ChatGPT & DALL-E 2 with Python

This is a simple Python script to access ChatGPT and DALL-E 2 within Jupyter Notebook.

The inspiration and core code comes from an article by Sophia Yang in Medium.

My humble additions are to loop the query, opt to do an image prompt (see below) and to save images and/or conversations.

Its been handy to have ChatGPT on call. It is definitely an interesting and valuable tool, though I have found it necessary to verify responses as errors are common. I’ve had the same answer to opposite questions and seen errors in basic arithmetic. A way I look at ChatGPT is similar to Wikipedia, in that it can help get you started, but you should still go the source for verification.

Similarly with coding - at times ChatGPT is a great help, and at other times is just plain wrong. Though, sometimes the approach to a solution can help prompt my own thinking to solve the problem at hand. So, another way to think of ChatGPT is like a sounding board or a friend who is really trying to help. Used in this way, I’ve found ChatGPT to be a useful tool on a few projects to get me thinking about things in a different way.

Getting ready

Get OpenAI Key - you’ll need to set up a developer account to create your personal API key:

https://platform.openai.com/overview

I believe the first month you get US$18 credit and then pay for monthly usage. With light to moderate usage, my bills are $1 per month or less

Create two folders in the same directory as the script for saving files:

chats
images

Prompts:

ChaptGPT: default - just enter prompt
DALL-E 2: type image: plus text prompt
Save image: after image is generated, type save
To finish: type exit

After you finish a chat, you will be asked if you wish to save the conversation.

Setup

from chatgpt import Conversation
import urllib.request
import sys
import os
import openai
from IPython.display import display, HTML
from IPython.display import Image

import pandas as pd

working_directory = os.getcwd()
working_directory = "/Users/adam/data/ChatGPT/"

# Enter personal OpenAI key here - see link above
openai.api_key = "<INSERT OPENAI KEY HERE>"

Functions

A list of models

If you wish to try other models, a list can be obtained with use of this code:

models = pd.DataFrame(openai.Model.list()["data"])

# Functions

def openai_completion(prompt):
    # Send prompt/Get ChatGPT reponse
    response = openai.Completion.create(
      model="text-davinci-003",  # this can be changed to access different models
      prompt=prompt,
      max_tokens=2000,           # you can change this to different values for longer shorter prompts/replies
      temperature=0.5            # the "straightness" of the replies 
    )
    return response['choices'][0]['text']


def openai_image(prompt):
    #Send prompt/Get DALL-E 2 response
    response = openai.Image.create(
      prompt=prompt,
      n=1,
      size="256x256"             # this can be different values "256x256", "512x512", "1024x1024"
    )
    image_url = response['data'][0]['url']
    return image_url

# 
def append_strings_to_df(question, response):
    # Conversation text as dataframe - ready for saving upon request (enter "save" as prompt)
    global conversation
    dict = {"Question": question, "Response": response}
    data = pd.DataFrame(dict, index=[0])
    conversation = pd.concat([conversation, data])

Main prompt loop

# Set up empty conversation df
conversation = pd.DataFrame(columns=["Question", "Response"])

# Main prompt loop
while True:
    question = input(">>")
    if question.lower() == "exit":
        append_strings_to_df(question, "")
        break
    elif question[0:7].lower() == "image: ":            #create image based on prompt after "image: "
        image_link = openai_image(question[7:])
        image_obj = Image(url=image_link)
        display(image_obj)
        file_name = question[7:18] + ".png"             #extracts text after "image: " to create filename
        append_strings_to_df(question, image_link)
    elif question.lower() == "save":                    #to save image
        file_path = working_directory + "images/" + file_name
        urllib.request.urlretrieve(image_link, file_path)
    else:    
        response = openai_completion(question)          #chatGPT response
        print(response)
        print()
        append_strings_to_df(question, response)

# After exit - option to save conversation        
while True:
    save_convo = input("Save the conversation (Y/N)? :").lower()
    if save_convo == "y":
        file_name = input("Name? :")
        conversation.to_csv(working_directory + "chats/" + file_name + ".csv", index=False)
        
        break
    elif save_convo == "n":
        print("No probs - thanks!")
        break