/kurul

a chat bot front-end that produces higher quality responses through a pipeline of skepticism.

Primary LanguagePythonMIT LicenseMIT

kurul

kurul leverages the OpenAI API to provide high-quality responses in a friendly and interactive manner. kurul refines its responses using a pipeline of skepticism to ensure relevant and thoughtful conversations with users.

kurul can be interacted through a Telegram bot.

Dependencies

  • Python 3.10
  • python-telegram-bot
  • openai python package
  • tenacity package

Setup

  1. Clone the repository.
git clone https://github.com/dyigitpolat/kurul.git
  1. Install the required packages.
pip install python-telegram-bot openai
  1. Configure the API keys and authorized chats.

  • Obtain an API key from the OpenAI platform after signing up for a free account.

  • Replace the placeholders in the src/api_keys.py file with your OpenAI API key and Telegram bot token. For example:

openai_api_key = "your_openai_key_here"
telegram_bot_api_key = "your_telegram_bot_token_here"
  • To add authorized users, you need to find their Telegram chat ID. Add the Telegram chat IDs of authorized users to the src/authorized_chats.py file. For example:
authorized_chat_ids = [
    123456789, # UserID 1
    987654321, # UserID 2
]
  1. Set up a Telegram bot by following this guide from the official Telegram documentation.

Usage

cd src
python main.py

The chatbot will automatically respond to messages sent by authorized users on the Telegram platform.

Architecture Overview

kurul's core functionality is encapsulated in the kurul class. It uses a pipeline of skepticism to improve the quality of Chat GPT responses. Main components include:

  • Chat GPT client for initial response generation
  • Context manager for handling conversations context
  • Skeptical prompt generation for creating skepticism queries
  • Response refinement based on the skepticism feedback

License

This project is licensed under the MIT License. Please refer to the LICENSE file for more information.

Contributing

Contributions are welcome! Feel free to submit a pull request, report a bug, or suggest new features by creating an issue.

Please make sure to update tests as appropriate and follow the existing coding style.

Acknowledgments