Prompt input for the LLM — prompt

A helper function to create the prompt input.

Usage

prompt_user(
  content,
  images = NULL,
  tool_calls = NULL,
  .content_envir = rlang::caller_env(),
  .collapse_regex = "[*]$"
)

prompt_assistant(
  content,
  images = NULL,
  tool_calls = NULL,
  .content_envir = rlang::caller_env(),
  .collapse_regex = "[*]$"
)

prompt_system(
  content,
  images = NULL,
  tool_calls = NULL,
  .content_envir = rlang::caller_env(),
  .collapse_regex = "[*]$"
)

Arguments

content: A single text with any content in "" evaluated as an R expression. If the bracket contains "" at the end (or regular expression match as specified in .collapse_regex) then it collapses into a comma separated list in a square bracket with each element single quoted. E.g. "1:3" expands to '1', '2', '3'.
images: A list of path or url to the images (or base64 encoded images).
tool_calls: Currently doesn't work as intended.
.content_envir: The environment used to evaluate content.
.collapse_regex: The regular expression match to collapse the content.
role: The role should be "user", "assistant" or "system".

Examples

prompt_user("What should I eat today? Pick one of {c('Ramen', 'Pizza', 'Burger')*}.")
#> $role
#> [1] "user"
#> 
#> $content
#> What should I eat today? Pick one of ['Ramen', 'Pizza', 'Burger'].
#> 
#> attr(,"class")
#> [1] "prompt" "list"