pyAGI

pyAGI is a Python package that functions as an autonomous agent for your Python application. By interfacing with OpenAI models, it utilizes a sequence of prompts to instruct the AI in generating various outputs such as application descriptions, architecture outlines, user experience flows, and even code samples. All these are achieved using the langchain framework to construct chains of language models.

Features

1. Supports several OpenAI models, including a range of GPT-3 and Davinci models.
2. Provides functionality for generating:
   1. Application descriptions
      1. Software architecture outlines
      2. User experience flows
      3. Code flow descriptions
      4. Coding steps
      5. Application code

Setup

To use pyAGI, you need to have an OpenAI account and obtain your API key. Once you have your OpenAI API key, set it as an environment variable OPENAI_API_KEY in your system.

export OPENAI_API_KEY='your-api-key-here'

Installation

To install pyAGI, use pip:

pip install pyAGI

The package is available at PyPI pyAGI (https://pypi.org/project/pyAGI/).

Usage

After installing, you can use pyAGI to generate content for your application as follows:

from pyAGI.pyagi import PyAGI

Create a PyAGI instance with a selected OpenAI model
py_agi = PyAGI.create_llm_chain(selected_model="text-davinci-003")

# Run the PyAGI instance with the objective
py_agi({"objective": "Your Objective", "selected_model": "text-davinci-003"})

Usage without installing the pip package

python main.py "Objective of your app" "OpenAI Model"

For example:

python main.py "Build a weather app" "text-davinci-003"

Available Models

Currently supported models:

• "text-davinci-002",
• "text-davinci-003",
• "gpt-3.5-turbo",
• "gpt-3.5-turbo-0301",
• "text-curie-001",
• "text-babbage-001",
• "text-ada-001"

Upcoming models (https://openai.com/waitlist/gpt-4-api):

• "gpt-4",
• "gpt-4-0314",
• "gpt-4-32k",
• "gpt-4-32k-0314"

Please note that upcoming models will not be supported until released.

Contributing

Contributions are welcome! Please open an issue if you encounter a bug, or a pull request if you have improvements to the existing code.

License

This project is licensed under the MIT License.