Case study · Smart home and IoT

Digilite: a smart home app for the whole family

Digilite Automation is an Android smart home app, and we built it. Families switch and schedule appliances from anywhere, replace a drawer of remotes, run routines, follow their power use and share the home with the people they live with.

  • Android
  • Google Assistant
  • Amazon Alexa
  • 1,000+downloads on Google Play
  • 4.9star rating from verified reviews
  • 35%+of active homes run routines every week
  • 25%less electricity used by homes after setup
01 · The brief

The brief: one app for every switch in the house

Digilite installs smart devices in family homes. Its customers wanted a single place to run them all, from the sofa or from the office. The app had to be simple enough for every person in the house, and safe enough to hold the keys to the front door.

We built Digilite Automation for Android. It talks to Zigbee and Wi-Fi devices, sends infrared to older appliances, and answers to Google Assistant and Amazon Alexa.

Six everyday problems shaped the build. Each one is below, with what we did about it.

Who the app is for

  • The home owner

    Sets up rooms and devices, decides who gets access, and checks the energy page.

  • The family member

    Has a separate login and controls the rooms the owner allows, without borrowing a password.

  • The guest

    Gets limited access for a short stay, such as the front door and the lights.

  • The voice user

    Never opens the app for simple jobs. A spoken command to Google Assistant or Alexa does it.

Six problems and what we did about each

  1. 01
    The problem

    Lights, fans and water heaters stay on when nobody needs them, and the electricity bill climbs every month.

    What we built

    Remote switching plus timed schedules, so each appliance switches at the right moment, even when the family is out.

  2. 02
    The problem

    Every TV, set-top box and air conditioner arrives with its own remote, and remotes go missing.

    What we built

    Infrared control inside the app. One phone now replaces the remotes for TVs, projectors, air conditioners and set-top boxes.

  3. 03
    The problem

    Repeating the same steps every night, such as switching things off at bedtime, wears people down.

    What we built

    Routines, scenes and workflows, each one firing a group of actions together, started by the clock, the room temperature or movement.

  4. 04
    The problem

    An ordinary electricity meter cannot tell a family which appliance is costing them the most.

    What we built

    Live power graphs for the home, plus a detailed energy page for each smart plug, with a device picker on every graph.

  5. 05
    The problem

    Every person in the house needs safe access of their own, including to doors, while the home stays protected when it is empty.

    What we built

    Separate logins for each member, session control, workflows for locks and doorbells, and the Lock Home away mode.

  6. 06
    The problem

    Devices drop off the network when the Wi-Fi or the power fails, and people cannot tell what to fix.

    What we built

    A step-by-step guide for devices that drop offline, with stronger Zigbee support, device history and sensor handling.

02 · Feature tour

What we built

Eight features carry the app. The screens are our designs of each one, with fictional homes, people and readings.

  1. 01

    Every appliance from one screen

    The home screen lists rooms and devices as tiles. A tap switches a light, a fan or a water heater, wherever the phone happens to be. Each tile shows its room and its current state.

    Schedules sit on the same devices. A water heater can switch itself off at 07:30, and a porch light can come on at dusk. People set these once and stop thinking about them.

    At the top, a live power reading shows what the home is drawing now. It links straight to the energy page.

  2. 02

    One phone in place of a drawer of remotes

    Older appliances only listen to infrared. Digilite sends the same signals their own remotes send, so a TV, a projector, an air conditioner or a set-top box works from the app.

    Each appliance gets a remote laid out for its job. An air conditioner gets a temperature dial, modes and fan speeds. A TV gets channels, volume and power.

    Infrared appliances can join schedules and routines too. So the bedroom air conditioner can switch off at 06:30 on weekdays without anyone touching it.

  3. 03

    Routines, scenes and workflows

    A routine runs several actions together. Bedtime might turn off the downstairs lights and the TV, cool the bedroom and lock the front door.

    Routines start from the time, the room temperature, movement or a door opening. A condition can hold a routine back, for example only cooling the bedroom when it is above 26°C.

    This is the feature families come back to. More than 35% of active homes run routines or workflows every week.

  4. 04

    Live energy use, appliance by appliance

    The energy page shows today’s use, the live draw and an hourly graph. A row of chips switches the graph between the whole home and each smart plug.

