Qubere Shipment Management
From creation to filing, the shipment detail page is the single system of record — canonical facts, containers/packages/line items, exceptions, PGA holds, screening, documents, tracking, and the full agent execution history, all reading and writing the same rows the rest of the app relies on.
The problem
What an ops manager says about managing shipments today:
"I have five tools open just to know what's still missing on one shipment."
No single record
"Nobody can tell me why a shipment is on hold without pinging three people."
No visible exception or audit history
"I don't know if the AI actually ran on this shipment or if someone just fixed it manually."
No agent execution visibility
How Qubere addresses it
The shipment workbench
Global search across shipment number and importer name, column filters for status/client/owner, a live countdown on ISF and Entry Filing deadlines, and KPI tiles computed from the same filtered rows — Total, In Progress, Ready to File, Exception Hold.
Canonical facts & cargo detail
The canonical facts panel is sourced from the same CanonicalShipmentService state the filing agents read. Containers, packages, and line items render as real, editable tables built from the shipment's actual extraction and entry data — the same rows the classification and duty-engine agents work from.
| Line item | HTS | Qty | Value |
|---|---|---|---|
| Cotton knit top | 6109.10.0012 | 4,800 | $18,220 |
| Steel fasteners | 7318.15.2000 | 62,000 | $9,140 |
Pipeline progress & agent execution
A live pipeline tracker polls run status every 3 seconds with a retry action. The agent execution timeline is built from real AgentExecutionRecord/PipelineJob rows — Document Intake, Document Intelligence, Product Intelligence, HTS Classification, Origin Rules, Valuation & Assists, Compliance Audit, and Filing Readiness.
Exceptions & PGA holds
The exceptions drawer reads the same ExceptionItem rows used across the app, layered with computed missing-document and reconciliation-conflict cards. PGA holds are kept strictly separate from screening warnings — only an actually-recorded agency hold shows here.
Party screening, embargo checks & documents
The party-screening panel reads the same restricted-party disposition model used across the compliance module, gated by compliance.restrictedParty.read. The shipment's Parties panel shows each role's Party Master link and latest screening status, read-only — review stays in Compliance Workbench. Embargo and PGA checks no longer have a card on the shipment page; they live on the Customs Filing page, reading the same shared embargo model. The document workspace is the same document components used in Document Management, scoped here.
| Document | Type | Status |
|---|---|---|
| INV-2026-4471.pdf | Commercial Invoice | accepted |
| PL-2026-4471.pdf | Packing List | accepted |
Tracking & audit trail
The tracking panel is a real webhook/event-driven projection — connection state shows as Connected, Waiting, Stale, or Not Configured depending on whether a carrier feed is actually wired up for that leg, never faked as live when it isn't. The audit trail ties field approvals, document ingestion, agent runs, and exception resolutions into one chronological record.
Objection handling
| "Is the tracking panel live carrier tracking for every shipment?" | It's a real webhook/event-driven projection, but connection state is shown honestly — Connected, Waiting, Stale, or Not Configured — depending on whether a carrier feed is actually wired for that leg. It never displays a fake "live" state. |
| "Does the agent timeline mean Customs Filing runs automatically too?" | No. Customs Filing is deliberately excluded from every auto-triggered pipeline path because it simulates an actual CBP transmission — it's always a manual, explicit action, never automatic. |
| "Can I bulk-update shipments from the workbench?" | Today the workbench supports inline broker reassignment one row at a time; there's no multi-select bulk-action toolbar yet. |
The close
No spreadsheets, no separate ops log, no guessing why something's on hold — because every panel on the page reads the same shipment the rest of the app relies on.
Next step: open a shipment with an open exception, and walk the audit trail back to the exact document or agent run that caused it.