Quick Start Guide

This guide will help you set up and run your MilesONerd AI Telegram bot in just a few minutes. By the end, you’ll have a fully functional AI-powered bot responding to messages on Telegram.

Prerequisites

Before you begin, ensure you have the following:

Python 3.8 or higher installed on your system

pip (Python package manager)

A Telegram account

Git (for cloning the repository)

GPU Recommended: While the bot can run on CPU, using a GPU with at least 40GB VRAM is highly recommended for optimal performance with the Llama 3.1-Nemotron 70B model.

Get Your Telegram Bot Token

You need a bot token from Telegram to connect your bot:

Open Telegram and search for @BotFather

Start a chat and send the command /newbot

Follow the prompts to name your bot:

Name: MilesONerd AI (or your preferred name)
Username: Must end in “bot” (e.g., milesonerd_ai_bot)

Copy the bot token provided by BotFather - you’ll need this in the next steps

Save your bot token securely. You’ll use it in the configuration step. The token looks like: 1234567890:ABCdefGHIjklMNOpqrsTUVwxyz

Clone the Repository

Clone the MilesONerd AI bot repository to your local machine:

git clone https://github.com/MilesONerd/telegram-bot.git
cd telegram-bot

Install Dependencies

Install all required Python packages using pip:

pip install -r requirements.txt

The installation includes:

python-telegram-bot (21.10) - Telegram Bot API wrapper
transformers (4.48.0) - Hugging Face model library
torch (2.5.1) - PyTorch for model inference
python-dotenv (1.0.1) - Environment variable management
Additional dependencies for model optimization

This may take several minutes as it installs PyTorch, Transformers, and other dependencies.

Configure Environment Variables

Set up your bot configuration:

Create a .env file in the project root:

cp .env.example .env

If .env.example doesn’t exist, create .env manually:

touch .env

Edit the .env file and add your Telegram bot token:

# Required: Your Telegram Bot Token from @BotFather
TELEGRAM_BOT_TOKEN=your_bot_token_here

# Optional: Default AI model to use
DEFAULT_MODEL=llama

# Optional: Enable continuous learning (future feature)
ENABLE_CONTINUOUS_LEARNING=true

# Optional: Google Search API key (future feature)
# SERPAPI_API_KEY=your_serpapi_key_here

Security: Never commit your .env file to version control. The token grants full access to your bot.

Start the Bot

Launch your bot with a single command:

python bot.py

You should see output indicating the models are being initialized:

2026-03-09 10:30:45 - __main__ - INFO - Initializing AI models...
2026-03-09 10:30:45 - ai_handler - INFO - Starting model initialization...
2026-03-09 10:30:45 - ai_handler - INFO - CUDA available: True
2026-03-09 10:30:45 - ai_handler - INFO - GPU Device: NVIDIA A100-SXM4-80GB
2026-03-09 10:30:46 - ai_handler - INFO - Loading BART model: facebook/bart-large
2026-03-09 10:30:52 - ai_handler - INFO - BART tokenizer loaded successfully
2026-03-09 10:30:58 - ai_handler - INFO - BART model loaded successfully
2026-03-09 10:30:58 - ai_handler - INFO - Loading Llama model: nvidia/Llama-3.1-Nemotron-70B-Instruct-HF
2026-03-09 10:31:15 - ai_handler - INFO - Llama tokenizer loaded successfully
2026-03-09 10:32:45 - ai_handler - INFO - Llama model loaded successfully
2026-03-09 10:32:45 - ai_handler - INFO - All models initialized successfully
2026-03-09 10:32:45 - __main__ - INFO - MilesONerd AI Bot is starting...

First Run: The first time you run the bot, it will download the AI models from Hugging Face (~150GB total). This can take 10-30 minutes depending on your internet connection. Subsequent runs will use cached models.

Test Your Bot

Now that your bot is running, let’s test it on Telegram:

Open Telegram on your phone or desktop

Search for your bot using the username you created (e.g., @milesonerd_ai_bot)

Start a conversation by clicking “Start” or sending /start

You should receive a welcome message:

Hi @yourusername! I'm MilesONerd AI, your intelligent assistant.
I can help you with various tasks using advanced AI models and internet search.
Use /help to see available commands.

Try Basic Commands

Test the bot’s functionality with these commands:

/start

/start

/help

/help

/about

/about

Expected responses:

/start

Hi @yourusername! I'm MilesONerd AI, your intelligent assistant.
I can help you with various tasks using advanced AI models and internet search.
Use /help to see available commands.

/help

Available commands:
/start - Start the bot
/help - Show this help message
/about - Learn more about MilesONerd AI

Message handling:
- Short questions: Quick responses using lightweight model
- Long messages: Summarization and detailed response
- Include 'summarize' or 'tldr' for text summarization
- Chat-related queries: Optimized conversation handling
- Regular messages: Comprehensive AI-powered responses

You can send me any message, and I'll process it using the most appropriate AI model!

/about

MilesONerd AI is an intelligent assistant powered by advanced AI models and internet search capabilities.

Features:
- Advanced language understanding
- Internet search integration
- Continuous learning from interactions
- Multiple AI models for different tasks

Created with ❤️ using python-telegram-bot and Hugging Face models.

Test Message Processing

Try different types of messages to see how the bot responds:

1. Short Question (uses Llama 3.1-Nemotron)

What is Python?

