- Specify expected behavior via python function doc string. Use
gpt_caller()decorator on the function to indicate a gpt function. - Uses
gpt-3.5-turboby default. You can also use switch to other models likegpt-4viagpt_caller(model="gpt-4") - Call gpt functions as reuglar python functions.
⚠️ gpt functions could hallucinate due to its natural. Use them at your own risk!- 🚀 unlock maximum potential by utilizing GPT functions and GitHub Copilot together.
import openai
from gpt_function import gpt_caller
#set your openai api key
openai.api_key = os.environ.get("OPENAI_API_KEY") or ""
#declare the function and expected behavior
@gpt_caller()
def get_US_states(region: str = "All") -> list[dict]:
"""Return a list of dictionaries containing the name, abbreviation, and capital of each US state filtered by region.
Parameters:
region (str): The region of the US states to return. eg: "West_Coast"
Returns:
list[dict]: A list of dictionaries containing the name, abbreviation, and capital of each US state. eg:
[
{
"name": "Alabama",
"abbreviation": "AL",
"capital": "Montgomery"
},
...
]
"""
pass
#use the gpt function as regular python function
states = get_US_states(region="West_Coast_Mainland")
for state in states: # California, Oregon, Washington
print(state["name"], state["abbreviation"], state["capital"])git clone https://github.com/xuanli/gpt-function.git
cd gpt-function
pip install -r requirements.txt
- The idea is inspired from Recurrency AI Hackathon