    Below the graph, appliances are ranked by what they used. That ranking is what turns a vague bill into a clear action, such as moving the water heater schedule.

    After setup, homes use 25% less electricity.

  5. 05

    Family access without shared passwords

    The owner invites people into the home. Each one signs in with their own account, so nobody shares the owner’s password.

    Members and guests get the access the owner chooses. A guest can have the front door and the lights for a weekend, then lose access on its own.

    Every account lists its signed-in devices. A person can sign out one device, or every device at once, which matters when a phone is lost.

  6. 06

    Doors, doorbells and an away mode

    Smart locks and doorbells have their own workflows. If a door stays open too long, every member gets an alert.

    Lock Home is the away mode. One switch locks the doors and sets the house up for an empty day.

    Lock and doorbell flows are the most sensitive part of the app. They go through extra security testing on top of the normal test rounds.

  7. 07

    Help when a device goes offline

    Devices drop off when the Wi-Fi fails or the power cuts out. The app marks the device as offline and shows when it was last seen.

    A short guide walks through the likely fixes in order: the hub’s power, the home network, the distance to other devices, and re-pairing.

    Device history and better Zigbee handling sit behind this screen, so a flaky sensor is easy to spot.

  8. 08

    A personal TV guide

    The entertainment guide keeps track of what is showing on the family’s TV. People can follow their own shows and see what is on now.

    Because the set-top box already answers to infrared, the guide can change the channel for them.

Planning something like this? Talk it through with the team that built it, in a free 30-minute call.

Book a free 30-minute call
03 · How it works

How it works

A smart home app is a remote control that works from anywhere, so the real work happens between the phone and the house. This is how we build that link for an app like Digilite, using the stack it runs on.

The architecture

Commands and device state travel over MQTT. People and permissions live behind the REST API. The phone never talks to a device directly when it is away from home.

Architecture: Commands and device state travel over MQTT. People and permissions live behind the REST API. The phone never talks to a device directly when it is away from home.

People and devices

  • Zigbee sensors, lock and doorbellLow-power radio devices paired to the hub
  • Home hubZigbee coordinator, Wi-Fi and infrared blasters in the house
  • Infrared appliancesAir conditioners and older appliances, reached by a blaster in each room
  • Wi-Fi plugsEach holds its own connection to the broker

Phones and voice

  • Android appKotlin: control, schedules, routines, energy and access
  • Voice assistantsGoogle Assistant and Alexa, after account linking

Our system

AWS IoT Core

  • MQTT broker on AWSEvery hub and Wi-Fi device keeps one secure connection open
  • Device stateDevice shadows: last reported and desired state for each device

App path

  • REST API in Node.jsSign-in, homes, devices and routines over HTTPS
  • Homes and accessMembers, roles and sessions
  • Routines engineTime, temperature, motion and door triggers
  • Live updatesWebSocket push to open apps

Data

  • PostgreSQLHomes, people, devices, routines and permissions
  • Time-series storePower readings and hourly totals

Outside services

  • Firebase Cloud MessagingAlerts when the app is closed
  1. Firebase Cloud Messaging to Phones and voice: Alerts
  2. Phones and voice and App path: App and voice requests over HTTPS; live state back over WebSocket
  3. App path to Device state: Desired state
  4. Home hub and MQTT broker on AWS: MQTT over TLS
  5. Wi-Fi plugs and AWS IoT Core: Commands down, reported state up, at QoS 1
  6. Zigbee sensors, lock and doorbell to Home hub: Zigbee radio: sensor events and lock state
  7. Home hub to Infrared appliances: Infrared codes
  8. App path to Data: Homes, routines and power readings

What happens when someone says “turn off the bedroom AC”

Seven steps run between the spoken words and the air conditioner going quiet. Permission is checked before any command reaches the house.

What happens when someone says “turn off the bedroom AC”. Seven steps run between the spoken words and the air conditioner going quiet. Permission is checked before any command reaches the house.
  1. 01Send the commandVoice to APIThe assistant sends an execute request for the device, with the token from the linked account.

  2. 02Check the personAPI to DataThe API finds the home behind the token and checks this person may control this device.

