test
test pg-cluster-postgresql-postgresql.data.svc:5432/recognition_engine_test
Checking...

Environment

Loading...
Auto-Seed: checking...

Infrastructure

Background Jobs

JobStatusStartedDurationLines
Loading...

Recent pipeline jobs

NameStatusActiveOKFailDurationOwnerImageAge
Loading...

Queues

Loading...
Quick Actions

Import Scan

When reprocess is enabled, ML models are auto-imported from the selected source and any missing weight files are copied to MinIO.

Reprocess Scan

Resets scan status and triggers the pipeline for a scan already in this environment. Missing model weights are pulled from the selected source on demand.

Scans with recent activity

ScanStore / ShelfAgeDurationLast eventStage progress
Loading...

Pods

NameStatusReadyRestartsAgeImageNode
Loading...

Lower-frequency import flows. The two most common actions (Import Scan and Reprocess Scan) live on Home.

Import Project

Core tables (accounts, users, categories, packages, labels) are always imported. Select additional sections:

Import Planogram

Imports planogram + planogram_facing_position rows + referenced products (objects). Banner / store / shelf chain is auto-resolved via the planogram's FKs. Use this to seed Studio / facing / extapi work with realistic prod POGs without going through the full scan pipeline.

Import Users

Import ML Models

Imports deployed detector, SC, OCR, and router models with weights from source S3 to MinIO.

System Object Sync

What this is: Detector-emitted marker objects (rows of the object table where object_type <> 'CPG' — e.g. Marker, Rotated, Box End, Shelf Tag, CPG Unknown / Wrong / Not Found). They are referenced by every recognition output through object_volume → resolved_volume_inference → observation.

Why you might need to run it: if any of these rows is missing locally, resolver-runtime crashes with a ForeignKeyViolation. Because that is an IntegrityError (not a ValueError) the worker framework's detect_scan_failures path does not mark the scan FAILED — the SQS message just gets retried forever and the scan stays IN_PROGRESS. Symptom: a scan stuck on the resolver stage with repeating STARTED → COMPLETED pairs but no RUNTIME_PATH_COMPLETED.

What it does: reads all system objects from the selected source, compares against the local object table, and inserts only the missing ones (ON CONFLICT (id) DO NOTHING). Idempotent — safe to re-run any time. Total volume is tiny (≈7 rows today), so this is fast and cheap.

Click Check to load the diff.

Kubernetes resources for this namespace. Pods moved to Home where they matter most.

Deployments

NameReadyImageAge
Loading...

ScaledJobs (Pipeline Workers)

NameMinMaxPausedQueueImageAge
Loading...

Rare / destructive operations and pipeline internals. Most developers never need these.

ArgoCD App Overrides

Loading apps...
Loading...

Reinitialize Database

Drops all tables in this environment's database, recreates schema + extensions, runs DB migrations, and applies baseline seed data (accounts, system users, reference catalogs). Then copies all pensa_internal users and their permissions from production so every Pensa developer can log in immediately (idempotent — safe to re-run). Use when the database is corrupted or you want a clean slate.

Export / Import Database

Export: downloads a pg_dump SQL snapshot of the current database — use as seed data or share with colleagues.
Import: upload a previously exported .sql file to restore a database state.