Cartify.ai

Inspiration

You wanted to cook Hyderabad biryani. Google gave you 23 ingredients and 14 YouTube links. By the time you found the right products, you had just ordered them.

You shop every week for the same items, but it takes a full hour, like it's your first day on Earth. You have a $100 winter budget, and figuring out what fits feels like a spreadsheet project. You always forget something. The cart never learns.

The average shopper wastes 30+ minutes per weekly grocery session. Cart abandonment tops 70%. None of this is the shopper's fault. It's a broken discovery experience. That's why we built Cartify.

What it does

Cartify is a universal AI shopping agent, a conversational companion that lives inside any e-commerce site. You type one sentence. It understands your intent, your mood, and your budget. It fetches real products, builds your cart, and checks you out in under 2 minutes.

"Hey, I want to cook Hyderabad biryani under $20." That's the entire interaction. Cartify handles the rest.

Emotion Intelligence Understands sentiment and mood from natural language. Adapts recommendations to match how you feel and what you actually want, not just what you typed.

Budget Optimisation Set a spending limit, and Cartify fills your cart intelligently. "Under $20" means under $20; it doesn't round up.

Voice-Enabled Shopping Hands-free interactions powered by ElevenLabs. Shop while cooking, driving, or doing anything else.

Smart Cart Memory Weekly bundles are saved. One click to repurchase everything. The Cart learns what you always forget.

Agent Analytics Dashboard Merchants see top searched products, cart value distribution, sales by category, and bundle conversion rates powered by real user data behavior.

How we built it

The user types or speaks a natural language request. Groq (Llama model) parses intent, mood, and budget constraints. SerpAPI fetches live Walmart products that match. The cart is built and ready for one-click checkout.

The agent is embedded as a universal widget that any e-commerce platform can drop Cartify in without platform-specific code. All API calls run through Netlify Functions to keep keys secure and handle CORS cleanly. Cart state and purchase history are persisted in Supabase (PostgreSQL), with Supabase Auth handling user sessions. Zustand manages client-side state.

Challenges we ran into

Real-time product data reliability SerpAPI results vary heavily by query quality. We built a prompt engineering layer on top of LLM output to normalise queries before hitting the API, which dramatically improved product match accuracy.

Budget constraint precisionGetting the LLM to reliably respect budget ceilings across diverse product combinations required careful system prompt design plus a post-processing validation step before cart confirmation.

Serverless API proxying on Netlify Handling SerpAPI keys securely in a client-side The React app meant migrating all API calls to Netlify Functions, including resolving CORS issues, environment variable scoping, and --legacy-peer-deps conflicts at build time.

Emotion detection in a shopping context Mapping free-text sentiment to actionable product recommendations isn't obvious. We iterated through multiple prompting strategies before settling on a structured intent-extraction approach.

Accomplishments that we're proud of

• A working end-to-end AI agent: natural language in, checked-out cart out
• Budget-aware recommendations that actually respect the constraint
• Voice shopping that feels natural, not gimmicky
• A persistent bundle history that makes weekly re-ordering a single click
• An analytics dashboard that gives merchants real signals, not vanity metrics
• Universal widget architecture, drop it into any site in minutes

What we learned

The hardest part of an AI shopping agent isn't the AI; it's the data. Getting clean, Relevant, real-time product information is the actual bottleneck.

Emotional intelligence in commerce isn't a gimmick either. When the agent understands why you're shopping, the recommendations improve dramatically. And architecting widget-first (platform-agnostic from day one) forced cleaner code and better separation of concerns throughout the project.

What's next for Cartify.ai

• Multi-platform expansion, Amazon, Target, and local grocery integrations
• Seasonal and celebration-aware recommendations (winter prep, Diwali, Thanksgiving)
• Deeper bundle analytics, which combinations convert, which get abandoned, and why
• Personalised weekly auto-cart generation from purchase history
• White-label SDK so any e-commerce platform can embed Cartify in under 5 minutes

Built With

• api
• css
• deno
• elevenlabs
• groq
• llama
• netlify
• postgresql
• react
• serpapi
• supabase
• tailwind
• typescript
• vite
• zustand