Vanijya AI
AI-powered commerce for India's street vendors.
Vanijya AI gives India's 63 million street vendors and kirana store owners the negotiation intelligence, price discovery, and multilingual communication tools that were previously only available to large enterprises — bridging the information asymmetry that keeps small traders trapped in poverty cycles.
The Story
Why this exists.
The Problem
India's informal traders negotiate blind. They don't know the current wholesale price of tomatoes in the nearest mandi. They can't communicate fluently with suppliers from other states who speak different languages. They make pricing decisions based on habit and guesswork — not data. Meanwhile, the buyer walking into their shop has a smartphone with real-time pricing data and zero reason to share it.
Why It Matters
India's informal economy employs over 400 million people. A 10% improvement in their negotiation outcomes — enabled by price intelligence and AI communication tools — translates into hundreds of dollars of annual income recovery per vendor. Multiply that by millions and it becomes one of the highest-leverage economic interventions possible.
The Solution
A Next.js platform powered by Gemini AI and Serp API that gives vendors real-time commodity price intelligence, AI-powered multilingual negotiation scripts, smart deal evaluation, and a direct buyer-seller connection layer — all accessible from a basic Android phone.
Product Features
What it does.
Real-Time Price Intelligence
Serp API scrapes live commodity prices from government mandi portals, wholesale markets, and e-commerce platforms. Vendors see the current buy/sell price range for any commodity before entering a negotiation.
AI Negotiation Coach
Gemini AI generates negotiation scripts tailored to the commodity, current market price, vendor position, and buyer persona — in the vendor's preferred language. Hindi, Kannada, Tamil, Telugu, Marathi supported.
Multilingual Communication
Real-time translation and communication tools that let vendors interact with suppliers and buyers across language barriers — Gemini handles translation with commercial context preservation.
Smart Deal Evaluator
Enter a deal offered by a supplier — Vanijya AI scores it against current market prices and signals whether to accept, counter, or walk away. Plain-language explanation in the vendor's language.
Direct Buyer-Seller Network
A verified directory of buyers and suppliers in the vendor's commodity category and region — removing brokers and reducing the margin leakage from intermediaries.
System Architecture
How every layer connects.
Technical Deep Dive
Under the hood.
Price Intelligence Pipeline
When a vendor queries the price of a commodity, Serp API fetches live results from Agmarknet (government mandi portal), IndiaMART wholesale listings, and regional news. The raw results are fed to Gemini with a structured extraction prompt that returns: current wholesale price range, trend direction (rising/stable/falling), and the recommended buy/sell window. The entire pipeline completes in under 3 seconds.
// Price intelligence pipeline
const serpResults = await serpApi.search({
q: `${commodity} wholesale price today India ${state}`,
num: 5,
});
const priceData = await gemini.generateContent({
contents: [{
parts: [{ text: PRICE_EXTRACTION_PROMPT + JSON.stringify(serpResults) }]
}],
generationConfig: { responseMimeType: "application/json" }
});Multilingual Negotiation Script Generation
The negotiation coach prompt is engineered with four components: (1) commodity and current market price context, (2) the vendor's target price and acceptable floor, (3) the buyer/supplier persona, (4) language and cultural register instruction. Gemini returns a structured negotiation script with an opening offer, expected counter, and final position — all phrased naturally in the target language. Cultural negotiation norms (e.g. relationship-building before price discussion in South India) are baked into the prompt.
Engineering Decisions
Challenge → Decision → Result.
Performance & Scale
By the numbers.
Deployment & Infrastructure
Production setup.
Deployment
Next.js on Vercel Edge CDN. MongoDB Atlas for vendor and deal data. Serp API for price scraping. Gemini 1.5 Flash for AI features.
CI/CD
GitHub Actions — TypeScript + ESLint check on PR. Auto-deploy to Vercel on merge to main.
Monitoring
Vercel Analytics for performance. MongoDB Atlas for query health. Custom logging for Gemini API latency.
Challenges & Failures
What broke — and how I fixed it.
What I Learned
Key takeaways.
- 01
Prompt engineering for multilingual outputs requires cultural context, not just language specification.
- 02
For real-time price data, a general search API is often more current than a domain-specific official API.
- 03
Designing for low-end devices means optimising for 3G latency and minimal JavaScript — Next.js with SSR is essential.
Future Roadmap
Where this is going.
v2.0 — 2026
- —Voice interface for vendors who prefer speaking over typing
- —WhatsApp bot integration — access Vanijya AI without installing anything
- —Offline price cache — last 7 days of prices available without connectivity
Screenshots
The product.
Vendor price intelligence and negotiation dashboard
AI negotiation coach in regional language
Technology Stack
Built with.
Frontend
AI
Data
Database
Deployment
Ready to dive in?