openapi (OpenAI API)

openapi is an R interface used to connect with the OpenAI API. It enables R users to call all API methods in the R environment and provides R6Class for various API returns.

With R6Class, it’s easy to extract and process the return information, and also beneficial for other R method development.

openapi also offers ChatGPT application in the R environment, which can be run in the console, Rstudio viewer, browser, or background, to assist R users in improving their work efficiency.

With openapi, various assistant can be created to help R users in different tasks such as code writing.

Additionally, these applications are created as Rstudio add-ins, making it convenient for R users to upload and handle their own code in ChatGPT.

Installation

if (!require("devtools", quietly = TRUE)) {
  install.packages("devtools")
}
devtools::install_github("zhanghao-njmu/openapi")

Authentication

To use openapi, the api_base and api_key must be provided.

The api_key needs to be generated through one’s own OpenAI account. The default api_base is https://api.openai.com, but it can also be set to other OpenAI API reverse proxy addresses.

With api_setup function, one can manually set the api_base and api_key, and then all functions in openapi do not need to specify these two parameters.

One can also modify .Rprofile manually, for example, by setting options(openapi_api_key = api_key) to specify api_key permanently.

library(openapi)

## Official OpenAI API
api_base <- "https://api.openai.com/v1"
api_key <- "sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
api_setup(api_base = api_base, api_key = api_key)

## Azure OpenAI API
api_base <- "https://xxxxxx.openai.azure.com"
api_key <- "xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
api_setup(
  api_base = api_base, api_key = api_key,
  api_type = "azure", azure_deployment = "openai/deployments/[your_deployment_name]", api_version = "2023-03-15-preview"
)

ChatGPT

Interact with chatGPT on the console

q <- ChatGPT$new()
q$chat("What is Bioinformatics?")
q$chat("Application?")
q$regenerate()
q$first()
q$forward()
q$forward()
q$last()
q$messages
q$latest_response

Interact with chatGPT within the RStudio’s viewer pane

Start in the current session

ChatGPT_gadget()
ChatGPT_gadget(viewer = shiny::dialogViewer("chatgpt", width = 800, height = 1000))
ChatGPT_gadget(viewer = shiny::browserViewer())

Start as background job

ChatGPT_job()

R code assistant

code <- "
truncate_text <- function(text, max_length = 50) {
  truncated_text <- sapply(text, function(x) {
    if (nchar(x) > max_length) {
      paste0(substr(x, 1, max_length - 3), \"...(\", nchar(x) - max_length, \" characters omitted)\")
    } else {
      x
    }
  })
  return(truncat_text)
}
"
edited <- code_check(code)
edited$extract("difference")
edited$extract("output")
edited$response

Othe functions

edited <- code_document(code)
edited <- code_comment(code)
edited <- code_explain(code)
edited <- code_improve(code)
edited <- code_refactor(code)
edited <- code_create_test(code)

Rstudio Addins

• ChatGPT

• ChatGPT (background job)

• Quote selection

• Document selection

• Check selection

• Improve selection

• Comment selection

• Refactor selection

• Explain selection

• Create test units for selection

  Prompts

prompts_en <- fetch_prompts(language = "en")
prompts_zh <- fetch_prompts(language = "zh")
prompts_new <- generate_prompts("Act as an R Package Development Assistant")

OpenAI API

Models

models <- list_models()
models$print()
models$extract(field = "id", filter = list(owned_by = "openai-internal"))

gpt <- retrieve_model(model = "gpt-3.5-turbo")
gpt$print()
gpt$extract("permission")

Text completion

comp <- create_completion(prompt = "Who won the world series in 2020?")
comp$print()
comp$extract(field = "choices")
edit$extract(field = "model")

Chat completion

messages <- list(
  list(
    "role" = "system",
    "content" = "You are a helpful assistant."
  ),
  list(
    "role" = "user",
    "content" = "Who won the world series in 2020?"
  ),
  list(
    "role" = "assistant",
    "content" = "The Los Angeles Dodgers won the World Series in 2020."
  ),
  list(
    "role" = "user",
    "content" = "Where was it played?"
  )
)
chat <- create_chat_completion(messages = messages, stream = TRUE)
chat$print()
chat$extract(field = "choices")
chat$extract(field = "model")

Edit

edit <- create_edit(input = "What day of the wek is it?", instruction = "Fix the spelling mistakes")
edit$print()
edit$extract(field = "choices")
edit$extract(field = "model")

Embedding

emb <- create_embeddings(input = "The food was delicious and the waiter...")
emb$print()
emb$extract("embedding")

Images

image <- create_image(prompt = "A magpie standing on the green grassland", n = 1, size = "512x512")
fetch_image(image$extract("url"), destfile = "magpie.png")
mask_image(image = "magpie.png", height_range = c(1, 100), destfile = "magpie_edit.png")
mask_image(image = "magpie.png", G_range = c(0.6, 1), destfile = "magpie_edit.png")

image_edit <- create_image_edit(image = "magpie_edit.png", prompt = "A magpie standing on the snowy ground")
fetch_image(image_edit$extract("url"))

image_variation <- create_image_variation(image = "magpie.png")
fetch_image(image_variation$extract("url"))

More applications are waiting to be developed …