Independent synthetic work-progress and resource-reference lab / 독립 합성 작업·참조 실험실
Python standard library. Run python test_work.py and python run_examples.py after extraction.
Original code, no client data, real accounts, network, file upload, object bytes storage or deletion.
Jobs persist cursor, generation and worker metadata in a local SQLite database. Reopening retains progress; a newly claimed generation fences a stale worker. Chunk tokens detect repeated count updates. Running workers are not reclaimed automatically: no lease expiry, process launcher, heartbeat, authorization or distributed scheduler. Job progress alone does not establish that business rows committed; production chunk writes and progress must share an appropriate transaction.
Resource registration computes a scoped content digest and byte count, NOT persistent object bytes. Same-company digest reuse returns the existing metadata identity; other companies remain separate. Actual document ownership/authentication must be validated by a server before calling link. Review flag required before linking. Current, history and backup references all protect metadata from the manual cleanup candidate list. No resource deletion API exists. Preview cannot prove that unknown external references do not exist.
All company/worker/document identifiers are trusted synthetic context, not authenticated client input. Rejects an unrelated existing SQLite file before mutation. No compatibility with real ERP schemas.
Recorded JSON is actual local example output; a browser record viewer does not run Python/DB. No physical-device or production-security audit.
수정·재배포·상업적 사용을 허용하며 무보증입니다. You may use, modify and redistribute these original synthetic files, including commercially, as-is without warranty.
