The Challenge
Clearing goods meant filing into a different national system per corridor — ICEGATE in India, ACE in the US, CDS in the EU — each with its own format and quirks. Filing was manual, error-prone, and when a submission failed there was no systematic recovery; shipments simply stalled until someone noticed.
The Solution
We built a connector abstraction on the Baalvion Operating System: one base interface with a concrete connector per national gateway. Submissions run through an asynchronous, queue-backed pipeline that retries transient failures and routes persistent ones to a dead-letter queue for review — so a single gateway hiccup never silently strands a shipment.
- Connector per gateway (ICEGATE, ACE, CDS) behind one interface.
- Async pipeline so filings don't block the order flow.
- Automatic retry with a dead-letter path for persistent failures.
- Normalised responses so downstream systems see one shape.
Architecture
Each connector implements a shared base contract, translating the platform's canonical submission into the gateway's native format and normalising the response back. A message queue decouples submission from clearing; a relay processes the queue with retry and failover, and persistent failures land in a dead-letter queue with full context. Clearing is triggered by shipment events, the same event-driven pattern behind our logistics control tower.
Technology Stack
A connector architecture, a message queue with retry and dead-letter handling, response normalisation, and event triggers — built through our automation and cloud solutions practices for the logistics sector.
Results
Manual filings dropped by roughly 85%, failed submissions now recover automatically instead of stranding shipments, and the same platform clears goods across three very different national systems through one consistent interface.
Lessons Learned
Abstracting the gateway and normalising responses meant adding a new country was a connector, not a rewrite. Every external call needed an explicit retry and dead-letter path — assuming success is how shipments get stuck. And treating failure recovery as a first-class feature, not an edge case, was what made the system trustworthy in production.