How to integrate GPT-4 model hosted on Azure with the gptstudio package?

Question

I am looking to integrate the OpenAI GPT-4 model into my application. Here are the details I have:

• Endpoint: https://xxxxxxxxxxxxxxx.openai.azure.com/
• Location/Region: yyyyyyyyyyyyyyyyyyyyyyyy
• Key: *******************
• Deployment Name: gpt-4o
• Model Name: gpt-4o
• Model Version: 2024-02-01

I edited the .Renviron file accordingly:

AZURE_OPENAI_TASK="completions"
AZURE_OPENAI_ENDPOINT="https://xxxxxxxxxxxxxxx.openai.azure.com/"
AZURE_OPENAI_DEPLOYMENT_NAME="gpt-4o"
AZURE_OPENAI_KEY="*******************"
AZURE_OPENAI_API_VERSION="2024-02-01"
AZURE_OPENAI_USE_TOKEN=FALSE

I want to integrate it with gptstudio. Could someone guide me on authenticating and making API requests to this endpoint?

I tried:

library(gptstudio)
chat(service = "azure_openai", prompt = "hello", model = "gpt-4)
#> $messages
#> $messages[[1]]
#> $messages[[1]]$role
#> [1] "system"
#> 
#> $messages[[1]]$content
#> As a chat bot assisting an R programmer working in the RStudio IDE it is important to tailor responses to their skill level and preferred coding style. They consider themselves to be a beginner R programmer. Provide answers with their skill level in mind.  
#> 
#> 
#> $messages[[2]]
#> $messages[[2]]$role
#> [1] "user"
#> 
#> $messages[[2]]$content
#> [1] "hello"
#> Error in `query_api_azure_openai()`:
#> ✖ Azure OpenAI API request failed. Error 400 - Bad Request
#> ℹ Visit the Azure OpenAi Error code guidance
#>   (<https://help.openai.com/en/articles/6891839-api-error-code-guidance>) for
#>   more details
#> ℹ You can also visit the API documentation
#>   (<https://platform.openai.com/docs/guides/error-codes/api-errors>)

Answer 1

I would recommend Hadley Wickham's new package elmer. It is very simple. You can save your Azure API configuration in the .Renviron by calling usethis::edit_r_environ() where you can setup: AZURE_OPENAI_ENDPOINT, AZURE_OPENAI_DEPLOYMENT_NAME, AZURE_OPENAI_API_VERSION and AZURE_OPENAI_KEY.

Then you can use them to configure your chat object:

library(elmer)
chat <- chat_azure(endpoint = Sys.getenv("AZURE_OPENAI_ENDPOINT"),
                    deployment_id = Sys.getenv("AZURE_OPENAI_DEPLOYMENT_NAME"),
                    api_version = Sys.getenv("AZURE_OPENAI_API_VERSION"),
                    api_key = Sys.getenv("AZURE_OPENAI_KEY"))

One example:

chat$chat("What is R?")
#> R is a programming language and free software environment used primarily for 
#> statistical computing and data analysis. It was created by statisticians Ross 
#> Ihaka and Robert Gentleman in 1993. The R environment provides a wide variety 
#> of statistical and graphical techniques, and it is highly extensible. This 
#> makes it a powerful tool for data manipulation, calculation, and graphical 
#> display.
#> 
#> R features include:
#> - An effective data handling and storage facility
#> - A suite of operators for calculations on arrays, in particular matrices
#> - A large, coherent, and integrated collection of intermediate tools for data 
#> analysis
#> - Graphical facilities for data analysis and display, either on-screen or on 
#> hardcopy
#> - A well-developed, simple, and effective programming language which includes 
#> conditionals, loops, user-defined recursive functions, and input and output 
#> facilities.
#> 
#> R has a large and active community, with many packages available to extend its 
#> capabilities. These packages are often contributed by users and cover 
#> specialized areas of statistics, machine learning, bioinformatics, and more. 
#> The Comprehensive R Archive Network (CRAN) is the main repository for these 
#> packages. Because of its power and flexibility, R is widely used in academia, 
#> research, and industry.

Answer 2

Below is the sample code to authenticate an Azure Openai GPT-4o model with a key in R using httr and jsonlite packages.

Code :

library(httr)
library(jsonlite)

azure_endpoint <- Sys.getenv("AZURE_OPENAI_ENDPOINT")
api_key <- Sys.getenv("AZURE_OPENAI_KEY")
api_version <- Sys.getenv("AZURE_OPENAI_API_VERSION")

api_url <- paste0(azure_endpoint, paste0("openai/deployments/",Sys.getenv("AZURE_OPENAI_DEPLOYMENT_NAME"),"/chat/completions?api-version="), api_version)

messages <- list(
  list(role = "system", content = "You are a helpful assistant."),
  list(role = "user", content = "Does Azure Open AI supports gpt-4 model?")
)

request_body <- list(messages = messages)
request_body_json <- toJSON(request_body, auto_unbox = TRUE)

response <- httr::POST(api_url,
                       httr::add_headers(`Content-Type` = "application/json",
                                         `api-key` = api_key),
                       body = request_body_json)

if (http_error(response)) {
  cat("HTTP error:", response$status_code, "\n")
  print(content(response))
} else {
  response_content <- jsonlite::fromJSON(rawToChar(response$content))
  print(response_content)
  
  if (length(response_content$choices) > 0) {
    cat("Response:", response_content$choices[[1]]$message$content, "\n")
  }
}

Output :

The following code ran successfully in RStudio as below,