Prepare a study's AI table requests in one call

The steps before the first request, one after the other: tfl_ai_study_init() (only when the study has no study.yml yet – one that is there is kept, so the call can be run again), tfl_ai_adam_summary() (to ai/adam_summary.md), tfl_ai_read_shell() (to ai/toc_shell.md; its summary is printed, with what to pass when no TOC or a report's shell was found) and tfl_ai_write_requests(). Nothing is sent anywhere.

Usage

tfl_ai_prepare(
  dir,
  toc,
  adam = "adam",
  study_id = "STUDY-01",
  title = NULL,
  toc_sheet = NULL,
  sheets = NULL,
  size = 3L,
  language = NULL,
  workbook = c("text", "attach", "both"),
  instructions = c("each", "once"),
  max_chars = NULL,
  overwrite = FALSE
)

Arguments

dir

The study folder.

toc

The TOC / SHELL workbook (.xlsx or .xlsm).

adam

The ADaM folder, from dir (relative) or absolute.

study_id, title

The study's id and title, for a new study folder.

toc_sheet, sheets

As tfl_ai_read_shell(): the TOC sheet when the guess is wrong, the sheets to keep.

size, language, workbook, instructions, max_chars

As tfl_ai_write_requests(): the reports a request holds, the language of the assumptions, how the workbook reaches the chat ("text", "attach", "both"), the instructions in each request or once, a request's length at most.

overwrite

TRUE to start the study folder again (tfl_ai_study_init() over its study.yml and spec/). The requests are written again on every call (the answers are kept).

Value

The requests, as tfl_ai_write_requests() gives them.

See also

The AI table guide: system.file("ai", "tflspec-ai-table-guide.md", package = "tflspec").

Examples

ex <- system.file("ai", "example", package = "tflspec")
d <- file.path(tempdir(), "prepare-study")
dir.create(file.path(d, "adam"), recursive = TRUE, showWarnings = FALSE)
if (requireNamespace("haven", quietly = TRUE)) {
  for (x in c("adsl", "adae", "adrs", "adlb")) {
    df <- getExportedValue("tflspec", x)
    for (v in names(df)) if (is.factor(df[[v]])) df[[v]] <- as.character(df[[v]])
    haven::write_xpt(df, file.path(d, "adam", paste0(x, ".xpt")))
  }
  r <- tfl_ai_prepare(d, file.path(ex, "onco_toc_shell.xlsx"),
                      study_id = "XYZ-ONC-01", overwrite = TRUE)
  r[c("request", "outputs", "chars")]
}
#> <tfl_ai_shell: onco_toc_shell.xlsx>
#> 4 sheet(s): TOC (toc), T14.1.1 (shell), T14.2.1 (shell), Safety (shell)
#> 4 report(s) in the TOC, 4 with a shell found
#> text: 3932 characters; format(x) gives it
#> 4 requests (4 reports) in ai/requests; the longest 33295 characters.
#> Next: paste ai/requests/demographics-1.md into a new chat; save the whole answer as ai/answers/demographics-1.md, then run tfl_ai_collect_answers().
#>          request  outputs chars
#> 1 demographics-1 T-14-1-1 33295
#> 2           ae-1 T-14-3-1 33176
#> 3         labs-1 T-14-3-5 32997
#> 4     efficacy-1 T-14-2-1 33168