Turning optional model steps on and off

Some model-driven steps — like rewriting a resident's search query for better retrieval — are optional per city. This screen is where you turn one on, write your own wording for it, or leave it to the platform's.

1. Open City → Model tasks

Each task shows its own name, a short description, a toggle, and a state pill: Active or Off, plus which layer supplied that answer — Platform or Tenant.

2. Understand what the toggle means — it depends on the platform's own default

If the platform ships a task off by default, switching it on for your city is literally that: activate. But if the platform ships a task live by default, the same toggle means something different — it's an override, and switching it back off hands the task back to the platform's own wording rather than turning the task off entirely. You cannot switch off a platform-live task from here; the copy on the toggle changes to say so.

3. Write your own wording, or start from the platform's

Click "Wording" on any row to open an editor with two panels: the platform's current instruction (read-only — this is what any tenant not overriding it is running) and your own, editable one. If you have no wording of your own yet, the editor starts you off with the platform's own tested paragraph rather than a blank box you'd have to rebuild from scratch.

4. Every activation and every wording change is confirmed

Both acts change what a model runs on every turn, for every resident of your city — so both show a plain-language confirmation dialog naming the actual consequence before they commit, never a bare "are you sure?"

5. Expect roughly five minutes for a change to fully propagate