Solo Dragon

India has more than a crore kirana shops, and most of them run on the owner's memory and a notebook. The owner is behind the counter all day, often speaks Hindi or another regional language, and has no time to type a thousand products into an app. So three things keep going wrong: The shop has no page online, and customers drift to quick-commerce apps. 1 : The owner finds out an item ran out when a customer asks for it. 2 : Customers who leave unhappy never say why. 3 : Most shop software expects the owner to sit down and type, usually in English. At a busy counter that doesn't happen. What Shelf Voice does : FOR SHOP OWNERS : 1 : Setup is four questions in the owner's own language. Tap "Use my location" and the address fills itself in from the phone's GPS. Speak changes in Hindi, English or nine other Indian languages. One sentence can carry many updates, and the owner checks each one before it is saved. 2 : Every product gets a price, a stock count and a clean photo. When the owner sells below the MRP printed on the pack, customers see the discount. Speech, product understanding, translation, review summaries and product images all run on Meta's models. The screenshot is the real Muse Spark usage from building the app. 3 : A kirana stocks hundreds of items, so speaking isn't the only way in. The owner can tick products from a ready-made list for their kind of shop, photograph a supplier's bill, photograph a shelf, scan barcodes, or upload the Excel or CSV export from their billing app. 4 : Sales are logged by voice too ("do Maggi bike"). The restock list then says what to order, how much, and why, with the items closest to running out at the top. 5 : Nothing the AI suggests is saved until the owner confirms it on a check screen, and any confirmed change can be undone for 10 seconds. FOR CUSTOMERS : 1 : Type a PIN code or share your location to see the shops near you on a map, nearest first. No account is needed. 2 : Each shop has its own page with real photos, prices and discounts. Customers can order there, or send the cart to the shop on WhatsApp. 3 : Customers scan the QR code at the counter and leave a 15-second voice review. The owner reads a short summary: whether the customer was happy, and which items they asked for. If a customer wanted something that was out of stock, that item's restock forecast goes up. META AI MODELS INTEGRATION : 1 : Muse Voice Transcribe 1.0 handles everything spoken 2 : Muse Spark 1.3 does the understanding. 3 : Muse Image 1.0 made the product photos for the ready-made catalogue.