Compile a program — compile • dsprrr

Optimize a dsprrr program with a teleprompter. This is the single user-facing compilation entry point and is ordered for the native pipe: program |> compile(teleprompter, trainset).

Usage

compile(program, teleprompter, ...)

Arguments

program: A dsprrr module or compositional program to optimize.
teleprompter: A Teleprompter defining the optimization strategy.
...: Additional arguments. The first is normally trainset, a data frame. Optimizers may also accept valset, .llm, and .trace_context.

Value

An optimized program.

Examples

if (FALSE) { # \dontrun{
classifier <- module(signature("text -> sentiment"))
trainset <- data.frame(
  text = c("I love it!", "Terrible experience"),
  sentiment = c("positive", "negative")
)

optimized <- classifier |>
  compile(LabeledFewShot(k = 2L), trainset)
} # }