A stockist's billing PC wasn't designed to be a cloud data source. It runs local software, on local storage, on a connection that might be solid 4G one hour and nothing the next. Surfacing near-real-time prices and stock from that environment is harder than it sounds.
Offline-first is the normal case, not the edge case
We assume the connection will drop. The connector captures changes locally and syncs when it can, rather than depending on a constant live link. If the godown's internet goes out for an afternoon, nothing breaks — the data catches up when connectivity returns, in order, without gaps.
Idempotent ingestion
Patchy networks mean the same update can arrive twice, or a sync can be interrupted halfway. Every ingestion is idempotent: replaying the same data produces the same result, never a double-count or a corrupted price. The pipeline is built to be retried safely, because in Tier-2 it will be.
Knowing what's stale
Not all data is equally fresh, and pretending otherwise is how wrong prices reach consumers. The system tracks how recently each stockist's feed updated, so stale entries can be aged out or flagged rather than shown as if they were live. Honest staleness beats confident wrongness.
Built for the network we actually have
It would be easy to build for a clean, always-on connection and demo it in a Bangalore office. It would also be useless in Madurai. Designing for intermittent connectivity from day one is part of why VeBy works where the model has to work.