✓ All 3 Phases Complete F&B Retail · GrabMart PH AI-Assisted Build 2026

Building a Real-Time
Inventory System
for a GrabMart Business

How a non-technical flight attendant built and deployed production-grade inventory software in 6 weeks — solving a 2-year problem for ₱0 in development cost.

1,184
SKUs tracked
50+
Bundle recipes solved
~$5
Monthly infra cost
6 wks
Concept to deployment
30s
Daily PDF processing

Nightcap — A Liquor Brand on GrabMart

Nightcap is a liquor retail business based in Pasay, Metro Manila, operating exclusively through GrabMart Philippines. Their catalog spans over 1,184 SKUs — wines, spirits, beers, mixers, and grocery items.

A significant portion of revenue comes from bundle products: curated combinations of individual items sold together as a single listing on Grab. The business is managed by owners who also hold full-time jobs as flight attendants — meaning any system built had to run as automatically as possible.

"We needed a system that works while we're 30,000 feet in the air."


2 Years of Inventory Chaos

GrabMart treats every listing — whether a single product or a bundle — as an independent SKU. When a bundle sells, GrabMart only decrements the bundle's own stock count. It has no native mechanism to deduct stock from the individual component items that make up the bundle.

📦

Stock discrepancies accumulated daily across 50+ bundle products

📊

Manual Google Sheets tracking proved too tedious and error-prone

No reliable way to know actual available stock at any point in time

⚠️

Items randomly marked out of stock by Grab even when stock was available

🔄

Risk of overselling increased constantly as discrepancies compounded

⏱️

The problem had persisted unresolved for over two years


Vibe Coding with AI — No Developers Needed

Rather than purchasing an off-the-shelf inventory tool (none of which solve the bundle deduction problem anyway), the decision was made to build a custom system through vibe coding — an approach where a non-technical business owner builds functional software by collaborating iteratively with an AI assistant.

The entire system was designed, built, debugged, and deployed through a single ongoing conversation with Claude (Anthropic). No prior coding experience was required. The business owner followed instructions, copied and ran code, and provided feedback at each step.

"The entire project proves that technical skill is no longer the barrier to building software."


5 Core Components

Component 01
Product Database

All 1,184 products imported from GrabMart CSV into Supabase via a custom Node.js script. Idempotent — safe to run multiple times.

Component 02
Bundle Tagger Web App

Identifies bundle products and maps their component recipes. Handles multi-product bundles and quantity bundles natively.

Component 03
PDF Order Processing Engine

Since Grab's API isn't available to merchants, the team uploads their daily receipts PDF. The engine extracts orders, matches items with 4-level fuzzy matching, and deducts stock automatically — in under 30 seconds.

Component 04
Live Inventory Dashboard

Full-featured web dashboard: 1,184 SKUs with live stock, bundle alerts, sales tracking, restock history, undo system, and COGS tracking. Accessible from any device worldwide.

Component 05
Cloud Deployment

Deployed to Railway — permanent public URL, 24/7 accessible, auto-deploys from GitHub on every push.

Bonus
Upload Streak Tracker

Calendar showing daily upload streaks and completion rate — keeps the team accountable to daily stock reconciliation.


What It's Built With

Technology Component Purpose
SupabaseDatabasePostgreSQL cloud storage for all inventory data
Node.js + csv-parseImport ScriptBulk import from GrabMart CSV export
Vanilla HTML/JSBundle TaggerWeb app for mapping bundle recipes
HTML/JS + pdfjsDashboardLive inventory + PDF order processing
GitHubVersion ControlCode hosting and deployment trigger
RailwayHosting24/7 cloud deployment (~$5/month)
GrabMart PDF ReceiptsAPI WorkaroundDaily order data source
Claude by AnthropicAI Co-BuilderFull system design, build, and debug

All 3 Phases Shipped

Phase 01 Account setup + database creation (Supabase, GitHub, Railway) Complete
Phase 02 Product import, bundle tagger, dashboard, PDF engine, sales & restock features Complete
Phase 03 Cloud deployment to Railway — live 24/7 on any device worldwide Complete

What Made This Hard

No GrabMart API Access
GrabMart's API requires going through an official partner program — not self-serve for smaller merchants. After weeks of waiting, a PDF-based workaround was built. The merchant downloads daily receipts and uploads to the dashboard — processed in under 30 seconds.
Two Distinct Bundle Types
During tagging, two bundle types emerged — multi-product bundles (e.g. Bombay Sapphire + Canada Dry) and quantity bundles (e.g. 12pcs Canada Dry). The engine handles both, with availability calculated from the component with least stock.
PDF Format Inconsistencies
The browser-based PDF reader extracted text differently from server-side parsing. Item lines appeared with spaces, prices appeared on separate lines, product names wrapped across lines, and order IDs sometimes contained letters. Each edge case required a specific parser fix.
Supabase Free Tier Pausing
Supabase pauses free tier projects after one week of inactivity, causing a "Failed to fetch" error on the deployed app. The fix: keep the dashboard active with daily PDF uploads, which naturally prevents inactivity pauses.
No Technical Background
The business owner had zero prior coding experience. Every step was executed through conversational AI instructions — from terminal commands to database queries to deployment configuration. This is also the most important part of the story.

What Changed After Launch

1,184
SKUs fully tracked in a cloud database for the first time
50+
Bundle recipes permanently documented — never completed before
30s
Daily PDF processing time — fully automated stock deduction
$5/mo
Total infrastructure cost — vs. $50-200/mo for off-the-shelf tools
6 wks
Concept to full cloud deployment
₱0
Development cost beyond time investment

What This Proves

This project demonstrates that the barrier to custom software development has fundamentally changed. Problems that would previously require hiring a development agency at significant cost can now be solved iteratively by business owners themselves — with AI as the technical co-builder.

Most importantly, the system is fully owned by the business. No vendor lock-ins. No per-SKU fees. No dependency on a third-party tool that could change pricing or discontinue a feature.

"A non-technical business owner can build and deploy production-grade software through AI-assisted vibe coding — without writing a single line of code from scratch."

— AroVerse by Studio Fabillar · 2026

Have a real problem
that needs a real system?

I build operational tools, websites, and digital products for businesses that need things that actually work — not just look good.

Work with me Book a Build Sprint