Skip to contents

Assign names + descriptions to every theme and subtheme in one AI call

Source: R/theme_algorithm_v2.R
ai_label_theme_set.Rd

Single AI call after convergence. The AI sees the FULL tree (themes -> subthemes -> codes) with every member code visible, and assigns researcher-facing names + descriptions to every node. Cross-theme name distinctness is explicitly enforced via the prompt; the AI is told that if two themes both deserve the same name, the underlying structure is wrong.

Usage

ai_label_theme_set(
  skeleton,
  codes,
  provider,
  research_focus = "",
  concept_str = "",
  reflexivity_block = "",
  audit_log = NULL,
  response_cache = NULL,
  methodology_override = NULL,
  walk_state = NULL
)

Details

The schema (.theme_labeling_schema()) requires the AI to return the same number of themes as the skeleton has, and the same number of subthemes per theme. The orchestrator binds names positionally via theme_index / subtheme_index.

On AI failure or malformed response, falls back to derived names (top member code per theme) so the run still produces an output – but logs a warning so the operator knows labeling didn't succeed.