Sprint Serve

Sprint Serve is a restaurant POS used by restaurants in Surat, India. We built two Muse Spark assistants on top of it, and both are grounded in the restaurant's own data. For guests: "veg for 2 under ₹700" becomes a real meal from the live menu. Muse Spark searches the menu and proposes dishes through tools, while the server enforces veg/egg limits, meal shape and a budget that includes GST. Every price comes from our quote engine, never from the model, and checkout re-prices the order. For owners: questions like "how did this week go?" or "why were sales down, and what should I do?" are answered in English, Hindi or Gujarati, by typing or by voice. A fast tier picks read-only reports from a fixed metric registry (no SQL, no customer names or phone numbers). A deeper tier writes the analysis with no tools, at a reasoning effort chosen per question. Every ₹ amount and percentage is checked against the reports before the owner sees it, and a prompt injection planted in a dish name can't change the answer. All demo data is synthetic. Open source under Apache-2.0.