Inventory Application.

Inventory counting uses a shared product catalog for configured companies. Receiving goods remains assigned to the selected company and warehouse. Separating these activities reduces the risk of counting or recording goods in the wrong context.

Before counting, the application retrieves a consistent catalog of products, EANs, and stock levels. Scanning then uses the local index. It does not require writing to Optima or an active internet connection for every code. An unknown stock level remains unknown–it is not replaced with zero.

Each scan creates an operation in the local SQLite database. Confirmation appears for the operator after the transaction is saved. Quantity can follow from a code multiplier, such as a carton. Undo compensates for a specific operation and preserves the history instead of erasing the trace of an error.

A code that cannot be matched unambiguously does not increase the quantity. The application shows the problem to the operator. An optional online suggestion is labeled as a suggestion, requires checking the packaging, and does not write anything to Optima.

Completion leads through a summary and identification of the person performing the work to a complete set of exports. A multipage PDF and data files preserve the product, code, source, counted quantity, and information needed to reconcile the result. I also checked report completeness at page boundaries.

Work began in August 2026 and included repeated improvements to the interface, data persistence, scan queue, resumption, and exports. Tests also covered large batches of scans and behavior after a restart. Acceptance on the target Windows environment, with a physical scanner, and an actual import remain separate conditions of the pilot.