DexFren AI Telegram chatbot Bot Documentation

Overview

DexFren AI chatbot is a Telegram bot powered by GPT-4 that helps users understand and use the DexAppBuilder platform. The bot uses a knowledge base built from DexKit's documentation and YouTube tutorials.

Setup Requirements

Dependencies

pip install python-telegram-bot pip install langchain-chroma pip install python-dotenv pip install swarm pip install youtube-transcript-api

Environment Variables

TELEGRAM_BOT_TOKEN OPENAI_API_KEY

Project Structure

project/
├── docs/                   # PDF documentation (if any)
├── knowledge/             
│   ├── __init__.py
│   └── data_ingestion.py  # Knowledge base processing
├── knowledge_base/         # Vector store data
├── .env                   # Environment variables
├── .gitignore            
├── README.md              # Simplified documentation
├── main.py                # Bot implementation
└── build_knowledge_base.py # Knowledge base builder

Features

• Natural language processing for DexKit-related queries
• Context-aware responses based on documentation and tutorials
• Support for both English and Spanish video transcripts
• Real-time responses using GPT-4
• Persistent knowledge base using Chroma vector store

Usage

Building the Knowledge Base

1. Place any PDF documentation in the docs/ directory
2. Update YouTube URLs in build_knowledge_base.py
3. Run the knowledge base builder:

python build_knowledge_base.py

Running the Bot

python main.py

Interacting with the Bot

1. Find the bot on Telegram using the bot username
2. Start a conversation with /start
3. Ask questions about:
   • DexAppBuilder usage
   • Platform features
   • Technical guidance
   • Best practices

Architecture

Components

• Telegram Bot: Handles user interactions
• Swarm Agent: Processes queries using GPT-4
• Knowledge Base: Stores and retrieves relevant context
• Vector Store: Manages embeddings for efficient searching

Data Flow

1. User sends message via Telegram
2. Bot queries knowledge base for relevant context
3. Context is sent to Swarm Agent with user query
4. GPT-4 generates response using context
5. Response is sent back to user