V-Stores
One platform running a two-branch store in Pune: the billing counter, the back office and the online shop, all on a single stock count.
16,000+products catalogued
2stores, one live stock count
8staff roles
Problem
Vikas Store sells stationery, toys, gifts and festive goods across two branches. V-Stores replaces scattered tools with one system the owners control. Staff bill at the counter and run a 16,000+ product catalogue from the back office today. The storefront opens to online customers once the catalogue is complete.
Systems
- Payments
- Razorpay behind a provider interface. Webhooks are verified with HMAC signatures, and settlement is idempotent: the browser and the webhook can both report a payment, and the order is still created exactly once.
- Counter billing
- Barcode scanning through the camera, printed receipts, and invoice numbers allocated inside the sale's transaction. When two counters sell the last unit at the same moment, only one sale goes through.
- Catalogue at scale
- 16,000+ products searchable by name, SKU and tag. PostgreSQL trigram indexes and single-query listing pages keep every page at one trip to the database.
- Inventory
- Stock tracked per store. Every sale, restock, import and edit is recorded as a stock movement with the staff member's name: a complete audit trail.
- Bulk import
- Excel and CSV stock sheets matched by SKU, so hundreds of products load in a single step.
- Access control
- Role-based access with 40 permissions across 8 roles, checked on every request and cached for speed.
- Customer sign-in
- Passwordless email OTP through Resend, with expiry, retry limits and rate limiting.
- Notifications and delivery
- Order updates over the WhatsApp Cloud API, and courier booking through Shiprocket.
- Discoverability
- A server-written page head for every route, with structured product data for Google, plus a sitemap and a Google Merchant shopping feed.
- Analytics
- First-party visitor tracking that ties each order back to the campaign or channel that brought the customer in.
Decisions
- One round trip per request. Relation joins and aggregate SQL brought the heaviest admin screen from 3.1 s to 0.3 s.
- Swappable providers. Each integration has a stand-in driver, so the whole shop runs in development without paid accounts. Production refuses to start without the real ones.
- Separate environments. UAT and production run as separate Docker services, and every change passes typecheck, tests and a build in CI before it ships.
React 19TypeScriptExpressPrismaPostgreSQLRazorpayWhatsApp Cloud APIDocker
Private repo · walkthrough on request