Telegram Multi-Agent AI Bot

Telegram Multi-Agent AI Bot is a powerful, AI-driven chatbot designed to intelligently process and respond to user messages on Telegram. Leveraging OpenAI GPT-4o-mini, LangChain, and MongoDB Atlas Search, this bot consolidates multiple short messages into a context-aware conversation, retrieves relevant information from stored documents, and generates coherent responses.

🚀 Features

• 💬 Telegram Bot Integration – Seamlessly interacts with users on Telegram.
• 🧠 AI-Powered Message Processing – Uses LangChain and OpenAI GPT-4o-mini for intelligent responses.
• 🪠 Contextual Memory (MongoDB Atlas Search) – Retrieves and utilizes previous conversations for enhanced response accuracy.
• 📄 File Upload & Retrieval – Stores documents in MongoDB Atlas Search for retrieval-augmented generation (RAG).
• 🗂️ MongoDB-Based Message Queue – Manages incoming messages efficiently.
• ✨ FastAPI Web Server – Provides a scalable and modern backend.
• 🛠️ Python Virtual Environment Support – Ensures easy dependency management.

🛠 Installation Guide

1. Set Up a Virtual Environment

# Create a virtual environment
make venv

# Activate the virtual environment (MacOS/Linux)
source venv/bin/activate

2. Install Dependencies

make install

3. Set Up MongoDB Atlas

• Create a MongoDB Atlas cluster.
• Enable Atlas Search and configure an index for vector search.
• Obtain your MongoDB connection string.

4. Configure Environment Variables

Create a .env file and add the following:

TELEGRAM_BOT_TOKEN=your_telegram_bot_token
OPENAI_API_KEY=your_openai_api_key
MONGODB_URI=your_mongodb_atlas_connection_string

▶️ Usage

1. Activate the Virtual Environment

source venv/bin/activate

2. Start the Application

make run

3. Interact with the Bot

• Start by sending /start to the bot.
• Send multiple short messages within 15 seconds for contextual merging.
• Upload files (PDF, TXT, etc.) to enrich responses with document-based knowledge.

🎨 System Architecture

1. Message Reception
   • The Telegram bot receives incoming messages.
   • Messages are stored in a MongoDB queue.

2. File Upload & Processing
   • Users upload files via Telegram.
   • Documents are stored in MongoDB Atlas Search for future retrieval.

3. Message Processing

   • Messages sent within 15 seconds are consolidated.
   • Context is retrieved from MongoDB Atlas Search, including previously uploaded documents.
   • OpenAI's GPT-4o-mini, via LangChain, generates responses.

4. Response Delivery

   • The response is sent back to the user.
   • Interaction data is stored in MongoDB Atlas Search.
   • Processed messages are removed from the queue.

📃 Makefile Commands