Description:
-
Analyze existing scenes: read their triggers, conditions, actions, and parameters to understand how they work.
-
Synthesize automations: produce an understandable description of the logic of a scene or a set of scenes, without having to examine them one by one.
-
Identify interactions: detect dependencies between scenes, virtual devices used, and sequences likely to influence the overall behavior of the installation.
-
Answer questions in natural language: explain why an action is triggered, under what conditions a device is controlled, or how an automation works.
-
Propose improvements: identify redundant scenes, simplifiable conditions, or possibilities for grouping and factoring automations.
-
Modify existing scenes: allow the user to request in natural language a modification of a condition, a trigger, or an action, or even the creation of a variant of an existing scene.
-
Have a validation mode: present the proposed changes and their impact before applying them, let the user accept or refuse them, and preserve the initial configuration in case of cancellation.
The goal is to facilitate the understanding, maintenance, and evolution of existing installations, without having to manually reconstruct all their logic. This feature would also help better support the evolution of Gladys: for example, analyze the current heating scenes to determine to what extent a new native feature, such as Thermostat, could replace or simplify certain automations.