An interface to query OpenAI's GPT 4 model

The GPT class provides functionality to query GPT 4 in code with the following features:

• Ask GPT has a ask function which can be used to interact with GPT 4
• Chat interface The store_chat_history parameter in ask allows you to track messages history
• Configuration A file called gpt.yaml that is used to initialize GPT with your API key, temperature, etc...
• Retry Thetenacity library automatically retries the query when encountering RateLimitErrors from OpenAI
• Logging The hydra library creates a new logging directory each time GPT is initialized and logs conversations by date

Getting Started

pip install openai
pip install tenacity
pip install hydra-core --upgrade

❗ Note, that you will need to fill in the gpt.yaml file with your OpenAI information such as the API key.

gpt: 
  _target_: gpt4.GPT
  openai_api_type:  
  openai_api_base: 
  openai_api_version: 
  openai_api_key: 
  engine: largeGPT4
  temperature: 0.7
  system_prompt: You are a helpful assistant.
  log_dir: ${hydra:run.dir}

Usage and Tutorials

Python

import hydra
from omegaconf import DictConfig, OmegaConf
from gpt4 import GPT 

@hydra.main(version_base=None, config_path=".", config_name="gpt")
def experiment(cfg): 
    gpt = hydra.utils.instantiate(cfg.gpt)
    reply = gpt.ask('What year is it?', verbose=True, store_chat_history=False)

if __name__ == "__main__":
    experiment()

Jupyter notebook

import os
import hydra 
from omegaconf import OmegaConf
from gpt4 import GPT 

with hydra.initialize(version_base=None, config_path="."):
    cfg = hydra.compose(config_name="gpt", return_hydra_config=True, overrides=["gpt.log_dir=${hydra.run.dir}"])
    os.makedirs(cfg.gpt.log_dir, exist_ok=True)

gpt = hydra.utils.instantiate(cfg.gpt)
reply = gpt.ask('What year is it?', verbose=True, store_chat_history=False)