Piyush Solanki

Queue Cure — Real-Time Clinic Queue Manager for Neighbourhood Clinics

Neighbourhood clinics have no system to manage patient queues. Receptionists track 50+ patients from memory under pressure, patients have zero visibility into wait time — so they can't step out, can't plan, can't relax. The doctor has no idea how many patients are left. The core problem: three roles, no shared source of truth, all working blind. Process I started by identifying the real problem — not "build a queue system" but eliminating uncertainty for three different roles simultaneously. I first considered HTTP polling (hit /api/state every 2 seconds) but rejected it: 50 patients polling every 2s = 25 requests/second, and the patient screen would always be 0–2 seconds behind. I chose Socket.io WebSocket instead — server pushes state to all clients the moment anything changes. I kept all state on the server as a single source of truth, so clients never talk to each other. Every mutation goes: client emits → server mutates → server broadcasts to ALL. This eliminated sync conflicts entirely. For wait time, I rejected hardcoded estimates and built a rolling average from actual consultation durations — by the 3rd patient, the system knows Results Queue updates reach all 3 screens in under 100ms via WebSocket push. A receptionist can add a patient in ~3 seconds (name → Enter, token auto-assigned). Wait time estimates self-correct after the first few patients using real consultation data. The system handles duplicate detection, one-level undo, keyboard shortcuts, QR code sharing, and sound notifications — all without a database, running entirely in Node.js memory with zero setup friction for a single clinic. Reflection The biggest gap is persistence — a server restart loses the entire queue. I'd add SQLite or Redis so state survives restarts. I'd also add multi-doctor support (one receptionist, multiple parallel queues) and a TV display mode showing the current token in large text for the waiting area. SMS/WhatsApp notification when a patient is 2 tokens away would remove the need to watch the screen entirely. If scaling beyond a single clinic, I'd move to Redis pub/sub to handle multiple Node processes behind a load balancer.