2. Conversation Request (uses Llama 3.1-Nemotron)

Let's chat about artificial intelligence

3. Summarization Request (uses BART)

Please summarize this: [paste a long article or text]

tldr: [paste a long article or text]

4. Long Message (uses BART + Llama 3.1-Nemotron)

Send a message with more than 100 words. The bot will:

Summarize it using BART

Generate a response using Llama based on the summary

Understanding the Bot’s Behavior

The bot intelligently routes your messages to different AI models based on content analysis:

Message Routing Logic

# From bot.py - How the bot decides which model to use

if len(user_message.split()) > 100:
    # Long messages: BART summarization + Llama response
    summary = await ai_handler.summarize_text(user_message)
    response = await ai_handler.generate_response(
        f"Based on this summary: {summary}\nGenerate a helpful response:",
        model_key='llama',
        max_length=200
    )

elif 'summarize' in user_message.lower() or 'tldr' in user_message.lower():
    # Explicit summarization: BART only
    response = await ai_handler.summarize_text(user_message)

elif 'chat' in user_message.lower() or 'conversation' in user_message.lower():
    # Conversation: Llama with extended length
    response = await ai_handler.generate_response(
        user_message,
        model_key='llama',
        max_length=200
    )

elif len(user_message.split()) < 10:
    # Short queries: Llama with limited length
    response = await ai_handler.generate_response(
        user_message,
        model_key='llama',
        max_length=100
    )

else:
    # Default: Llama with standard length
    response = await ai_handler.generate_response(
        user_message,
        model_key='llama',
        max_length=150
    )

Model Selection Table

Message Type	Word Count	Keywords	AI Model Used	Max Length
Long message	> 100 words	-	BART → Llama	200 tokens
Summarization	Any	”summarize”, “tldr”, “summary”	BART	130 tokens
Conversation	Any	”chat”, “conversation”, “talk”	Llama	200 tokens
Short query	< 10 words	-	Llama	100 tokens
Default	10-100 words	-	Llama	150 tokens

Troubleshooting

Bot Token Error

ERROR - No token found! Make sure to set TELEGRAM_BOT_TOKEN in .env file

Solution: Verify your .env file exists and contains the correct token:

cat .env

Ensure the line reads:

TELEGRAM_BOT_TOKEN=your_actual_token_here

Model Initialization Failed

ERROR - Failed to initialize AI models. Exiting...

Possible causes:

Insufficient GPU/CPU memory
Network issues downloading models
Missing dependencies

Solutions:

Check GPU memory:

nvidia-smi

Verify PyTorch installation:

python -c "import torch; print(torch.cuda.is_available())"

Reinstall dependencies:

pip install -r requirements.txt --upgrade

Out of Memory Error

RuntimeError: CUDA out of memory

Solution: The Llama 3.1-Nemotron 70B model requires significant memory. Options:

Use a smaller model (modify ai_handler.py:35)
Enable CPU offloading (already configured with device_map='auto')
Use quantization (future enhancement)

Bot Not Responding

Checklist:

Is bot.py running without errors?
Did you use the correct bot username in Telegram?
Is your internet connection stable?
Check logs for error messages

Model Download Takes Too Long

The models are large (~150GB combined). To monitor progress:

watch -n 1 du -sh ~/.cache/huggingface/

You can manually pre-download models using the huggingface-cli:

huggingface-cli download nvidia/Llama-3.1-Nemotron-70B-Instruct-HF
huggingface-cli download facebook/bart-large

Configuration Options

Environment Variables

Customize your bot’s behavior by editing .env:

.env

# Required
TELEGRAM_BOT_TOKEN=your_bot_token_here

# Optional: Choose default model (llama or bart)
DEFAULT_MODEL=llama

# Optional: Enable/disable continuous learning
ENABLE_CONTINUOUS_LEARNING=true

# Optional: Future feature - internet search
SERPAPI_API_KEY=your_api_key_here

Model Parameters

You can adjust generation parameters in ai_handler.py:

# Default parameters in generate_response() at line 109
temperature: float = 0.2  # Lower = more focused, Higher = more creative
top_p: float = 0.4        # Nucleus sampling threshold
max_length: int = 300     # Maximum response length in tokens
max_attempts: int = 5     # Retry attempts for valid responses

Next Steps

Now that your bot is running, explore more:

API Reference

Learn about all available commands and functions

Advanced Configuration

Customize model parameters and behavior

Deployment Guide

Deploy your bot to production servers

Model Details

Contribute to the MilesONerd AI project

Getting Help

If you encounter issues:

Check the GitHub Issues
Review the source code
Contact the author: MilesONerd

Logging: The bot outputs detailed logs to the console. Use these logs to diagnose issues:

python bot.py 2>&1 | tee bot.log

Congratulations! You now have a fully functional AI-powered Telegram bot. Start chatting and explore its capabilities!

Get Started

Guides

AI Models

Quick Start

Quick Start Guide

Understanding the Bot’s Behavior

Message Routing Logic

Model Selection Table

Troubleshooting

Bot Token Error

Model Initialization Failed

Out of Memory Error

Bot Not Responding

Model Download Takes Too Long

Configuration Options

Environment Variables

Model Parameters

Next Steps

API Reference

Advanced Configuration

Deployment Guide

Model Details

Getting Help

Build docs developers (and LLMs) love

Get Started

Guides

AI Models

​Quick Start Guide

​Understanding the Bot’s Behavior

​Message Routing Logic

​Model Selection Table

​Troubleshooting

​Bot Token Error

​Model Initialization Failed

​Out of Memory Error

​Bot Not Responding

​Model Download Takes Too Long

​Configuration Options

​Environment Variables

​Model Parameters

​Next Steps

API Reference

Advanced Configuration

Deployment Guide

Model Details

​Getting Help

Build docs developers (and LLMs) love

Quick Start Guide

Understanding the Bot’s Behavior

Message Routing Logic

Model Selection Table

Troubleshooting

Bot Token Error

Model Initialization Failed

Out of Memory Error

Bot Not Responding

Model Download Takes Too Long

Configuration Options

Environment Variables

Model Parameters

Next Steps

Getting Help