Tap-Tap is an offline-first point of sale built for high-volume bars and multi-venue restaurants. Your data lives on your till. The cloud is a sync target, not a single point of failure.
Postgres runs on the venue's till PC. The server is a sync replica, not the source of truth. Lose the internet, lose Cloudflare, lose the data centre — orders still ring, tickets still print, totals still tally. The outbox drains on reconnect.
Measure, mixer, serve — picked in three taps. A "Double" is +100% via a surcharge, not a separate SKU. Keg yields know that a 20L barrel is 35–38 pints, not 40. Inventory reconciles against pour counts, not guesswork.
One terminal, switch venues from the sidebar. Shared barcode and product catalogue across all sites. Stock, sales, reports — per venue or rolled up. docker compose up moves the whole stack to another PC in minutes.
Guests speak their order in any language. AI parses, confirms, applies dietary guards (vegetarian, vegan, halal, keto, allergens), and sends to kitchen. Phone-OTP gate keeps it real customers, not bots.
Scan a barcode at intake, Tap-Tap reads it from a shared catalogue or falls back to Open Food Facts. Receipts get attached to inventory lots. Yield rules turn one keg into N pints, one bottle into N nips. Nightly health snapshot tells you what's quietly running out.
Recommendations push slow-moving stock first, not whatever has the fattest margin. Freshness beats markup — and the kitchen wastes less.
Most "offline POS" products mean "cached for five minutes." Tap-Tap means it: every terminal carries its own IndexedDB outbox, every server keeps an outbound_message queue with backoff drainer, and the source of truth is the till PC itself. The cloud is a convenience, never a dependency. If the day the cellular tower dies is the day you most need to close out, Tap-Tap will close out.
Most POS systems were designed for restaurants and have a "bar" tab bolted on. Tap-Tap was designed by operators who run both. Spirit modifier groups, double-by-surcharge pricing, keg yield reconciliation, ABV-aware reporting, happy-hour rules with floors and ceilings — all primitive types in the data model, not afterthoughts.