Darly

Privacy-first ambient caretaking system using Ring entryway telemetry to monitor elderly daily routines and detect safety emergencies without indoor cameras.

Comment

Inspiration

When I joined the Amazon Developer Hackathon, I explored the requirements across all tracks. In the Ring track, one priority category immediately stood out to me: Caretaking.

I thought about how deeply this applies to our closest relatives — our parents and grandparents. As family members, we want to look after them and fulfill our filial duties, but due to work, distance, and daily circumstances, we cannot always be beside them 24/7.

At the same time, traditional surveillance solutions put cameras inside private rooms (bedrooms, living rooms), which feels intrusive and strips away an older person's dignity and personal independence. They often reject indoor cameras, and wearable panic buttons are frequently forgotten on nightstands.

I created Darly to solve this dilemma: using only three non-invasive Ring entryway devices (Ring Video Doorbell, Ring Alarm Contact Sensor, and Ring Alarm Motion Detector) strictly on the perimeter of the home. This allows children and caregivers to have complete peace of mind that their loved one woke up, made their morning tea, and is safe, without ever placing a camera inside their personal living quarters.

What it does

Darly monitors natural household rhythms exclusively through entryway telemetry and provides an intuitive, block-based dashboard for caregivers:

  1. Ambient Routine Baseline Monitoring: Establishes waking windows and daytime activity intervals (such as morning porch visits) using foyer motion and door contact signals. Flags unusual inactivity when expected routine intervals elapse without detected movement.

  2. Night Wandering & Door Safety Watchdog: Monitors front door contact status during night curfew hours (22:00–06:00). Instantly flags anomalies if an exterior door opens in the middle of the night or remains unlatched beyond a configurable safety limit (e.g., 5 minutes).

  3. Entryway Event Classification & Escalation: Correlates exterior motion and doorbell events with interior door contact state. If a delivery arrives on the porch but the door remains closed, it is filtered as a benign courier drop-off, eliminating false alarm fatigue. If an emergency occurs, caregivers have a two-tier escalation workflow (neighbor check and 911 dispatch with address display).

  4. "What The System Sees" (Privacy Telemetry Inspector): An in-dashboard inspector displaying side-by-side payloads: the raw Amazon Ring JSON:API webhook payload vs. Darly's normalized internal event. Demonstrates clearly that the system receives only discrete state transitions (open/closed, motion, ding) — never audio, video, or private household data.

How I built it

  • Ring Integration & Webhook Pipeline: Adapted core Ring connection patterns and token management from the official AmazonAppDev/ring-api-helloworld starter template (under the MIT License). Built a resilient webhook receiver (POST /api/webhook) that validates incoming Amazon Ring JSON:API payloads at runtime using Zod schemas.
  • Interactive 3D Entryway View: Built an interactive 3D house and porch model with Three.js (House3DView.tsx), visualizing door swing angles, doorbell chime rings, and motion halos in real time.
  • Deterministic Care Engine: Developed a TypeScript safety engine (src/lib/engine/) consisting of routine baseline schedules (routine-baseline.ts), deterministic heuristic rules (rules.ts), and an anomaly detector (anomaly-detector.ts). It avoids non-deterministic AI hallucinations for safety-critical alerts, providing instant, predictable verdicts (Safe, Info, Warning, Emergency).
  • Modular Block-Based Dashboard: Designed a clear, clean interface using Next.js 16 (App Router, Turbopack), React 19, and Tailwind CSS 4. The layout is organized into focused, modular blocks (cards) so that caregivers of any technical background can understand the situation at a glance.
  • Web Audio API Sound Engine: Procedural synthesizer (audioChime.ts) generating ambient doorbell chimes and telephone tones without external audio file dependencies.

Challenges I ran into

  • Developer Sandbox Friction: Accessing the Ring Developer Playground required mandatory government passport/ID verification (KYC), which blocked immediate live cloud testing. To overcome this, I developed a standalone offline test harness with 4 pre-configured care scenarios (normal_morning, night_wandering, missed_routine, delivery_filter), enabling complete evaluation of the engine and UI without physical hardware binding.
  • Noise Filtering: Distinguishing between benign external activity (mail delivery, neighbors walking by) and actual resident departures without indoor cameras. I solved this through deterministic event correlation between exterior doorbell/motion triggers and interior door contact states.

What's next for Darly

  • Integrating ambient voice prompts via Alexa+ Agent Skills for gentle spoken reminders if morning routine movement is delayed.
  • Predictive multi-day routine drift analysis to identify gradual changes in mobility or sleep habits over weeks and months.

Built With

Share this project:

Updates

Submission history