Wheelness

Cleaners work in basements and stairwells where 4G drops constantly. A task board that fails on a dead connection means lost proof and angry customers, so the board treats the network as optional.

Every status change goes into a localStorage queue first. Proof photos go into IndexedDB. A sync loop replays the queue whenever connectivity returns, in order, with the server reconciling duplicates. The cleaner just works; the phone catches up on its own.

• Status flow: pending, in progress, arrived, completed or skipped
• Photos persist in IndexedDB until upload is confirmed
• Unresolved tasks carry over for 7 days automatically

Four roles share one schema, which is exactly how data leaks happen. Instead of trusting every query to remember a WHERE clause, isolation lives in Postgres itself.

Each request opens an AsyncLocalStorage context carrying the authenticated user and role. The database wrapper reads that context and sets the session variables that Row-Level Security policies check. A query that forgets to filter still cannot read another tenant's rows, because the database refuses. No ORM in the path, just typed raw SQL at full speed.

• AsyncLocalStorage scopes identity per request
• RLS policies on 30+ tables check session variables
• Raw SQL with TypeScript types, no ORM overhead

Referral credits in rupees attract industrial-scale abuse. The engine assumes every signup is suspicious and makes it prove otherwise.

Each new account passes through layered checks: self-referral blocks, hashed device fingerprints, hashed phone numbers matched against accounts deleted in the last 12 months, IP rate limits, and per-referrer velocity caps. Rewards draw from a monthly budget with a FIFO queue, and wallet credits commit atomically so a credit can never be issued twice.

• Device fingerprint and phone hashes catch recycled identities
• 10 signups per IP per hour, 15 referrals per referrer per day
• Monthly budget cap with FIFO reward queue