Control Every Room. Intelligently.
A unified smart home platform connecting every switch, sensor, and scene — orchestrated by real-time WebSockets and an AI assistant that understands natural language.
Everything Your Home Needs
Built on a production-grade stack — not a prototype. Every feature is designed for reliability, extensibility, and day-to-day use.
Device Control
Manage smart switches with granular on/off, manual/auto modes, and load-level slider (0–1000) — all synced in real time across every client.
Room Management
Structure your home as Homes → Rooms → Devices. Bulk-control entire rooms or drill down to individual channels with a tap.
Real-time Sync
Bidirectional WebSocket events dispatch commands and surface device responses in under 50ms — no polling, no stale state.
AI Assistant
LangGraph-powered NLU understands intent and entities in natural language — turn on lights, set scenes, or query energy usage conversationally.
How It Works
Connect Your Devices
Flash the WemosD1 slave firmware and pair each switch module to the ESP32 master gateway. The gateway joins your local network and establishes a persistent WebSocket tunnel to the backend.
ESP32 + WemosD1 Firmware · WebSocketOrganize Rooms & Scenes
Log in with your account, create a Home, then group devices into Rooms. Assign names, icons, and default modes (manual / auto). Changes propagate to every connected client instantly.
JWT Auth · Flutter App · Real-time SyncTalk to Your Home
Ask the AI assistant anything: 'Turn off all bedroom lights' or 'What is the kitchen power usage today?' LangGraph resolves intent, extracts entities, and dispatches the right device commands — safely.
LangGraph · gpt-4o-mini · Safety ControlsYour Home Understands You
Built on LangGraph with OpenAI's gpt-4o-mini at its core, the AI assistant interprets natural language across multi-turn conversations to control devices, query analytics, and manage automations — with safety guardrails at every step.
Phrase commands in plain English — intents and entities are extracted automatically.
The assistant remembers conversation history to handle follow-up questions naturally.
All commands pass through invariant checks before execution to prevent unintended states.
Ask for energy summaries, device uptime leaderboards, or trend analysis by room.
Hey, can you set the living room to movie mode?
Sure — dimming living room lights to 30% and closing the smart blinds.
What's today's energy usage so far?
You've consumed 4.2 kWh today. Kitchen is your top draw at 1.8 kWh.
Built on a Modern Stack
Each layer of the platform was chosen for production reliability — from IoT firmware to AI orchestration.
The system uses a full real-time WebSocket bus — not REST polling — so device state changes appear on every connected client (mobile, web, legacy Android) within milliseconds, regardless of which client sent the command.
Start Automating Today
Open-source, self-hostable, and production-ready. Deploy the backend with Docker in minutes and start pairing your first device.