Smart Tablekeeper: AI Restaurant Booking

Smart Tablekeeper is an AI-powered restaurant reservation, discovery and intelligence platform built through a BAND Desktop software factory. Instead of a rigid booking form that ends in "no availability", it turns a plain-language request like "dinner for 5 this Saturday around 7 PM, quiet place, budget ₹3,000" into structured requirements, checks real availability, and returns ranked options. The product rests on four pillars: Understand: an AI Booking Assistant extracts party size, date, time, occasion, seating preference, location and budget, validates the output against a strict schema, and asks for any missing critical field. Results appear as transparent cards showing rating, review count, cuisine, estimated group spend (clearly labelled as an estimate), distance, seating match and a Book Table action, with plain reasons such as "Within budget" or "Quiet seating available". Recover: when the preferred slot is taken, Smart Fallback offers nearby times or alternative restaurants, and a Smart Waitlist notifies the diner when an opening appears (Notify → User Confirms → Book). An optional Scheduled Booking Agent can attempt a booking at a future time. Guarantee: reservation logic uses transactional locking and database constraints so simultaneous requests can never double-book a table, and idempotency keys make retries return the original result. Malformed input returns documented validation errors, never generic 500s. Improve: aggregated demand and feedback signals become restaurant-side insights, such as requested times, party-size patterns, budget bands and conversion, to help venues improve service without exposing private user data. The build follows a Planner → Builder → Verifier factory loop with three generic agent mandates. Every task produces evidence: changed files, tests run, and a PASS/FAIL verification report. The whole stack is Dockerized, builds offline, and is delivered in staged folders: JSON API, Web UI, Concurrency and Domain Extension.