Skip to main content

Scene Generation Agent

The Scene Generation Agent is Builder Agent's second Agent on stage: activated when the user, after completing device recommendations, wants to orchestrate "one-tap" smart scenes.

What Is a "Scene"​

A scene = a set of device states executed simultaneously with one tap.

The user presses a button or says a voice command, and multiple devices act in concert. For example, "Good Night Mode" — turn off the lights, close the curtains, lower the AC temperature — all in one tap, no conditions required. :::Core distinction from automation A scene is actively triggered (by a person), while automation is passively triggered (by a condition). See Scene vs Automation for details. :::

Simple vs Complex: How to Tell​

Whether a scene is "simple" or "complex" comes down to two dimensions:

  • Device count: 1-3 devices in a single room = usually simple
  • Logic complexity: Does it need multi-condition checks, branching, sequential execution, or cross-room coordination? The more convoluted the logic, the more complex it is
DimensionSimple SceneComplex Scene
Device countFew (1-3)Many (cross-room, cross-system)
Logic structureA set of actions runs togetherMulti-condition + multi-branch + sequential
ExampleGood Night (lights off + curtains + AC)Welcome Scene (different behavior day vs night)

Simple Scene Orchestration​

Best for single triggers with a group of devices acting together. For example:

  • "Good Night Mode": All lights off + curtains closed + AC set to 26°C sleep mode
  • "Movie Mode": Living room lights dimmed + curtains closed + TV turned on

Complex Scene Orchestration​

For complex logic involving multiple conditions, devices, or branches. For example:

  • "Welcome Scene": Guest arrives during the day → Full brightness in living room + background music playing; Guest arrives at night → Living room lights at 50% + warm ambient lighting
  • "Away Patrol": No one detected → Lights off in sequence → AC off → Security camera activated + door lock armed

Output & Display​

Once scene generation is complete, the Agent outputs:

  • Conversational text: Describes the scene name, trigger method, and devices involved
  • Right-panel Flow preview: A visual representation of the trigger logic and device relationships — clear at a glance