Skip to contents

The tasks AI-assisted drafting knows

Source: R/ai_context.R
tfl_ai_tasks.Rd

What each task of the drafting functions (tfl_ai_prompt(), tfl_ai_parse(), tfl_ai_check(), tfl_ai_diff()) fills, which of tflspec's checks reads its answer, and what its items are matched by.

Usage

tfl_ai_tasks()

Value

A data frame: task, fills, checked_by, keyed_by, per (study, report or batch: one prompt a study, a report, or a few reports).

Details

  • toc: the study's table of contents, one row a report, read through tfl_read_toc() (one prompt for the study).

  • figure: one report's whole figure design (tfl_fig_design()), checked by tfl_check_fig_design().

  • table: the report files of the YAML spec (spec/outputs/<id>.yml) of a batch of 5 to 10 tables – each its ard: analyses, data:, codelists:, content: items (one call each: normalize_ard, table_plan, plan_*) and page: – over the study's defaults; read by tfl_read_output(), checked item by item against the schema (tfl_schema_check()) and the ADaM summary, and written as files by tfl_ai_unpack().

The ard task (the ARD definition alone) comes in a later version.

Examples

tfl_ai_tasks()
#>     task
#> 1    toc
#> 2 figure
#> 3  table
#>                                                                                                      fills
#> 1             TOC rows: output_id, type, title lines, population, footnote lines, section, datasets, label
#> 2                                                    one figure's whole design (data, stats, plot, layers)
#> 3 the report files (spec/outputs/<id>.yml) of a batch of tables: ard, data, codelists, content items, page
#>                                                                  checked_by
#> 1                                                            tfl_read_toc()
#> 2                                                    tfl_check_fig_design()
#> 3 tfl_read_output(), tfl_schema_check(), tfl_ard_code(), tfl_content_code()
#>                                  keyed_by    per
#> 1                               output_id  study
#> 2 section, then step / layer and position report
#> 3                               output_id  batch