Universal LLM ChatBot

Universal LLM ChatBot is a versatile Telegram bot that leverages Large Language Models (LLMs) for natural language processing, voice transcription, and text-to-speech capabilities. It supports multiple languages and provides a customizable interface for interacting with AI models.

Features

• Multi-language support
• Voice message transcription
• Text-to-speech responses
• Customizable AI model settings
• User-specific configurations
• Rate limiting to prevent abuse
• Admin commands for user management
• Ollama, Groq, Gemini, OpenAI, and Anthropic support

Prerequisites

• Python 3.10+
• Telegram Bot Token (obtain from BotFather)
• Ollama server running locally or remotely
• PyTorch development environment (see setup instructions below)

Installation

1. Clone the repository:

   git clone https://github.com/KPEKEP/universal-llm-chatbot.git
   cd universal-llm-chatbot
   

2. Install the required dependencies:

   pip install -r requirements.txt
   

3. Copy the config_template.yml to config.yml and update it with your settings:

   cp config_template.yml config.yml
   

4. Set up the required environment variables (see Configuration section).

PyTorch Development Environment Setup

To set up a PyTorch development environment using Conda, follow these steps:

1. Install Anaconda or Miniconda if you haven't already.

2. Create a new Conda environment:

   conda create -n pytorch-env python=3.10
   

3. Activate the environment:

   conda activate pytorch-env
   

4. Install PyTorch and related packages:

   conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch
   

5. Install other required packages:

   pip install -r requirements.txt
   

Configuration

The bot uses a combination of environment variables and a config.yml file for configuration.

Environment Variables

Set the following environment variables:

• UNI_LLM_BOT_TOKEN: Your Telegram Bot Token
• UNI_LLM_ADMIN_USER: Comma-separated list of admin user IDs
• UNI_LLM_ACCESS_MODE: Set to either "public" or "whitelist"

Example:

export UNI_LLM_BOT_TOKEN=your_bot_token_here
export UNI_LLM_ADMIN_USER=123456789,987654321
export UNI_LLM_ACCESS_MODE=public
export UNI_LLM_GROQ_API_KEY=your_groq_api_key_here
export UNI_LLM_GEMINI_API_KEY=your_gemini_api_key_here
export UNI_LLM_OPENAI_API_KEY=your_openai_api_key_here
export UNI_LLM_ANTHROPIC_API_KEY=your_anthropic_api_key_here

Config File

The config.yml file contains additional settings for the bot. Key configurations include:

• AI provider settings
• Rate limiting parameters
• Database settings

Refer to the comments in config_template.yml for detailed explanations of each setting.

Usage

To start the bot, you can use the provided start.bat file as an example. Here's a sample content of start.bat:

SET UNI_LLM_BOT_TOKEN=your_telegram_bot_token_here
SET UNI_LLM_ADMIN_USER_IDS=your_telegram_user_id_here
SET UNI_LLM_ACCESS_MODE=public
SET UNI_LLM_GROQ_API_KEY=your_groq_api_key_here
SET UNI_LLM_GEMINI_API_KEY=your_gemini_api_key_here
SET UNI_LLM_OPENAI_API_KEY=your_openai_api_key_here
SET UNI_LLM_ANTHROPIC_API_KEY=your_anthropic_api_key_here

@call conda activate pytorch-env
@call python main.py

Modify the start.bat file with your specific environment variables and run it to start the bot.

Available Commands

User Commands

• /start: Initialize the bot and receive a welcome message.
• /settings: Access and modify bot settings.
• /reset: Reset your conversation history.
• /history: Export your conversation history.
• /language: Change the bot's language.
• /speaker: Change the bot's voice speaker.

Admin Commands

• /whitelist <user_id or username> [on/off]: Add or remove a user from the whitelist.
• /blacklist <user_id or username> [on/off]: Add or remove a user from the blacklist.
• /grant_admin <user_id or username> [on/off]: Grant or revoke admin privileges for a user.
• /broadcast <message>: Send a message to all users.
• /getid <username>: Get the user ID for a given username.

Command Usage Examples

1. Whitelist a user:

   /whitelist @username on
   

2. Blacklist a user:

   /blacklist 123456789 on
   

3. Grant admin privileges:

   /grant_admin @username on
   

4. Send a broadcast message:

   /broadcast Hello, this is an important announcement!
   

5. Get a user's ID:

   /getid @username
   

Note: Admin commands are only available to users with administrative privileges as defined in the UNI_LLM_ADMIN_USER environment variable.

Extending Beyond BasicProvider

The Universal LLM ChatBot is designed to be extensible. To create a new provider:

1. Create a new file in the bot/providers/ directory (e.g., custom_provider.py).
2. Implement a class that inherits from the Provider base class in bot/provider.py.
3. Override the abstract methods: generate_response, transcribe_voice, and text_to_speech.
4. Update the config.yml to use your new provider.

Example:

from bot.provider import Provider

class CustomProvider(Provider):
    def __init__(self, provider_name, config):
        super().__init__(provider_name, config)
        # Initialize your custom provider here

    async def generate_response(self, model, messages, options):
        # Implement your custom response generation logic

    async def transcribe_voice(self, input_filename):
        # Implement your custom voice transcription logic

    async def text_to_speech(self, text, output_filename, language="en", speaker=None):
        # Implement your custom text-to-speech logic

    async def get_models(self):
        # Implement your custom available models list retrieval logic
        pass        

Ollama Setup

For detailed instructions on setting up Ollama for local inference, please refer to the Ollama documentation.

Acknowledgements

This project utilizes several open-source libraries and models:

• python-telegram-bot for Telegram integration
• Ollama for local LLM inference
• Whisper for voice transcription
• Coqui TTS for text-to-speech generation

Appreciation

If you use this project in your research or application, please give it a star

License

This project is licensed under the MIT License - see the LICENSE file for details.