Three EU column presets added to tabular review (sub-processors, transfer mechanism, AI Act role)
counselos adds three column presets to `columnPresets.ts` that fire on EU/GDPR-flavored column headers and produce tightly constrained output: structured bullet lists for sub-processors, a closed enum for transfer mechanisms, and a defined actor taxonomy for AI Act roles. One file, 18 lines, no schema changes.
The change is purely additive - three new entries prepended to the PROMPT_PRESETS array in frontend/src/app/components/tabular/columnPresets.ts. Each preset matches by column name regex and then constrains the LLM's answer format.
The sub-processor preset matches /\bsub[- ]?processors?\b/i and enforces a strict per-bullet format: name - service - country. If a DPA allows sub-processors without listing them, the prompt instructs the model to say so explicitly rather than leave the cell empty.
The transfer mechanism preset targets SCC/IDTA/international-transfer header names and pins the answer to a closed enum: "EU SCCs (module [N])", "UK IDTA", "UK Addendum to EU SCCs", "Adequacy decision ([country])", "BCRs", "Article 49 derogation ([type])", or "No transfer outside EEA/UK". If a module number is unspecified, the prompt says to note that rather than guess.
The AI Act role preset matches GPAI/provider-deployer header patterns and enumerates the six actor roles from Regulation 2024/1689, with appended " - GPAI" or " - systemic risk" suffixes where applicable.
The ColumnPreset interface used here has four fields: name, matches, format, and prompt. If your fork diverged from that shape, verify before lifting.
Spotted something wrong? Or know the PR text has fresher detail than the writeup above?