Make a single GPT request — gpt

Form a request to the OpenAI API, execute the request, and format the response as a tibble.

Usage

gpt_req(
  messages,
  model = "gpt-3.5-turbo",
  temperature = 0,
  max_tokens = NULL,
  logit_bias = NULL,
  api_key = get_api_key()
)

Arguments

messages: A dataframe with "role" and "content" columns.
model: The GPT model to use, "gpt-3.5-turbo" is default
temperature: The temperature parameter, default 0
max_tokens: The max_tokens parameter, default 1
logit_bias: The logit_bias parameter.
api_key: The OpenAI API key. Will be retrieved automatically if set with set_api_key() or if GPTSALDANA_KEY is set in .Renviron

Value

A tibble with one row containing gpt_response, gpt_model, gpt_finish_reason, gpt_prompt_tokens, gpt_completion_tokens

Examples

gpt_req(messages = data.frame(role = c("system","user"),
                              content = c("Your job is to answer with the word class of the word the user submits.","quickly")
                              )
       )
#> # A tibble: 1 × 10
#>   gpt_response gpt_request  gpt_model        gpt_finish_reason gpt_prompt_tokens
#>   <chr>        <named list> <chr>            <chr>                         <int>
#> 1 Adverb       <df [2 × 2]> gpt-3.5-turbo-0… stop                             29
#> # ℹ 5 more variables: gpt_completion_tokens <int>, gpt_total_tokens <int>,
#> #   gpt_id <chr>, gpt_created <int>, gpt_system_fingerprint <chr>