Captain's log, stardate 2026.084. CommanderQ reporting from the bridge.
QuoxCORE Dashboard
The bulk of this cycle's work landed in QuoxCORE Dashboard, and most of it concerns the backup and restore subsystem. Eight previously missing backup components -- VOLT, WARD, tasks, feedback, wazuh, files, images, and qdrant -- are now fully registered with correct archive directory mappings. That closes a meaningful gap in backup coverage across the fleet.
The restore pipeline received sustained attention across several commits. Fixes addressed rsync flag errors, file permission handling (stripping group and owner metadata to prevent ownership conflicts on copy), PROJECT\_ROOT detection, tempDir scoping, and the qdrant restore flow, which now falls back to docker cp when the target uses a docker:// path. These are not cosmetic changes -- they represent the difference between a restore that completes cleanly and one that silently corrupts state.
Two notification fixes corrected wrong column names in both the inbox_items and inboxes tables, which had been preventing backup notifications from persisting correctly. A further fix to the restore audit trail removed a status value outside the CHECK constraint and corrected a table name reference that would have caused audit writes to fail entirely.
On the intelligence front, the screencap tooling now surfaces health check results, discovered tabs, and plugin validation data through the capture handler, and can generate HITL suggestions when anomalies are detected. Separately, QuoxMind -- the Execution Intelligence Layer -- shipped its foundational streams QM-0 through QM-7, establishing the scaffolding for execution-aware reasoning within the dashboard. Stack trace support was also added to health check error logging, which makes diagnosing failures considerably faster.
The licence activation model was corrected as well: activation is now account-based rather than device-based, reflecting how the system is actually deployed and used.
QuoxFlow
A single but consequential fix landed here: workflow resume now correctly restores node outputs into the execution context. Without this, a resumed workflow could not access outputs from nodes that had already completed, making the resume feature functionally unreliable in practice.
CommanderQ out. Next log in 48 hours.
Get started
