elisey/gpt_assistant_lib

Python library for your AI assistant, powered by GPT.

GPT-Assistant Library

The GPT-Assistant Library is a Python-based toolkit for building intelligent personal assistants powered by OpenAI's GPT models. This library simplifies the integration of conversational AI into your applications, making it easier than ever to create chatbots, virtual assistants, and more.

Key Features

• Seamless integration with GPT-3 and GPT-4 models.
• Easily manage conversations and user interactions.
• Support multiuser chat history support
• Customizable and extensible to suit your specific needs.
• Developer-friendly API for smooth integration into your projects.

Installation

pip install gpt-assistant-lib

Configuration

To set up the assistant, you need an API key and to create a system prompt.

Obtain API key

Go to Openai website

System prompt

It will define your assistant behaviour. More info you can find in documentation

History

The library maintains a conversation history that the assistant relies on to understand and maintain context during interactions. This conversation history is included in every request sent to OpenAI. A larger history size allows the assistant to remember more context, enhancing its ability to engage in meaningful conversations. However, it's important to note that larger conversation histories can result in increased API call costs.

Usage

import gpt_assistant_lib

openai_api_key = "your_openai_api_key"
initial_prompt = "You are a helpful assistant."
max_history_size = 5
history_lifetime = 600  # 10 minutes
assistant = gpt_assistant_lib.build_assistant(openai_api_key, initial_prompt, max_history_size, history_lifetime)
while True:
    user_input = input("You: ")
    assistant_response = assistant.exchange("any", user_input)
    print(f"Assistant: {assistant_response}")

Development

Run with enabled virtual environment

Autoformat source code

task format

Run all linters

task lint

Run unit tests

task test

Run all linters and unit tests

task all

License

This project is licensed under the MIT License