How actions run¶

Ambience re-evaluates your scenes continuously. Whenever the winning scene changes — because a condition changes state, the time crosses a boundary, or something else shifts — it applies the new winner's actions immediately.

If you want to test a scene's actions without waiting for its conditions to match, use the Run actions option in the scene's action menu. This runs that scene's actions once, independently of the normal evaluation cycle.

Run actions.

Action execution model¶

The actions of the winning scene are applied as follows:

In parallel, not sequentially — every action is turned into a coroutine and they're all launched together via asyncio.gather. Because they run concurrently, completion order is nondeterministic. If the same entity exists in two actions, whichever arrives last wins.
Not fire-and-forget — each call is blocking=True, and the whole batch is awaited. Ambience waits for all the service calls to finish before the apply completes (it then records last-applied, traces, etc.). It does not just dispatch and move on.
Fault-isolated — return_exceptions=True means one action raising an exception is logged ("ambience: action raised: …") but doesn't abort the others. Malformed and unexposed actions are skipped with a warning before the gather (so they never become coros).

Next: Settings reference.