AddressSync

Update your address once. Every agency you consented to gets it automatically.

Most systems make you submit an address-change form to each bank, utility, and office separately, then wait for each one to "approve" it. That's backwards. Your address is yours — you should change it in one place, and agencies you trust should subscribe to it.

This is a working prototype of that flip: citizen-owned address + consent + push-based sync, built like India Stack's account aggregator / DigiLocker consent model.

How it works

 ┌─────────────────────────────────────────────┐
│ AddressSync │
│ │
citizen ───────►│ OTP login ──► one versioned address │
(Aadhaar) │ │ │
│ update v1 ──► v2 │
agency ◄────────│ pull API (needs active consent) │
│ ▲ │
agency ◄────────│ signed webhook ◄──┤ │
│ outbox worker │
└─────────────────────────────────────────────┘

• The citizen has one canonical address; every change is a new version, old rows are never mutated (free audit trail).
• Consent is a first-class row per (citizen, agency): granted or revoked.
• Every state change writes an event to an outbox table in the same DB transaction. A background worker delivers those events as webhooks to each consented agency, HMAC-SHA256 signed so the agency can verify the sender.
• Agencies can also just pull the current address via GET /agency/addresses/{consent_id} — the API returns 403 the moment consent is revoked.
• Aadhaar is never stored raw: only a masked ref (XXXX XXXX 1234) and a hash for login lookup.

Run it

cd backend
python3 -m venv .venv && .venv/bin/pip install -r requirements.txt
.venv/bin/python seed.py
.venv/bin/uvicorn app.main:app --port 8000

Open http://localhost:8000/demo — a single page with the citizen panel on the left, an agency panel in the middle, and the live sync feed on the right. Demo credentials are prefilled automatically (Aadhaar 111122223333; in dev mode the OTP is shown on screen instead of being sent by SMS).

Docker instead: ./start.sh from the repo root.

Prove it without a browser

./demo.sh # replays the whole story with curl, step by stepcd backend && python smoke_test.py # asserts every step end-to-end

The story it tells, in order:

1. Citizen logs in via OTP, updates their address (v2)
2. A new agency registers and points its webhook at the built-in fake inbox
3. Citizen grants consent → current address is pushed immediately
4. Citizen updates again → agency receives the change with no action needed
5. Citizen revokes → agency is notified, and its next pull returns 403

API

Method	Path	Who	What
POST	/citizen/otp/request	public	request login OTP
POST	/citizen/otp/verify	public	verify OTP → JWT
GET	/citizen/me	citizen	profile, current address, consents
PUT	/citizen/address	citizen	new address version (+ notify)
POST/DELETE	/citizen/consents[...]	citizen	grant / revoke agency access
POST	/agencies/register	public	create agency (API key shown once)
POST	/agencies/login	public	API key → JWT
PUT	/agency/webhook	agency	set push endpoint
GET	/agency/addresses/{consent_id}	agency	pull current address

Interactive docs: /docs.

Webhooks agencies receive: address.updated, consent.revoked — JSON body with X-Signature: HMAC-SHA256(webhook_secret, body) and X-Event-Id headers. Failed deliveries retry 3 times with backoff, then mark as failed.

Stack

FastAPI · SQLAlchemy · SQLite · PyJWT · httpx. One process: API + worker.

What's simulated vs real

Simulated: the UIDAI layer (OTP is returned by the API in dev mode instead of arriving via SMS) and the agencies themselves (the demo page hosts a fake receiver so pushes have somewhere to land).

Real: consent lifecycle, address versioning, transactional outbox, HMAC-signed delivery with retries, revocation propagation, JWT auth for both sides.

What production would add: actual UIDAI eKYC integration, Postgres, a real job queue (Celery/SQS) instead of the in-process asyncio loop, Alembic migrations, encrypted webhook secrets, rate limiting, and an audit-log UI.

Deploy your own

Free public URL on Render (no credit card needed):

1. Push this repo to your GitHub account
2. dashboard.render.com → New → Web Service → connect the repo
3. Render auto-detects the root Dockerfile. Instance type: Free
4. Add environment variables:
   • WEBHOOK_BASE_URL = https://<your-service>.onrender.com (so webhook pushes loop back correctly)
   • SECRET_KEY = any long random string
   • DEV_MODE = true (OTP shown on screen instead of SMS)
5. Set Health Check Path to /health, then deploy → open https://<service>.onrender.com/demo

Free-tier notes: sleeps after 15 min idle (~30–60s cold start on next visit); the SQLite file resets on redeploy/restart, but seed.py reseeds on boot so the demo always works — history just isn't permanent.

License

MIT