# New Part Document 2026-09-17 04:43 Asia/Shanghai. `Part:desktop:Part_NewDoc` passed its first formal acceptance: 15 scenarios in 9.235 seconds. Evidence: `business-evidence/f53fc82beb195554c975004d-20260916T204328632190.json`. The existing document creation implementation was retained. Fixed its mounted command response handling: the JSON editor's active-document update was inside the Slice-only tree parser branch, so NewDoc never reached it. NewDoc now updates that context directly from its validated empty-document response. Native and browser volume checks use a finite relative tolerance for OCCT integration; the previous strict 210 comparison rejected 209.99999999999997. Cases prove distinct empty native documents, active document changes, selection clearing, retained existing documents, invalid-request rejection, subsequent geometry edits and undo/redo, FCStd save/close/reopen, mounted desktop/mobile controls and OPFS restoration across page reload. The case retains actual native volume output. It does not certify every document-management command. One formal run correctly refused acceptance because the concurrent background implementation changed Part inputs during execution. The existing systemd service was then briefly frozen for a stable acceptance window, and thawed immediately after success. Its current task was not terminated or restarted. Future evidence must still reject changed inputs; no fingerprint rule was relaxed. Related blockers fixed in this continuation: - The bounded reader accepts exact current-module native task, frontend and business-case paths before coverage is declared. Existing byte/window limits, duplicate suppression, module separation and outside-root rejection remain. Seven focused source-reader tests passed, including those boundaries. - The unfinished Part import panel omitted required native-view selection props. It now maintains actual selected-object state. This fixes compilation only; no Import business acceptance is claimed or rerun. TypeScript and the production frontend build passed. Local sources and artifacts were sufficient; no downloads or old-item regression sweep were needed. Coverage at this snapshot is 189/1508, leaving 1319 requirements without executable cases. Coverage is distinct from current accepted proof. Full scope remains incomplete.