OpenAI4R

OpenAI is an artificial intelligence research laboratory founded by a group of technology leaders in 2015, with the aim of advancing digital intelligence technology for the benefit of humanity as a whole. OpenAI conducts research in various areas of artificial intelligence, including natural language processing, robotics, and reinforcement learning. It has developed several powerful AI models, such as GPT (Generative Pre-trained Transformer) and DALL-E, which can generate human-like language and images from textual descriptions.

As a community-maintained R package, OpenAI4R is aimed to enable a clear and intuitive calling of OpenAI's powerful AI models within the R language.

Install

Execute in R:

devtools::install_github('yimingsun12138/OpenAI4R')

To install the under-development version, execute:

devtools::install_github('yimingsun12138/OpenAI4R',ref = 'dev')

Notice

OpenAI4R now provides ChatGPT and completions API only. More APIs will be added later.

Usage

Load package and set up authentication:

library(OpenAI4R)

# Input your own OpenAI API key and organization.
Auth_OpenAI(key = readLines('/content/script/openai_API_key'))

A global variable named OpenAI_API_key will appear in the current environment.

Chat model

For chat model, initialize a chat session first. You can post your first command here as 'system' to set the global settings, modify the parameters of the chat model, or import an existing chat session to continue the conversation.

chat <- Init_chat_session()

Then, you can use this initialized function to have a continuous conversation. Here is an example:

> chat('Play a game with me. generate a number between 1 to 10 and i will guess it. If I am wrong, tell me how to adjust')
ChatGPT:
Sure, I can generate a random number between 1 to 10. Alright, I've got one in mind. Go ahead and make your first guess!

> chat('5')
ChatGPT:
Sorry, your guess was incorrect. The number I generated was less than 5. Please guess again, but this time try a lower number!

> chat('3')
ChatGPT:
Great job! You guessed it correctly. The number I was thinking of was indeed 3. Congratulations!

> chat('That is fun')
ChatGPT:
I'm glad you enjoyed the game! Feel free to play again or try out some other games. Let me know if you need any help.

You can also modify the model parameters during the conversation and export the conversation history. Refer to ?chat_func_char for more details.

Completion model

The completion model (endpoint) can be used for a wide variety of tasks. It provides a simple but powerful interface to some of OpenAI models. You input some text as a prompt, and the model will generate a text completion that attempts to match whatever context or pattern you gave it. Some examples are provided below.

> completion('Write a tagline for an ice cream shop.\n')
completion model:
Write a tagline for an ice cream shop.

"Cool down with a scoop of happiness!"

If you do not want to echo your prompt:

> completion('Write a tagline for an ice cream shop.',echo = FALSE)
completion model:
"Cool down with a scoop of fun!"

Create a text classifier:

> completion("Decide whether a Tweet's sentiment is positive, neutral, or negative.
+ 
+ Tweet: I loved the new Batman movie!
+ Sentiment:")
completion model:
Decide whether a Tweet's sentiment is positive, neutral, or negative.

Tweet: I loved the new Batman movie!
Sentiment: Positive

You can even insert the completion into your content, with the suffix parameter:

> completion(prompt_content = "I went to college at Boston University. After getting my degree, I decided to make a change",suffix = "Now, I can’t get enough of the Pacific Ocean!")
completion model:
I went to college at Boston University. After getting my degree, I decided to make a change and moved to California. I was looking for something new and exciting. The change of scenery and the warm weather was perfect. Now, I can’t get enough of the Pacific Ocean!

That is so cool!

Limitation

• stream, logit_bias and user parameter are not supported in Chat model and Completion model currently.

To do list

• Add more OpenAI APIs in this package.

• More detailed document and guidance.

• Use a universal function to initialize all APIs, when more APIs are added in the future.

• Fix the limitation.