deep-issue-id

Translation issue identification in ULT and UST text is handled by orchestrating multiple agents to compare text, detect structural and lexical problems, and apply domain-specific rules across multi-wave workflows.

You can analyze specific verse ranges by providing optional verse limits alongside the book and chapter, allowing the multi-agent workflow to scope detection and generate per-verse range TSV outputs.

Multi-agent issue detection works through a deterministic, multi-stage workflow that orchestrates sub-agents across distinct waves for structure analysis, challenger rulings, and merge adjudication to surface actionable translation issues.

Translation issues are output as consolidated per-chapter TSV files, and per-verse range TSVs when specified, making them suitable for downstream review and publishing workflows.

Editor notes are optional context you can provide to inform analysts and align outputs with the human text, but the workflow can run using the human-authored ULT and UST text directly from the repo master.

Automated issue identification is limited to detecting structural and lexical problems through domain-specific rules and sub-agent adjudication, meaning it surfaces actionable flags for review rather than automatically correcting translation text.