InceptiveTan
Built at Meta Global AI Developer Hackathon · Oct 26, 2026 · Remote

Mane Bazaar is a marketplace that home entrepreneurs run entirely from a chat, in their own language. The problem. Across Karnataka, women make and sell holige, pickles, snacks, children's clothes and tailoring from home, and small farmers sell their produce nearby. Their businesses are real, but they are invisible. Orders live in a paper notebook, sales travel by word of mouth, and customers outside their street never find them. Online marketplaces expect English, typing, forms, product photography and a website, which shuts out exactly the sellers who need reach most. Their records stay on paper, so they have no clear picture of what sells, who comes back, or what to prepare. The solution. Mane Bazaar keeps her in the chat she already knows and does the digital work for her. She says hi, picks Kannada, English or Hindi, and an AI agent sets up her shop by asking one simple question at a time, with buttons to tap. She lists products by sending a Kannada voice note, a photo of her notebook page, or a short video where she shows her batch and says the prices. The agent reads everything back and saves only when she says yes. It can clean up her product photos, answer licence questions such as FSSAI with official sources, and tell her each day what needs doing first. What buyers get. Every shop appears instantly on a public website where buyers browse by category, region and delivery terms, or simply ask "pickles under 200 in Bagalkot?" and get an answer from live listings. "Order in chat" places the order in conversation, notifies the seller straight away, and gives the buyer a UPI link that pays the seller directly. Mane Bazaar never holds her money. What she gets back. A private dashboard in Kannada or English to manage orders and payments, and a weekly review that tells her in plain words how much came in, what sold best, what did not, how many customers returned, and what to make more of next week. Her listings stay from day to day; each morning the agent simply asks whether anything changed. Why it matters. Visibility, a record of her sales and a simple way to get paid, without asking her to learn a new app or a new language. Built with Muse Spark, Muse Wonder, Muse Voice Transcribe, Muse Image and SAM 3.1 on Meta Model API. The live demo runs on Telegram because WhatsApp test messaging requires business verification; WhatsApp is the intended production channel. All code in this repository was written during the hackathon.