Skip to content

Duo Mentis

Small merchants in India struggle to market their businesses because existing AI tools are expensive, English-first, and difficult to use.

Code Video Cerebral Valley

Core trick

Gemini Live directs the merchant's camera while parallel image generation produces ten campaign variants, scored by a Gemini critic.

Tags

voiceproductivity

Description

Small merchants in India struggle to market their businesses because existing AI tools are expensive, English-first, and difficult to use. During this hackathon, we built Vyapari, an AI platform that helps local merchants create personalized marketing campaigns using natural voice interaction. The merchant simply shows their shop and product through the camera and speaks in their native language. Our multimodal pipeline creates a Brand Brief, and using NB2 Lite, we explore 10 marketing campaign strategies in parallel instead of generating just one advertisement. The AI automatically evaluates every campaign and selects the best one. Unlike existing tools that rely on templates or AI avatars, Vyapari creates advertisements featuring the merchant and their own product. The selected campaign is converted into a video advertisement using Omni Flash, personalized for customers, delivered through WhatsApp, and printed with a real UPI QR code. Unlike existing tools that stop after generating content, Vyapari doesn't stop until a real customer makes a purchase.
vijaym2k6/Google-DeepMind-VYAPARI★ 00 forksPythonCreated Jul 11, 2026 · last push Jul 11, 2026

README excerpt

# ವ್ಯಾಪಾರಿ VYAPARI — Speak. Sell.

**Voice-to-first-sale engine for India's neighborhood merchants.**
A shopkeeper talks to a camera in Kannada. The AI directs her like a film crew,
reasons about her market in the open, manufactures a scored ad campaign,
prints a poster, delivers personalized ads over WhatsApp — and doesn't stop
until a real customer pays real rupees.

Built at the Google DeepMind Bangalore Hackathon, July 11 2026.

## Stack
Gemini 3.1 Flash Live (director + perception) · Gemini 3.5 Flash (brief, strategy,
batched critic, prediction) · Nano Banana 2 Lite (hypothesis grid + personalization)
· Gemini Omni Flash (film + conversational editing) · Gemini 3.1 Flash TTS
(vernacular voice) · FastAPI + vanilla JS PWA · Pillow deterministic Indic typography.

## Run
```bash
cd backend
pip install -r requirements.txt
cp .env.example .env          # add GEMINI_API_KEY, set MOCK_MODE=0
python -m uvicorn main:app --port 8000
```
Open http://localhost:8000 — press **SPACE** (or click) to start.
Hotkeys: **E** film edit · **D** deliver · **P** payment received · **F** fallback audio.

`MOCK_MODE=1` (default) runs the complete demo flow with generated placeholders —
no API key needed. Smoke-test real models after key swap: `GET /api/smoke`.

## Architecture
See [VYAPARI.md](VYAPARI.md) (full spec) and [RUNBOOK.md](RUNBOOK.md) (demo-day ops).

Key design decisions:
- **Deterministic Indic typography** — Pillow + Noto fonts own every glyph of
  Kannada/Devanagari; the image mod

Team