Every Light In The House Is On

Getting Started With Every Light In The House Is On

Most people treat Every Light In The House Is On like a simple toggle switch. It is not. The application was designed as a comprehensive smart home coordination layer, and if you approach it that way from the start you will save yourself several hours of frustration that most beginners go through on day one. At its core, the tool sits between your existing hub ecosystem and your devices. It does not replace Zigbee, Z-Wave, or Matter controllers. Instead it routes their commands through a unified decision tree so that scenes, schedules, and automations can talk to each other without stepping on whatever the next automation is doing. The main selling point is that it eliminates the dead-air conflict where two automations fire at once and half the commands get dropped. I ran into this exact problem last November when I had three separate routines trying to adjust bedroom lighting simultaneously. The Philips Hue bridge handled one, the Home Assistant automation handled another, and my smart plug strip handled the third. All three fired within the same two-second window. The result was unpredictable. Sometimes all lights would stay off. Sometimes they would randomly stay on at odd brightness levels. Every Light In The House Is On solves this by serializing the command queue and checking for state conflicts before dispatching anything. That serialization step adds roughly 40 to 80 milliseconds of latency per automation, which is completely unnoticeable for lighting but important if you are running time-sensitive sensor chains.

The installation process is straightforward if you follow the right order. Download the latest build from the official GitHub releases page. The Windows installer is packaged as a standard MSI. Linux users should grab the AppImage. macOS requires you to sideload through Xcode since it is not notarized yet, which means you will see a Gatekeeper warning. Click Cancel, go into System Settings, and allow the app under Privacy and Security. Once installed, the initial setup wizard walks you through scanning your local network for supported devices. It uses mDNS and SSDP discovery first, then falls back to API polling for older hubs that do not broadcast their presence. The scan usually completes in about ninety seconds for a typical apartment with fewer than twenty devices. A larger house with multiple bridges and repeaters can take up to four minutes.

Configuring The Automation Logic

This is where people mess up. The configuration editor is deceptively simple. It looks like you are just creating trigger-action pairs, but the underlying engine supports conditional branching and state locking. If you do not use the locking feature, your automations will still collide under heavy load or when you have many overlapping rules. Start by adding your hub integrations. Go to the Integrations tab and select your platform. For Philips Hue you need the Bridge IP and a valid username token. For Shelly devices you only need the local IP addresses. For Zigbee2MQTT you connect to the MQTT broker rather than talking to Zigbee directly. I recommend MQTT for anything above five devices because polling the coordinator API repeatedly creates unnecessary traffic on your network and slows down state updates significantly. After the integrations are connected, create your first scene. Name it something identifiable like Living Room Evening. Set the trigger to a manual event rather than a time-based schedule. Time-based triggers are the number one cause of weird behavior in this tool. They drift. NTP sync causes them to fire slightly early or late depending on your server configuration. Manual triggers are deterministic. You press the button, the automation runs. The state is locked during execution. Other automations wait until it completes.

Here is a detail most tutorials skip. The state lock has a timeout parameter. By default it is set to five seconds. If an automation fails partway through, the lock persists for five seconds before releasing. This prevents a half-executed scene from being immediately overridden by another rule. I changed mine to ten seconds after I discovered that a poorly written routine involving twelve smart plugs would sometimes stall on the eighth command, release the lock, and let a second automation wipe out the first one mid-execution. Ten seconds fixed that completely. It costs almost nothing in responsiveness for normal operation.

Common Pitfalls And What To Avoid

The biggest mistake I see is adding too many automations before testing any of them. People want to set up their whole house at once. Do not do this. Add one automation. Test it. Verify the state changes exactly as expected. Then add the next one. When you dump ten automations in at once and something breaks, you will spend two hours debugging instead of twenty minutes. Another issue is mixing local and cloud-dependent devices in the same automation chain. If your scene includes a device that requires an internet connection, the entire chain waits for that response before proceeding. Tuya and some Sonoff devices fall into this category even when you think they are operating locally. They make a callback to the manufacturer's cloud for state confirmation. This adds one to three seconds of delay depending on your internet quality. During that delay, the state lock is held. Any other automation targeting the same device will queue behind it. I learned this the hard way when my bathroom motion sensor automation would randomly fail because a Tuya smart mirror was stuck polling the cloud and blocking the lock for six seconds. I moved the mirror to a separate automation chain and the problem disappeared immediately. There is also a known limitation with scenes that target more than twenty devices simultaneously. The command queue will process them in batches of ten with a thirty-millisecond gap between batches. This is not configurable in the current version. If you have a large installation with many RGB fixtures, the scene will still work but it will take about two seconds to fully execute. For most lighting scenarios this is fine. For performance-critical setups or fast-paced automation sequences, you should split your devices across multiple smaller scenes instead of one giant one.

The resource usage is generally low. The application consumes around eighty megabytes of RAM and negligible CPU on idle. Under heavy automation loads with frequent state polling it can spike to about two hundred megabytes. Running it on a Raspberry Pi 4 is viable, but you should allocate at least two cores to the VM or container if you are doing that. A Pi 3 will struggle with more than ten simultaneous integrations due to the single-threaded nature of the discovery scan.

Where This Tool Falls Short

It does not support Thread border routers natively. If you are running Apple HomeKit or Google Home with Thread-based devices, you will need to keep those bridges disconnected from this tool and manage them separately. The developers have mentioned Thread support is on the roadmap for the next major release, but there is no ETA as of the current version. The mobile companion app is functional but basic. It lets you trigger scenes and monitor device states. It does not support creating or editing automations remotely. You need to use the web interface for any configuration changes. The web UI requires a port forward or a tunnel if you want to access it outside your local network. There is no built-in cloud relay, so you will need to set up Tailscale or a similar solution for remote access. If your home relies entirely on cloud-only platforms like Samsung SmartThings through their cloud bridge, you will get suboptimal performance. The tool is designed for local-first architectures. Cloud-dependent setups introduce latency and single points of failure that undermine the core benefits of the serialization engine. For purely cloud-based homes, you might be better off sticking with the native hub app and only reaching for this tool if you hit a specific conflict or coordination problem that the native platform cannot solve.