OpenAI Decorator

The OpenAI Decorator is a Python package that enriches your Python functions with the capabilities of OpenAI's API. It leverages the function's signature and docstring to generate specifications comprehensible by OpenAI's API, subsequently making a request to the API.

Table of Contents

• Key Features
• Installation
• Usage Guide
  • Basic Usage
  • Type Annotations & Optionals
  • Passing a Callable Prompt
  • Using the Decorator as a Regular Function
• Contributing
  • Setup
• Future Enhancements
• License

Key Features

• Automatic generation of specs from the function's signature and docstring.
• Integration with OpenAI's API using the generated specs and a customizable prompt.
• Inference of function arguments from the OpenAI API response, negating the need for manual input.
• Handling of various edge cases related to function signatures, including default and optional values.

Installation

Install the package using pip:

pip install openai-decorator

Save your OpenAI key in an environment variable named OPENAI_API_KEY:

export OPENAI_API_KEY=<your key>

Usage Guide

Basic Usage

Apply the openai_func decorator to a function with type annotations and a well-formatted Google Style docstring. Then, it will seamlessly interact with OpenAI's API to generate the parameters for your function based on the prompt.

from openai_decorator import openai_func

@openai_func(prompt="Your OpenAI prompt here")
def example_function(arg1: str, arg2: int) -> str:
    """
    This is an example function.

    Args:
        arg1 (str): Description for arg1
        arg2 (int): Description for arg2

    Returns:
        str: Description for return value
    """
    return f"Your output here: {arg1} and {arg2}"

result = example_function()
print(result)

Type Annotations & Optionals

The decorator efficiently handles complex function signatures:

• Default Values: Parameters with a default value or Optional type annotation are marked as optional in the JSON specification.
• Type Annotations: Supports primitive types (int, str, bool, float) and complex types (List, Dict, Tuple, Set, Optional, custom classes).

from openai_decorator import openai_func

@openai_func(prompt="Perform mathematical operations")
def example_function(a: int, b: int, c: float = 0.0, d: List[int] = []):
    """
    This is an example function.

    Args:
        a: The first integer.
        b: The second integer.
        c: An optional float.
        d: An optional list of integers.
    """
    # Function body here

Passing Callable Prompts

In addition to static strings, you can also pass a callable to the prompt parameter in the openai_func decorator to dynamically generate prompts. This callable should take no arguments and return a string. Here's an example:

from openai_decorator import openai_func

def generate_prompt():
    # Dynamically generate a prompt.
    # In a real-world scenario, this could depend on various factors,
    # such as the current time or the latest trending topics.
    return "Your OpenAI prompt here"

@openai_func(prompt=generate_prompt)
def example_function(arg1: str, arg2: int) -> str:
    """
    This is an example function.

    Args:
        arg1 (str): Description for arg1
        arg2 (int): Description for arg2

    Returns:
        str: Description for return value
    """
    return f"Your output here: {arg1} and {arg2}"

result = example_function()
print(result)

In this example, every time example_function is called, `generate_prompt

Using the Decorator as a Regular Function

openai_func can also be used as a regular function. First, call openai_func with the prompt and any other arguments to get the actual decorator. Then, apply this decorator to your function.

from openai_decorator import openai_func

def example_function(arg1: str, arg2: int) -> str:
    """
    This is an example function.

    Args:
        arg1 (str): Description for arg1
        arg2 (int): Description for arg2

    Returns:
        str: Description for return value
    """
    return f"Your output here: {arg1} and {arg2}"

# Use openai_func as a regular function
prompt = "Your OpenAI prompt here"

example_func_with_generated_params = openai_func(prompt=prompt)(example_function)

result = example_func_with_generated_params()
print(result)

This approach gives you more flexibility in determining the prompt at runtime.

Contributing

Contributions are welcome! Feel free to submit issues and pull requests.

Setup

1. Install Python 3.11 and the latest version of poetry
   • pyenv can help manage multiple Python versions.
2. Clone the repository: git clone.
3. Set your OpenAI key as an environment variable:

export OPENAI_API_KEY=<insert your openai key>

4. Install dependencies: poetry install --no-root

Future Enhancements

• Handle optional parameters.
• Handle parameters with default values.
• Add test to ensure default values are used if OpenAI doesn't return parameters.
• Publish the package to PyPI.
• Expand test coverage.
• Show tests passing & coverage as github badges.
• Fix CI/CD -- Address the issue where the pre-release deployment to PyPI fails due to the need for version update.
• Add docstrings.
• Generate docs (from docstrings).

License

This project is under the MIT License. See the LICENSE file for more details.