  3. 03PublishAPI to BrokerThe API writes the new desired state, and the broker sends the change to the hub at QoS 1, so it arrives at least once.

  4. 04Reach the houseBroker to HubThe hub receives the message on its open connection and fires the air conditioner’s infrared code.

  5. 05Report the resultHub to BrokerThe hub reports the new state. Infrared is one-way, so it reports the command it sent.

  6. 06Update everyoneBroker to APIThe API stores the state and reports it back to Google or Amazon, so the assistant knows too.

  7. 07Refresh the appAPI to AppAny phone with the app open gets the new state over its WebSocket and redraws the tile.

One message bus between the phone and the house

Devices in a home sit behind the family’s router, so nothing outside can reach them directly. The answer is for the house to call out. Each hub and Wi-Fi device opens one long-lived connection to an MQTT broker and keeps it open.

MQTT is publish and subscribe. A hub subscribes to the topics for its home and publishes on others. AWS IoT Core runs the broker for us and supports both MQTT 3.1.1 and MQTT 5.

MQTT defines three delivery levels. QoS 0 sends once with no guarantee, QoS 1 delivers at least once, and QoS 2 delivers exactly once. AWS IoT Core supports QoS 0 and 1, not 2. So we send commands at QoS 1 and make every command safe to repeat: “set to off” rather than “toggle”.

Every device connects with its own X.509 certificate, registered with AWS IoT before it can talk. One stolen device certificate cannot be used to speak for another home.

Knowing when a device has gone quiet

The offline guide only helps if the app knows a device is offline. MQTT has a feature built for this, the Last Will and Testament. A client leaves a message with the broker when it connects, and the broker publishes it if the client disconnects without warning.

AWS IoT Core also publishes its own events when a client connects or disconnects, on topics under $aws/events/presence. We listen to these and mark the device offline with the time it was last seen.

For devices that drop off often, persistent sessions help. The broker keeps a client’s subscriptions and stores unacknowledged QoS 1 messages while it is away. In MQTT 5 that is set with Clean Start and a Session Expiry Interval.

What the app asked for, and what the house reports

A command and a result are two different facts. The app asks for the light to be off. The light later reports that it is off, or never answers.

AWS IoT device shadows hold both. Apps write the desired state, devices write the reported state, and the service reports the difference as a delta. A device that comes back online reads the delta and catches up.

Shadow updates carry a version number. AWS notes that a device can safely discard any delta older than the version it already holds. That stops an old command from undoing a newer one.

The app shows both facts. A tile that has been asked to change but has not confirmed it shows a waiting state, rather than pretending the job is done.

Zigbee and Wi-Fi devices

Zigbee devices use their own low-power radio, not Wi-Fi. They need a coordinator: a Zigbee adapter attached to the hub that bridges the radio to the network.

The open-source Zigbee2MQTT project is a good model of that bridge. Each device gets a friendly name and a topic. Publishing {"state": "ON"} to its set topic switches it, and an availability topic reports online or offline when it changes.

Wi-Fi plugs and switches skip the coordinator and connect to the broker themselves. The app does not care which radio a device uses. It sees one device list with one set of states.

Infrared for older appliances

Infrared is how TVs, air conditioners and set-top boxes have always been controlled. Some Android phones have an infrared emitter. Android exposes it through ConsumerIrManager, which needs the FEATURE_CONSUMER_IR feature and reports the carrier frequencies it can send on.

Most phones have no emitter, and the phone is often not in the room anyway. So the dependable route is a small infrared blaster in each room, connected to the hub. The app sends the command over the network, and the blaster sends the light signal.

Infrared only goes one way. The air conditioner never confirms it has changed. So for these appliances the app shows the last command sent. A routine that matters can pair it with a smart plug or a temperature reading as a check.

Routines that run without the phone

A routine cannot depend on someone’s phone being switched on. So routines are stored on the server and run there, or on the hub for the ones that must work without the internet.

Each routine is a trigger, optional conditions and a list of actions. Triggers are times, sensor readings such as temperature or motion, and door events. Conditions are checked at the moment the trigger fires, from the latest reported state.

Times are stored with the home’s own time zone, so a 23:00 routine still runs at 23:00 after the clocks change. Every run is logged, which is how the app can show that a routine ran six times this week.

