Discord Knowledge Bot

Demo Video: https://www.youtube.com/watch?v=2LhrUmU_F2E

A Discord bot that indexes server content and provides AI-powered chat functionality using local vectorDB, local embeddings, and OpenAI for final chat completion (other model support coming soon).

Key Features

• Local Vector Database: Uses ChromaDB for local vector storage - no external database required
• Local Embeddings: Uses sentence-transformers/all-MiniLM-L6-v2 for local embedding generation - no external embedding API needed
• OpenAI Integration: Only external dependency is OpenAI API for final chat completion (other model support coming soon)
• Self-Contained: Runs entirely locally to the host machine except for OpenAI chat completion

Features

• Server Indexing: Index all text channels in a Discord server
• Channel Indexing: Index specific channels with pagination support
• AI Chat: Natural language conversation with context from server content
• Context-Aware Search: Channel-specific or server-wide search capabilities
• Reindexing: Clear and rebuild indexes as needed
• Rate Limiting: Respects Discord API limits
• Text-Only: Focuses exclusively on text content

Requirements

• Python 3.8 or higher
• Discord Bot Token
• OpenAI API Key (required today for response generation)

Environment Variables

The bot requires the following environment variables to be set in a .env file:

• DISCORD_TOKEN: Your Discord bot token from the Discord Developer Portal
• DISCORD_APP_ID: Your Discord application ID from the Discord Developer Portal
• OPENAI_API_KEY: Your OpenAI API key from the OpenAI platform

These are loaded automatically when the bot starts.

Installation

1. Clone the repository:

   git clone <repository-url>
   cd discord_knowledge_bot

2. Install dependencies:

   pip install -r requirements.txt

3. Set up environment variables: Copy the example environment file and configure it:

   cp env.example .env

   Then edit .env and add your actual values:

   DISCORD_TOKEN=your_discord_bot_token_here
   DISCORD_APP_ID=your_discord_application_id_here
   OPENAI_API_KEY=your_openai_api_key_here
   

4. Configure the bot: Edit config.yaml to customize bot settings.

Usage

Starting the Bot

python main.py

Bot Commands

All commands are now slash commands (/) that appear in Discord's command interface.

Indexing Commands (Admin Only)

• /index-server - Index all text channels in the server
• /index-channel [channel] - Index a specific channel (or current channel)
• /reindex-server - Clear and reindex all text channels
• /reindex-channel [channel] - Clear and reindex a specific channel

Management Commands

• /help - Show help information and available commands
• /status - Show bot and indexing status
• /stats - Show detailed statistics
• /clear - Clear all indexed data (Admin only)

Chat Commands

• /ask <question> - Ask a question about this channel's content
• /ask-server <question> - Ask a question about the entire server's content

Direct Mentions

Mention the bot with @bot_name to chat directly! The bot will search through indexed content to provide relevant answers.

Context-Aware Search

The bot now provides two types of search:

1. Channel-Specific Search (/ask):

   • Searches only within the current channel's indexed content
   • Provides more focused, relevant responses
   • Use when asking about channel-specific topics

2. Server-Wide Search (/ask-server):

   • Searches across all indexed channels in the server
   • Provides broader context from the entire server
   • Use when asking about server-wide topics or cross-channel information

Permissions

The bot needs "Read Message History" permission to index channels.

Configuration

The bot can be configured through config.yaml:

# Bot Configuration
bot:
  prefix: "!"
  intents:
    message_content: true
    guilds: true
    messages: true

# OpenAI Configuration
openai:
  model: "gpt-4o-mini"
  max_tokens: 1000
  temperature: 0.7

# Local Embeddings Configuration
embeddings:
  model_name: "sentence-transformers/all-MiniLM-L6-v2"

# ChromaDB Configuration
chromadb:
  persist_directory: "./data"
  collection_name: "discord_knowledge"

# Indexing Configuration
indexing:
  chunk_size: 1000
  max_messages_per_request: 100
  rate_limit_delay: 1.0

Architecture

Components

1. Discord Bot Core (bot/)

   • Main bot functionality
   • Slash command handling
   • Event processing

2. Indexing System (indexing/)

   • Message collection with pagination
   • Text processing and chunking
   • ChromaDB storage management

3. AI Chat System (chat/)

   • OpenAI integration
   • Context building with channel filtering
   • Response generation

4. Utilities (utils/)

   • Configuration management
   • Text processing helpers

Data Flow

1. Indexing: Messages → Text Processing → ChromaDB Storage
2. Chat: User Query → Context-Aware Search → AI Response
3. Search: Query → ChromaDB Search (with optional filtering) → Formatted Results

Data Flow & Privacy

• Stays local: Discord message data, embeddings (vector representations), and the vector database (Chroma) all reside on your machine.
• Sent to OpenAI (required today): For each response, the bot sends the user's prompt plus a small, selected set of retrieved context snippets to OpenAI for final response generation.
• Planned: Support for selecting alternative model providers via LlamaIndex so responses can be generated without OpenAI.

Technical Details

• Local Vector Database: ChromaDB provides local vector storage - no external database required
• Local Embeddings: Uses sentence-transformers/all-MiniLM-L6-v2 for local embedding generation - no external embedding API needed
• OpenAI Integration: Only external dependency is OpenAI API for final chat completion (required today)
• Text Processing: MVP functionality with basic cleaning and chunking
• Vector Search: ChromaDB with LlamaIndex for semantic search
• Context Filtering: Channel-specific search with metadata filtering
• Slash Commands: Modern Discord slash command interface
• Permission System: Admin-only commands for destructive operations

Model Providers (Planned)

We plan to support selecting among multiple LLM providers supported by LlamaIndex for response generation. See the official LlamaIndex documentation for provider options:

• LlamaIndex LLMs (model providers)

Once implemented, you will be able to choose a provider in configuration; until then, OpenAI is required.

FAQ

• Do you send our data to OpenAI?
  • Yes, today the bot sends the user prompt and selected retrieved context to OpenAI to generate the final response. Raw server data and the full vector database remain local.
• Does OpenAI train on our API data?
  • No. Per OpenAI's API data usage policy, data submitted via the API is not used to train OpenAI models unless you explicitly opt in. See: OpenAI API Data Usage Policy
• What is “final chat completion”?
  • The last step where an LLM takes your prompt plus retrieved context and generates the natural-language reply.
• Can I avoid OpenAI?
  • Not yet. OpenAI is required today. Planned support will allow selecting other providers via LlamaIndex; see the link above.
• How long does indexing take?
  • Depends on server size and rate limits. For a mid-size server (thousands of users, dozens of channels, multiple years), initial indexing can take hours.

Command Examples

Indexing

/index-server
/index-channel #general
/reindex-server

Chat

/ask What was discussed about the new feature?
/ask-server Who mentioned the meeting yesterday?

Management

/status
/stats

Security

• Admin-Only Commands: Indexing and destructive operations require Administrator permissions
• Permission Checks: All commands verify user permissions before execution
• Confirmation Dialogs: Destructive operations require explicit confirmation
• Error Handling: Comprehensive error handling and logging