Q-Dispatch

AI-powered hospital insurance claim dispatch platform — Quantum AI Ltd. (quantumai.co.uk)

Q-Dispatch automates the cashless insurance claim lifecycle at a private hospital billing counter across four stages:

1. Packet Assembly & Pre-Audit — a clerk enters the patient details and itemised bill; an AI auditor (Claude) validates every line item before submission and blocks dispatch on hard errors.
2. Instant Claim Dispatch — the validated bill is converted to an NHCX-style claim email and sent to the insurer/TPA over SMTP, with a unique tracking token generated and inbox surveillance armed.
3. Smart Inbox Surveillance — a poller checks the reply inbox over IMAP every 90 seconds; when a reply quoting the tracking token arrives, Claude parses the decision (approved / partial / rejected / more-info), amounts, deductions and document requests.
4. Automated Counter Clearance — the patient copay (total − approved) is pushed to the checkout screen, and the Quantum AI 0.5% service fee is logged to the hospital's monthly invoice. The fee is a hospital charge — it is never added to or shown to the patient.

Tech stack

Runtime	Node.js 20 + TypeScript (CommonJS)
Framework	Express 4
Database	SQLite via better-sqlite3 (PostgreSQL-ready schema)
AI	Anthropic Claude — claude-sonnet-4-6
Email out	Nodemailer (SMTP)
Email in	node-imap
Scheduler	node-cron (poll every 90 s)
Auth	Session cookies (cookie-parser)
Frontend	Single public/index.html, vanilla JS, no build step
Deploy	Railway.app (reads PORT from the environment)

Local development

cp .env.example .env      # then fill in real credentials
npm install
npm run dev               # ts-node, http://localhost:3000

Or run the compiled build:

npm run build
npm start

On first start with an empty database the app seeds one demo claim (Ramesh Nair, with intentional audit issues) so the full pipeline can be exercised immediately.

Default login

Set LOGIN_USER / LOGIN_PASS in .env. The defaults in .env.example are qdispatch_admin / ChangeThisPassword123 — change them before deploying.

Environment variables

See .env.example for the full list — Anthropic key, SMTP and IMAP credentials, per-insurer claim email addresses, the app login, the session secret, and the QAI_FEE_RATE (default 0.005).

On Railway these are injected by the platform; locally they are read from .env.

Deploy to Railway

railway.toml is preconfigured:

[build]
builder = "nixpacks"
buildCommand = "npm install && npm run build"

[deploy]
startCommand = "node dist/index.js"

Add the environment variables from .env.example in the Railway dashboard and deploy. The SQLite file lives under data/ (gitignored); mount a Railway volume at /app/data if you need it to survive redeploys.

Demo flow without live email/AI keys

The Simulate TPA Reply panel (Stage 3) lets you paste a raw insurer reply and have it parsed instantly — no IMAP round-trip required. A realistic sample reply is pre-filled for the selected claim.

API reference

Method	Path	Purpose
GET	/login	Login page
POST	/api/auth/login	Authenticate, set session cookie
POST	/api/auth/logout	Clear session
POST	/api/claims	Create a claim (DRAFT) with bill items
GET	/api/claims	List claims (newest first)
GET	/api/claims/:id	Full claim with items + audit log
POST	/api/audit/:claimId	Run the AI pre-audit (→ AUDITED)
POST	/api/dispatch/:claimId	Dispatch the claim (→ DISPATCHED)
POST	/api/simulate-reply/:claimId	Inject + parse a TPA reply (→ REPLIED)
POST	/api/clear/:claimId	Counter clearance + ledger (→ CLEARED)
GET	/api/ledger	All ledger entries
GET	/api/ledger/summary	Pending / this-month fee summary
GET	/api/ledger/invoices	Monthly invoice rollups
GET	/api/health	Health + last inbox poll time

All /api/* routes (except login) require a valid session cookie.

---

© 2025 Quantum AI Ltd.