 - Async HTTP client
• Pillow ~=11.0.0 - Image processing

Install all dependencies with: pip install -r v3/requirements.txt