Daily stock decisions, answered by AI
An agentic AI demand forecasting tool for street vendors in Hazratganj, Lucknow —
combining live weather, local events, and custom inventory to deliver daily procurement recommendations.
| Name | GitHub |
|---|---|
| Tushar Bajpai | @Tushar |
| Sankalp Saini | @Sankalp |
| Subrat Dwivedi | @Subrat |
| Utsav Singh | @Utsav |
Repository: github.com/subrat-dwi/stall-brain
"Hazratganj's iconic street market has vendors who over-stock or under-stock daily, leading to waste and lost revenue. Build an agent that reads footfall patterns, weather, local events, and social buzz to recommend optimal daily procurement."
Lucknow context: Hazratganj · Janpath Market
Every morning, thousands of street vendors in Hazratganj make the same high-stakes guess: how much stock should I buy today?
They have no tools. They rely on gut feel, last week's memory, and WhatsApp forwards. The consequences are real:
- A vendor overbought potatoes on a rainy day — the crowd never came, the stock rotted.
- The next day, fearing rain again, he bought less — but it was an LSG match day, the streets were packed, and he sold out by 4 PM, losing thousands in revenue.
This isn't bad luck. It's a data problem. The signals exist — weather forecasts, event calendars, day-of-week patterns — but no one has put them together for a Hazratganj vendor. Bazaar Brain does exactly that.
Bazaar Brain is a multi-page agentic web app that acts as a daily procurement advisor for street vendors. A vendor picks their stall type, customizes their item list, and the AI agent synthesizes live weather data, a Lucknow-specific event calendar, and weekly footfall patterns to produce a precise, actionable stock recommendation — in English and Hindi.
The AI doesn't just answer a prompt. It reasons across multiple data signals before generating output, making it genuinely agentic rather than a simple chatbot.
Stall Profiles Seven vendor types built around Lucknow's actual street economy — Chaat Stall, Chai Tapri, Juice Corner, Flower Vendor, Bhutta Stall, Snacks & Namkeen, and Balloon Seller — each with pre-loaded default inventory.
Editable Inventory Vendors can add, remove, or rename items in their stall list before forecasting. The AI uses only the vendor's actual items — no generic suggestions.
Live Lucknow Weather Fetches real-time weather from Open-Meteo using Hazratganj's exact coordinates (26.85°N, 80.95°E) and maps conditions to footfall impact multipliers. Rain doesn't just mean "bad weather" — it means 0.4x footfall for chaat but a boost for bhutta.
Lucknow Event Radar A built-in calendar of local events — Lucknow Mahotsav, Navratri, IPL LSG home matches at Ekana, public holidays, and weekly footfall patterns — feeds into the demand calculation before the AI ever sees the prompt.
Demand Multiplier Engine Before calling the LLM, the system computes:
Demand Multiplier = Weather Impact × Event Impact × Day-of-Week Pattern
This structured signal is passed to the model, grounding the AI output in real numbers rather than vibes.
Bilingual Output Every forecast is available in English and Hindi (Devanagari), because a tool is only useful if the vendor can actually read it.
Forecast History & Feedback Past forecasts are stored locally. Vendors can mark predictions as accurate or not, creating a personal accuracy log over time.
Landing Page → Setup Page → Forecast Page
(welcome + (pick stall (live signals +
how it works) + edit items) AI output)
The app is structured across three pages with a clear stepper, ensuring vendors always know where they are and what to do next.
| Layer | Technology |
|---|---|
| Frontend | React 19 + Vite |
| Styling | Tailwind CSS v4 + CSS Variables |
| Routing | React Router v6 |
| AI / LLM | Groq API — llama-3.3-70b-versatile |
| Weather | Open-Meteo API (free, no key required) |
| Icons | Lucide React |
| Storage | localStorage (forecast history + feedback) |
| Deployment | Vercel |
Coding Agent used : Copilot with GPT Codex
1. Clone the repository
git clone https://github.com/placeholder/bazaar-brain.git
cd bazaar-brain2. Add your Groq API key
Create a .env file at the project root:
VITE_GROQ_API_KEY=your_groq_api_key_here
Get a free key at console.groq.com — no credit card needed.
3. Install and run
npm install
npm run devOpen http://localhost:5173.
- WhatsApp delivery — send the daily forecast directly to a vendor's phone each morning
- Crowdsourced social signals — parse local hashtags like
#Hazratganjand#LucknowFoodiesto catch viral footfall spikes - Group buying — aggregate demand across nearby vendors to unlock wholesale pricing on shared items
Built with 💛 for the street vendors of Hazratganj, Lucknow.
APL Hackathon 2026