Resync Unverified Works

Resync Unverified Works

Scan all stored ORCID items and queue ingestion for every person who has works with a now-eligible work type so the ingestion pipeline re-classifies them as works instead of leaving them in the unverified review queue.

Script: scripts/db/resync-unverified-works.mjs npm alias: pnpm resync:unverified-works

Token is loaded automatically from .env at the repo root (CLOUDFLARE_API_TOKEN or CF_API_TOKEN).

---

When to use this

Run this after expanding the ORCID_WORK_TYPE_KEYWORDS list. Existing orcid_works rows that were stored before the change still carry their original classification. This script finds those rows and queues ingestion jobs for the affected ORCIDs so the pipeline re-processes them with the updated keyword list.

Newly-ingested works (after the keyword change is deployed) are handled automatically — no manual resync needed.

---

Commands

Dry-run (plan only, no queues touched)

Prints the affected ORCID count and a work-type breakdown without sending anything:

pnpm resync:unverified-works --dry-run

Live run (default: remote D1)

pnpm resync:unverified-works

This defaults to low-pressure ingestion mode (skip_existing=true, concurrency=1, enqueue-delay-ms=1000) to reduce worker invocation pressure.

Scope to a specific entity

pnpm resync:unverified-works --entity-id=cesam

Add more time to watch

pnpm resync:unverified-works --watch-timeout=300

By default, watch mode has no timeout and waits until all queued jobs reach a terminal state (complete or failed).

Use local D1 (for testing against local state)

pnpm resync:unverified-works --local

Control queue send concurrency (default: 1)

pnpm resync:unverified-works --concurrency=10

Control delay between queue chunks (default: 1000ms)

pnpm resync:unverified-works --enqueue-delay-ms=2000

Re-enrich already-known publications (higher load)

pnpm resync:unverified-works --no-skip-existing

---

All options

Flag	Default	Description
--dry-run	off	Print plan only; no messages sent
--remote	on	Use remote D1 databases
--local	off	Use local D1 databases
--entity-id=<id>	cesam	Entity ID written into each sync job message
--concurrency=<n>	1	Concurrent queue send requests
--enqueue-delay-ms=<n>	1000	Delay between queue chunks in milliseconds
--no-skip-existing	off	Re-enrich already-known publications (higher load)
--no-watch	off	Queue jobs only, skip live status polling
--watch-timeout	off	Optional timeout in seconds for watch mode

---

What happens after queueing

Each queued message triggers the ingestion pipeline for one ORCID. The worker fetches ORCID works and runs classifyOrcidWork with the updated keyword list. Items that now resolve to "work" are written to the works table via handleNonDoiOrcidWork and will no longer appear in the unverified review queue.

The script also polls ingestion_jobs live and prints counts for missing, queued, processing, complete, and failed jobs so you can see if processing started and whether it finished.