Commander Capital console
My card-deck business ran on two Excel workbooks: one to build orders, one for bookkeeping. When an order showed up about 120 cards short with nothing on screen to show it, I rebuilt both as one app. It plans orders, prices them, splits the buying across shops, refuses to let an order go out short, and keeps the books with FIFO cost of goods sold.
- Order Builder → Cart: the red Order coverage panel lists the 17 copies with no shop. Drag a card from Undecided into a shop and watch the count drop.
- Build → Price this order: every line is priced against the store and market, with over-market cards flagged.
- Books: revenue, fees, FIFO COGS and ROI by sale. Then General ledger for the trial balance.
What's under the hood
- One invariant, checked everywhere: no copy of a card is dropped between the order and the shops. Coverage is counted in copies, not card names, and a split across shops can't be applied until it adds up.
- Books: sales come in from eBay's API, Mercari CSV exports and manual entry. COGS is FIFO against the cost of each built batch, plus per-deck fixed costs such as mailers and sleeves. A journal and trial balance are derived from the books, and every entry balances.
- Data integrity: every save is atomic, keeps a restore point, and refuses a write built on a stale revision so two open tabs can't overwrite each other.
- Scale: about 23,000 lines across 34 test files, including end-to-end browser tests that drag real cards between real columns.
In the demo, the store's partner API is replaced by a stand-in that serves sample prices. Card names and market prices are real (from Scryfall). Everything else is the production code.