active-model

Injects a one-line <active-model> call-out into the system prompt each turn, read live from ctx.model. It tells the orchestrator which model it is running and updates on a mid-session /model switch.

Purpose: let orchestration adapt to the live model. When the active model is strong, fan work out to cheaper subagents; when it is already cheap or local, do the work directly. Pairs with the subagent-routing extension, which reads this identity when deciding how to route and escalate.

No configuration.

Install

pi install npm:@guygrigsby/pi-active-model