A base client for interacting with the OpenAI API and custom clients/mixins to make developing things like chatbots super quick!
Project description
OpenAI API Python Client
A (very rough WIP) base Python 3.9+ client for OpenAI APIs and some concrete "stateful" clients for common operations like running a chatbot with context
Overview
Right now this just provides a base client that allows a reusable way to do common things (like loading the API key or backing-off/retrying on an APIError) plus some easy overrides (like which model to use) and some mixins for when state dependence is important.
Now this also includes some pre-built "recipe" clients:
- Text/Code Completion Clients
- ChatBot Client
- Image Generation Client
- Image Edit Client
- Image Variation Client
Example
All I've got in lieu of real docs for now :)
- Install this library as a PyPi package
pip install OpenAI-API-Python-Client
- Export your OpenAI API key in your shell environment
export OPENAI_API_KEY=AKIAIOSFODNN7EXAMPLE
- Use a recipe class to create apps!
Chatbot Client
# Import this library :)
from openai_client import Chat_Bot_Client
# Simple D&D chatbot app :)
if __name__ == "__main__":
# API Key is read from OPENAI_API_KEY environmental variable
client = Chat_Bot_Client()
# Add a high level directives to guide the model
client.add_directive("You are a Dungeons and Dragons Dungeon Master. Use the 5th edition of the Dungeons and Dragons Player Handbook, Dungeon Master Guide, and Monster Manual")
client.add_directive("At the beginning of your chat with the user you will assist them in creating a character. This character will have a description and stats as outlined in the 5th edition of the Dungeons and Dragons Player Handbook.")
client.add_directive("Let the user choose race and class before assigning a personality, stats, and starting inventory. Provide the user with a list of races and classes they can be. Tell the user they can ask for more details about a class or race")
client.add_directive("Once you introduce the character, give the player the start of an adventure campaign and ask the player what they would like to do")
client.add_directive("As outlined in the handbook, if a roll is necessary based on the situation, roll for the user")
client.add_directive("Finish by asking the player what they'd like to do next")
# Simple loop to run a chat session
while True:
# Send it to the chatbot and get the response
response = client.run_prompt()
# Print the response
print("\n" + response + "\n")
# Get question from the user
client.get_user_input()
Image Generation Client
# Import this library :)
from openai_client import URL_Image_Client
# Simple animal mashup app :)
if __name__ == "__main__":
# API Key is read from OPENAI_API_KEY
client = URL_Image_Client()
# Add a context to always include before the prompt that is sent to the API
client.add_pre_prompt_context("Generate a hybrid animal using the following animals:")
# Add a context to always include after the prompt that is sent to the API
client.add_post_prompt_context("This rendering should be hyperrealistic. The background should be a savannah during the daytime")
# Prompt the user for input
animals = input("Choose two animals to create a hybrid of:\n>>> ")
# Send the request and get the image URL
image_url = client.run_prompt(animals)
# Get the image URL
print(image_url)
Image Edit Client
# Import this library :)
from openai_client import URL_Image_Edit_Client
# Simple image edit app :)
if __name__ == "__main__":
# You can open the file yourself
image = open("src/openai_client/clients/images/demos/image.png", "rb")
# Or just pass a string path
mask = "src/openai_client/clients/images/demos/mask.png"
# API Key is read from OPENAI_API_KEY
client = URL_Image_Edit_Client(image, mask)
# Prompt the user for input (e.g. "a goofy looking cartoon smiley face")
face = input("Describe a face to generate:\n>>> ")
# Send the request and get the image URL
image_url = client.run_prompt(face)
# Get the image URL
print(image_url)
Image Variation Client
# Import this library :)
from openai_client import URL_Image_Variation_Client
# Simple image variation app :)
if __name__ == "__main__":
image = open("src/openai_client/clients/images/demos/image.png", "rb")
# API Key is read from OPENAI_API_KEY
client = URL_Image_Variation_Client(image)
# Send the request and get the image URL
image_url = client.run_prompt()
# Get the image URL
print(image_url)
Text Completion Client - Tweet Classifier
# Import this library :)
from openai_client import Completion_Client, Example
# Simple classifier app :)
if __name__ == "__main__":
e1 = Example('Tweet: "I loved the new Batman movie!"', 'Sentiment: "Positive"')
e2 = Example('Tweet: "I hated that ice cream."', 'Sentiment: "Negative"')
# API Key is read from OPENAI_API_KEY
client = Completion_Client("Decide whether a Tweet's sentiment is positive, neutral, or negative.", [e1,e2])
# Add another example
client.add_examples(Example('Tweet: "I did not like the explosion"', 'Sentiment: "Negative"'))
# Get tweet to classify
tweet_to_classify = "Amusement parks are ok"
# Print the tweet to classify
print(f"\nClassifying tweet: '{tweet_to_classify}'")
# Send to the model with examples
response = client.run_prompt(tweet_to_classify)
# Print result
print(response)
Code Completion Client - Code Generator
# Import this library :)
from openai_client import Code_Completion_Client
import json
# Simple code generation app :)
if __name__ == "__main__":
# API Key is read from OPENAI_API_KEY
client = Code_Completion_Client()
# Get a prompt to generate code with
prompt = "Create a Python dictionary of all US states without any non-code text"
# Print the tweet to classify
print(f"Creating code: '{prompt}'")
# Send to the model with examples
response = client.run_prompt(prompt)
# Load AI generated code into Python object
us_states = json.loads(response)
# Print result for California
print(f"Getting result for key: CA")
print(us_states["CA"])
- Use Mixins and the base class to create new "stateful" clients on top of the base client. See the implementation of Chat_Bot_Client for an example
Contributing
Contributions are welcome! Please not the following when contributing:
-
Unittests must be added under the
tests/
directory for the PR to be approved. You can run unittests from the root project directory with the following command:python setup.py test
-
PRs cannot be merged without all unittests passing (they will execute automatically)
-
Merges to
main
will automatically create a new release on PyPi unless it is from a forked Repo
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for OpenAI API Python Client-1.0.15.tar.gz
Algorithm | Hash digest | |
---|---|---|
SHA256 | 3ae9eb420d0247284b6ab426c1679aedf7bc784e6715d9fad42cb49311afaf7d |
|
MD5 | 26758c954517350cefa119e470edb9dd |
|
BLAKE2b-256 | 5ed262f83e4b859fe325dc8566ecb789fde10bade9c41e6c289e19eb6f2e390e |
Hashes for OpenAI_API_Python_Client-1.0.15-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | b1f463610a44cd632addeaffcf33d7ef9177896fab4ddd92ab8b53e49c8886d1 |
|
MD5 | 04ce0a0d87c387212cac4f9fdd40f580 |
|
BLAKE2b-256 | 2c62321e60018a754c3248188145727f96f6ac43e7e74aef3d7b1dafd0aa97a8 |