Energy data that stays fast

Smart plugs report power readings all day, and those rows add up quickly. Graphs that read millions of raw rows get slow. The case calls for a time-series store beside PostgreSQL, and that is the reason.

TimescaleDB is one such store, built on PostgreSQL. Its hypertables split time-series data into chunks by time automatically, so a query for today only touches today’s chunks.

Its continuous aggregates hold summaries, such as hourly totals, and refresh in the background as new readings arrive. The energy graph reads the hourly totals, and only the live figure reads the latest raw reading.

Google Assistant and Alexa

Both assistants talk to the home through our cloud, not through the phone. Google calls this a cloud-to-cloud integration. Our fulfilment service answers smart home intents sent by Google.

When someone links their account, Google sends a SYNC intent and we return every device with its traits. QUERY and EXECUTE intents follow. Report State then lets us push each change to Google Home Graph, so Assistant reads the current state instead of asking.

For Alexa we build a smart home skill. Its code runs in an AWS Lambda function. Alexa discovers devices, then sends directives such as TurnOn and TurnOff, and the power controller reports a powerState of ON or OFF. We send a ChangeReport when a device changes for any other reason.

Both platforms link accounts with OAuth 2.0. Google smart home supports only the authorization code flow, and Alexa asks for the same grant type. So one authorisation server in our API serves both.

Accounts, members and sessions

A home has an owner, members and guests. Each person has their own account and their own sessions. Permissions sit on the home and the device, and every command is checked against them on the server.

Sign out of one device, or all of them, works because sessions live on the server. Signing out revokes that session’s refresh token, so a lost phone stops working on its next request.

On the phone, the keys that protect the session are held in the Android Keystore. Google describes it as making keys harder to extract, with the key material staying non-exportable.

Alerts that reach a closed app

Door left open, doorbell pressed and device offline all need to reach people with the app closed. Firebase Cloud Messaging carries those alerts to the phone.

Since Android 8.0, every notification must belong to a channel, and people can change each channel’s behaviour. So we keep security alerts and routine updates in separate channels. Someone can silence routine updates without missing a door alert.

Privacy and store rules

Google Play asks every app to complete a Data safety form, which appears on the store listing. Digilite’s listing says nothing is shared with third parties, and that traffic is encrypted while it travels.

The app collects only personal details, app activity and app performance data. Anyone can ask Digilite to delete their data. Google Play also requires apps with accounts to offer deletion inside the app and through a web link.

Today we would also plan for Matter. The Connectivity Standards Alliance says Matter is deployed across Apple Home, Google Home, Amazon Alexa, Samsung SmartThings and Home Assistant. Supporting it lets new devices from many brands join the same home.

The stack and why we chose each part

LayerTechnologyWhy we chose it
Android appKotlin The main language for modern Android, with direct access to the platform APIs that infrared and notifications need.
BackendNode.js, REST APIs Suits an API that mostly passes small messages between apps, devices and assistants, in one language across the backend.
Real-time messagingMQTT, WebSockets MQTT is light enough for small devices and keeps a link open into the home. WebSockets push changes to open apps.
DatabasePostgreSQL, time-series store PostgreSQL for homes, people and permissions. A time-series store keeps power graphs fast as readings pile up.
Device protocolsZigbee, Wi-Fi, IR Zigbee for low-power sensors, Wi-Fi for plugs and switches, infrared for the appliances families already own.
VoiceGoogle Assistant, Amazon Alexa Covers the two assistants most homes use, both linked through one OAuth server.
NotificationsFirebase Cloud Messaging Delivers door and offline alerts when the app is closed.
CloudAWS A managed MQTT broker, device certificates and device shadows, beside the rest of the backend.
04 · Build plan

The build plan

This is how we would plan an app like Digilite today. The weeks assume one team working on the device layer, the backend and the Android app in parallel.

