Bundle a run's transparency artifacts into a single report — bundle_transparency

Generates a self-contained HTML methodological-transparency report for a pakhom run, plus a machine-readable JSON companion. Per AC4 (methodology stamped on every output), the report is mode-stamped at the top; per the transparency spec (anti-Jowsey compliance), it maps every pipeline step to a Lincoln & Guba (1985) credibility / dependability / confirmability / transferability checkpoint citing the exact decisions logged.

Usage

bundle_transparency_report(run_dir, output_path = NULL)

Arguments

run_dir: Path to the run output directory (the directory containing run_metadata.json + ai_decisions.jsonl etc.).
output_path: Optional path for the HTML output. Defaults to file.path(run_dir, "transparency_report.html"). The JSON companion is written alongside (same basename, .json extension).

Value

Invisible list with html_path, json_path, and the parsed report_data (the machine-readable contents).

Details

The bundler reads ONLY from disk artifacts produced by a completed (or in-progress) pakhom run – it never re-executes the pipeline, never calls an AI provider, and is safe to invoke any number of times. Missing artifacts degrade gracefully (the corresponding section renders an "unavailable" notice rather than crashing).

References

Lincoln, Y. S. & Guba, E. G. (1985). Naturalistic inquiry. Sage Publications. Olmos-Vega, F. M. et al. (2023). A practical guide to reflexivity in qualitative research: AMEE Guide No. 149. Jowsey et al. (2025). PLOS One doi:10.1371/journal.pone.0330217.