MJ Ablanque
Idol Fairies Beauty logoFull-stack E-commerceStorefront + Operations + Automation

Idol Fairies Beauty

Live store · my own clearance stockAI-assisted build

A responsive Korean beauty shopping platform built around the full order journey—from live product availability and guest checkout to payment verification, delivery tracking, referral rewards, and back-office control.

Built with
Next.jsReactTypeScriptSupabasePostgreSQLVercelResendGeminin8nGoHighLevel

Problem

A growing online beauty store needs accurate stock, low-friction checkout, local payment and shipping support, and a practical way to manage every order without stitching together disconnected tools.

Solution

A single full-stack platform keeps the catalog, checkout, payment proof, shipping, order status, referrals, Messenger support, and admin workflows connected to the same operational data.

Result

Customers can complete and track an order without creating an account, while the store manages products, inventory, payments, shipping, referrals, and fulfillment from one system.

System Map

How the System Works

Start with the three system branches, then open the parts you want to inspect.

Idol Fairies Beauty Platform
Demonstration

See the Automation in Action

Selected Storefront & Automation Evidence

Public-safe visual assets from the Beauty storefront, checkout, admin, and the live GHL + n8n Messenger AI workflow. Names and addresses are cropped from test orders.

Responsive Beauty Storefront

Problem

Beauty shoppers need a clear way to browse a changing catalog across phones and larger screens without seeing stale availability.

Solution

The Next.js storefront presents skincare and makeup collections, product details, variants, live stock states, a persistent cart, and responsive shopping layouts.

Result

The public catalog and administrative product records use the same inventory source instead of separate manual listings.

Verified Workflows

Test Results

Download regression report

62 checks across two independent audits plus a live re-verification pass. 61 pass; 1 hallucination was caught while testing the live Messenger bot and fixed the same day.

Test AreaWhat We VerifyResult
Inventory & CheckoutStock handling from cart through order creation, including the earlier issue where stock was held indefinitely on unpaid ordersPASS — 1-hour reservation window auto-releases stock; confirmed in code and via a live unpaid test order
Guest Checkout & ShippingCart math, shade/variant stock caps, and destination + weight-based shipping calculationPASS — verified across multiple SKUs and destinations
Policies & SEOReturns/shipping/privacy/terms pages exist; robots.txt and sitemap.xml resolvePASS — previously 404, now live in production
Payment Proof FlowGCash/GoTyme instructions, upload accepts common image formats, Gemini extraction flags mismatches for admin reviewPASS — human-in-the-loop decision confirmed, no auto-approval
Referral ProgramReferral code capture at checkout, dashboard, and reward clearing lifecyclePASS — verified in checkout, dashboard, and admin reward flows
Messenger AI — Accuracy35 test questions written from typical customer questions: authenticity, pricing, condition/expiry, payment methods, delivery, returns, and owner privacyPASS — 35/35 after one fix. Testing caught one hallucinated answer in the live Messenger bot, which I corrected and re-tested.
Messenger AI — Live ArchitectureGHL trigger → n8n Gemini agent → tool calls (Search Products, Check Order Status, Shipping Estimate, Referral Info, Human Handoff) → reply returned to GHLPASS — confirmed directly in the published, live n8n workflow
Site HealthConsole errors, broken image requests, product-level SEO metadataPASS — zero console errors, zero broken images across full audit session
UI RegressionMessenger widget overlapping the last shade-selector button on variant-heavy product pagesFIXED — pointer-events hit-strip logic confirmed in code

Build a Storefront That Runs Beyond Checkout.

Connect the customer journey to the inventory, payment, fulfillment, referral, and support workflows behind it.

Discuss your process