ht-migration-continuous-worker

Execute exactly one ACM HyperText paper migration work unit; do not merely report or plan. The memory base is imports/uploaded-archive-2026-09-14. Read MIGRATION-SCHEDULE.md, CAMPAIGN-CONTROL.json, campaign-ledger.json, inventory.json, reconciliation.json, and recent runs. Enforce maximum concurrency one with the campaign lock: if another worker is active and the lock is fresh, stop without mutation. Select exactly one eligible paper using the control-file priority, preferring localreadyunpublished, then pilot/OCR validation, structured sources, existing reconciliation, and remaining PDF papers. Complete the paper sequentially. Prefer supplied XML/HTML; otherwise use coordinate-aware PDF extraction and OCR only when required. Fail closed on source identity, attribution, metadata, complete body, reading order, figures and captions, tables, equations, references, and appendices. If uncertain, record a specific heldforreview or manualremediation state with evidence and do not publish. If QA passes, publish assets, dry-run the Seed write, write below the existing year parent in the ACM HyperText account using signer z6MknjjfAYZdrU3Ksw2imMhq5JLSE8Cuc7Cev9fcr4oxJaGC, then read and verify the exact resulting version. Serialize every ledger and Seed mutation. Update the ledger only after evidence exists. Append one timestamped report under runs, regenerate LATEST-PROGRESS.md, update controls, and clear the lock even after handled failure. Do not fan out paper conversion. Keep the three unsupplied omissions outside the 1,665 supplied-paper denominator. End after this single record so the next interval continues the queue.

Do you like what you are reading? Subscribe to receive updates.

Unsubscribe anytime