Are you dreaming of a freshly
designed website? Book a consultation today!
Contact Us
ListMint.ai
Case Study
AI-Powered Peer-to-Peer Marketplace
ListMint is an AI-powered local marketplace where sellers snap photos of items and multimodal Gemini 2.0 Flash handles the rest — extracting attributes, creating studio white cutouts, generating optimized listings, and staging cross-posts across 11 external marketplaces with automated relisting cadences.
The platform is invite-only, keeping the community trusted and spam-free. Sellers build verified reputation through ratings and Trusted Seller badges, while buyers browse a geo-filtered feed showing items in their state.
Project Type:
Web Application
Status:
Live — Invite Only
Platform:
Web (Mobile-First)
AI Engine:
Gemini 2.0 Flash
Category:
Marketplace / SaaS
Built With:
Next.js + Firebase
Platforms Supported
AI Category Detectors
Firestore Collections
Product Showcase
The Live Marketplace
Problem Statement
The Challenge
Selling secondhand items online is time-consuming and fragmented. Sellers must manually write descriptions, research pricing, pick categories, and then repeat the process across multiple platforms to maximize reach. Most existing platforms are either too large and impersonal, or too niche to attract buyers.
The goal was to build a platform that eliminates the friction of listing creation entirely, handles multi-platform distribution automatically, and fosters a trusted, community-driven marketplace experience — all from a single, mobile-first interface.
Platform Capabilities
Core Features
AI-Powered Listing Generation
One-Click Cross-Posting
Studio White Cutout
Staggered Relist Queue
Listing Boosts & Credits
Invite-Only & Admin Dashboard
AI Engine — Gemini 2.0 Flash
Photo-to-Listing AI Pipeline
The AI pipeline is ListMint's core innovation. A seller uploads 1–10 photos and Gemini 2.0 Flash analyzes all of them simultaneously to produce a complete, market-ready listing in seconds.
Beyond generic descriptions, the AI performs category-specific detection, extracting specialized attributes based on item type:
- 👗 Clothing: US/EU/UK sizes, material, color, gender
- 💻 Electronics: Brand, model, storage, RAM, screen size
- 🚗 Automotive: Make, model, year, OEM part numbers, fitment
- 🪑 Furniture: Dimensions, material, style, assembly required
- ⚽ Sports: Sport type, sizes, skill level
Fields auto-populated by AI are visually tagged with an “AI” badge so sellers know exactly what was detected vs. what they entered manually. Retail MSRP is also estimated to provide pricing context.
Distribution Engine
One-Click Cross-Posting to 11 Platforms
Once a listing is finalized, sellers can distribute it across all major secondary marketplaces with a single click. ListMint generates platform-specific titles and descriptions formatted for each destination's character limits, tagging taxonomies, and buyer audience expectations.
Category attributes are automatically mapped to marketplace-specific fields, while clothing and sports titles are auto-appended with sizing and gender metrics to maximize organic algorithmic discovery.
Built With
Technology Stack
Framework
Next.js 16 (App Router + Server Components)
Language
TypeScript 5
UI Library
React 19 + ShadCN UI (Radix primitives)
Styling
Tailwind CSS v4
Database
Cloud Firestore
Authentication
Firebase Auth (Email/Password + Google OAuth)
Storage
Firebase Cloud Storage (WebP images)
AI Engine
Firebase Genkit — Gemini 2.0 Flash
Hosting
Firebase App Hosting
Image Processing
Canvas API (WebP), heic2any (iOS HEIC), Studio Cutout
Distribution Engine
Multi-platform staggered relist queue & clipboard sync
Platform Security
Security-First Architecture
ListMint was designed with security as a first-class concern, not an afterthought. Every mutation flows through server-side authentication middleware, ensuring no client-side trust vulnerabilities exist.
- Server-side auth via
requireSessionUser()/requireAdminUser() - Input sanitization strips HTML/scripts before every Firestore write
- MIME validation on all image uploads (server-side allowlist)
- Invite-only registration with email-lockable invite codes
- httpOnly session cookies (secure, SameSite=Lax)
- Firestore security rules as defense-in-depth layer
- Admin-only actions protected by role middleware
- Content reporting system with reason categories
Firestore Data Model
usersProfiles, credits, location, settingslistingsPhotos, pricing, attributes, statusconversationsBuyer–seller message threadsmessagesIndividual messages + timestampscreditTransactionsCredit purchase/spend audit trailratingsSeller ratings and reviewswatchlistUser-saved listing referencesinviteCodesCode management (used, email-locked)reportsFlagged content reportsreferralsUser referral trackingFull Feature Set
Additional Platform Features
Ratings & Reputation
Buyers rate sellers after purchases (1–5 stars + comment). Trusted Seller badges earned through positive ratings. Public seller profiles with full rating history.
Watchlist & Notifications
Save listings to a personal watchlist and track price changes and availability over time with persistent saved state.
Mobile-First Design
Swipeable category chips with fade indicators, touch-friendly photo capture, HEIC/HEIF auto-conversion for iOS/Android, and WebP compression for optimized file sizes.
Studio White Cutout
1-click neural background isolation isolates items onto clean studio white backgrounds, elevating click-through rates and satisfying eBay, Mercari, and Amazon catalog standards.
Staggered Relist Cadence
Intelligent scheduling queue spaces out relisting requests across connected accounts to prevent burst API triggers and refresh item freshness in active buyer feeds.
Mark-Sold & Delist Hub
Two-step sold flow: select which platform the item sold on from all 11 options, then trigger automatic or guided delisting reminders across connected channels.