Shopping should feel confident.
Not complicated.
Smart Shopping Recommender combines transparent product matching with conversational AI. You bring your needs. We help you understand your options.
Your priorities come first
Budget, purpose, brand, and features shape every product score. The ranking is deterministic and explainable.
AI, with context
AI-generated advice uses the supplied catalogue and review excerpts. It is not live web research or a promise of current availability.
Your shopping stays yours
Wishlists, searches, preferences, and comparison history are private to your signed-in account.
The match score, explained
A 0–100 score combines six signals. This is a weighted recommender, not a claim of a trained predictive model.
User preference match
40%Category, query, preferred brand, purpose, and previously explored categories.
Budget compatibility
20%Products within the requested budget get full budget points; the recommendation list excludes over-budget products.
Product rating
15%Sample rating normalized against a five-star scale.
Popularity
10%Clearly labeled sample popularity normalized to 100.
Feature match
10%The fraction of requested feature keywords found in product information.
Discount / value
5%An equal blend of normalized discount and product rating.
Within the preference component: category is 35%, query 20%, brand 20%, usage 15%, and viewed/wishlist category affinity 10%. Unspecified preferences receive neutral full compatibility. Rating, budget, category, and brand constraints are applied before sorting. Feature requests affect the score; the Explore feature filter requires matching keywords.
What’s connected — and what isn’t
Working integrations
React, TypeScript, Vite, Tailwind CSS, and hash routing power the interface. Platform email/password authentication includes verified signup and password resets. The built-in AI generates recommendation briefs, chat replies, and review summaries. The built-in key-value store saves account data. The billing SDK handles existing-customer portals and verifies payment returns.
Storage and backend
The current app saves data in the built-in key-value store. A real managed relational database is a Pro feature. You have two options in the editor’s Database panel: upgrade to Pro for a managed database, or connect your own Supabase project for free. This deployment does not run an Express server or MongoDB, and does not implement a second JWT authentication system; it uses the platform’s secure authentication and SDK services instead.
Catalogue and administration
All 30 products, prices, ratings, review counts, popularity values, specifications, and review excerpts are sample data. Demo items cannot be purchased. Owner access is verified by a server function using platform ownership. The owner workspace saves private product/category drafts, not public listings. Public catalogue publishing, user administration, real purchase fulfillment, and global analytics require a connected production data layer and server-authorized mutations.
AI availability
AI advice is generated from supplied context, not live web results. The weighted ranking continues to work if AI is unavailable. Advice can be imperfect; verify seller details, current prices, and product specifications independently.