  1. DiscoveryWeeks 1 to 2
    • List every device type, radio and infrared appliance
    • Agree roles for owners, members and guests
    • Choose the hub hardware and MQTT topic layout
    • Clickable prototype of home, routines and energy
  2. Device layerWeeks 3 to 8
    • MQTT broker, device certificates and topics
    • Zigbee coordinator bridge and Wi-Fi devices
    • Infrared code library and room blasters
    • Online and offline detection
  3. Backend and dataWeeks 3 to 10
    • REST API for homes, people and devices
    • Desired and reported state for each device
    • Time-series store and hourly totals
    • WebSocket live updates
  4. Android appWeeks 5 to 14
    • Home, rooms and device tiles in Kotlin
    • Infrared remotes and schedules
    • Routines, scenes and workflows
    • Energy page and per-plug graphs
  5. Voice, access and securityWeeks 11 to 18
    • OAuth server and account linking
    • Google cloud-to-cloud and Alexa smart home skill
    • Members, guests and session control
    • Lock, doorbell and Lock Home workflows
  6. Testing and launchWeeks 17 to 24
    • Device lab runs with real hardware
    • Power and Wi-Fi failure tests
    • Extra security tests on locks and doorbells
    • Play Store release and 3 months of fixes
  1. 01Weeks 1 to 2

    Discovery

    • List every device type, radio and infrared appliance
    • Agree roles for owners, members and guests
    • Choose the hub hardware and MQTT topic layout
    • Clickable prototype of home, routines and energy
  2. 02Weeks 3 to 8

    Device layer

    • MQTT broker, device certificates and topics
    • Zigbee coordinator bridge and Wi-Fi devices
    • Infrared code library and room blasters
    • Online and offline detection
  3. 03Weeks 3 to 10

    Backend and data

    • REST API for homes, people and devices
    • Desired and reported state for each device
    • Time-series store and hourly totals
    • WebSocket live updates
  4. 04Weeks 5 to 14

    Android app

    • Home, rooms and device tiles in Kotlin
    • Infrared remotes and schedules
    • Routines, scenes and workflows
    • Energy page and per-plug graphs
  5. 05Weeks 11 to 18

    Voice, access and security

    • OAuth server and account linking
    • Google cloud-to-cloud and Alexa smart home skill
    • Members, guests and session control
    • Lock, doorbell and Lock Home workflows
  6. 06Weeks 17 to 24

    Testing and launch

    • Device lab runs with real hardware
    • Power and Wi-Fi failure tests
    • Extra security tests on locks and doorbells
    • Play Store release and 3 months of fixes
05 · Testing and release

Testing and release

A smart home app can look perfect on a phone and still fail in a real house. So we test against real devices, real networks and real failures.

A device lab, not just emulators

We keep a shelf of the real devices the app supports: Zigbee sensors, Wi-Fi plugs, a lock, a doorbell and infrared appliances. Every release runs its main flows against that shelf.

Infrared is checked by eye. A command is sent and someone watches the appliance respond, because infrared gives no answer back.

Pulling the plug on purpose

We cut the hub’s power and switch off the router in the middle of a routine. The app must mark devices offline, show when they were last seen, and recover cleanly when power returns.

Commands sent while a device is away must either arrive once it is back or show as not delivered. They must never apply twice.

Locks and doorbells get extra rounds

Door flows go through extra security testing. We try expired sessions, guests past their end date and removed members, and check that each one is refused.

We also check that a door-open alert still arrives with the app closed and the phone asleep.

Voice and store release

Account linking is tested for both assistants, including unlinking and linking again. We check that a change made in the app reaches each assistant without a fresh query.

For Google Play, the Data safety form is checked against what the app really collects before every release.

06 · Results

Results

Digilite now runs family homes with less effort, and the people using it rate it highly.

  • 1,000+downloads on the Google Play Store
  • 4.9star rating from verified reviews
  • 35%+of active homes run routines or workflows every week
  • 25%cut in electricity use by homes after setup
  • 1.5 yearsof automation running a user’s whole house, in their own review

One user reports that every electrical appliance in their house has run smoothly on automation for a year and a half. Another singles out the appliance schedules and calls the automation very good.

Weekly routine use matters more than downloads. It shows families handing real jobs to the app, night after night. The energy graphs do their job too: once families can see which appliances use the most, homes use a quarter less electricity.

Later updates brought stronger Zigbee support, the lock and doorbell workflows and the Lock Home away mode.

What the client says

  • Android app development

    “A smart home app only works if the whole family trusts it. AppAgentix built our Android app so every appliance and routine sits in one place, with each family member given the right access. Commands reach the device in a moment, the app shows honestly when something is offline, and voice control works through Google Assistant and Alexa. The listing holds a 4.9 star rating. Every release was tested on real devices in real homes.”

    Eric N.Senior Developer, Digilite
07 · Since launch

Since launch

A smart home app keeps changing as new devices arrive. These are the improvements an app like this receives in its first six months.

  1. 01More devices supported

    New Zigbee and Wi-Fi models added to the device list after testing on the device lab shelf.

  2. 02Routine run history

    Each routine shows when it ran, what it did and any step that failed.

  3. 03Low battery warnings

    Battery-powered sensors and locks raise an alert before they run flat.

  4. 04Clearer offline reasons

    The app says whether the hub, the home network or the single device has dropped off.

  5. 05Android updates kept current

    Each new Android version tested, with the target SDK raised as Google Play requires.

  6. 06Energy alerts

    A notice when an appliance draws far more than its usual pattern.

  7. 07Matter groundwork

    Device model and permissions reviewed so Matter devices can join the same home later.

Want a result like this for your own product? Get a written view of scope, time and price within two working days.

  • 95%+ delivered on or before the agreed date
  • 60% of clients come back for a second project
  • 3 months of fixes after launch in the build price
Get a written view in two working days
08 · Cost today

What this would cost to build today

These are our published prices, in pounds, excluding VAT, as of November 2026. An app at Digilite’s scope, with device control, infrared, routines, energy, voice and family access, sits in the full app range.

  • Discovery£3,000 to £6,0002 weeks

    Device list, scope, clickable prototype and a fixed quote. It comes off the build price if you continue.

  • First version£15,000 to £40,0008 to 14 weeks

    Android app with rooms, remote switching and schedules for Wi-Fi devices, on a managed MQTT broker.

  • This caseFull app like Digilite£40,000 to £90,0004 to 8 months

    Zigbee and infrared, routines, energy graphs, Google Assistant and Alexa, family access, lock and doorbell workflows.

  • Support and runningfrom £750 a monthMonthly

    Android and device updates, fixes and monitoring. Minimum 3 months, then 30 days’ notice.

  • Fixes after launch are included in the build price for 3 months.
  • Cloud costs are separate: the MQTT broker, storage and messages are billed by your providers and grow with the number of homes.
  • Hub, sensor and blaster hardware is priced separately from the software.

Estimate your app in a minute with our Android calculator

09 · FAQ

Questions about building a smart home app

How much does it cost to create a smart home?

For the app, a first Android version with remote switching and schedules costs £15,000 to £40,000 with us, over 8 to 14 weeks. A full app like Digilite, with Zigbee, infrared, routines, energy and voice, costs £40,000 to £90,000. Prices exclude VAT. Devices and cloud running costs are extra.

How can I build my own smart home?

Pick devices that share a standard, such as Zigbee or Matter, and a hub to bridge them to your network. Then connect the hub to a cloud broker over MQTT so an app can reach it from anywhere. Routines, voice and energy graphs are built on top of that link.

What does the smart home app do?

It is the remote control for the whole house. In Digilite, families switch and schedule appliances, use infrared remotes for older appliances, run routines, follow their power use, manage locks and doorbells, and share access with family members.

Do I need Wi-Fi for a smart home?

Not for every device. Zigbee sensors use their own low-power radio and talk to a hub. The hub then needs the home network and internet so the app and voice assistants can reach it from outside the house.

Can Alexa and Google Home work together?

They can control the same devices. A smart home cloud can link to both: Google through a cloud-to-cloud integration and Alexa through a smart home skill, each using OAuth account linking. Digilite works with both assistants.

Why do I need MQTT?

Devices sit behind a home router, so the cloud cannot call them. With MQTT, each hub keeps one light connection open to a broker and receives commands on it. It also gives delivery levels and a Last Will message that reports a device dropping off.

What are the disadvantages of MQTT?

It needs a broker that every device trusts, and that broker becomes critical. Delivery guarantees vary: AWS IoT Core supports QoS 0 and 1 but not QoS 2, so commands must be safe to receive twice. Apps also need a separate path, such as WebSockets, for live updates.

Planning a smart home app?

Tell us about your devices and the people who use them. A developer reads your brief. We reply within 30 minutes.