Build a terminal model picker from the account's live model catalog, then apply the selection through a dynamic model-policy callback.
catalog query -> getAvailableModels() -> model/context/reasoning picker
|
execution query <- resolveModel(policy) -----------+
Complete the repository setup, then:
npm install
npm startThe picker displays enabled models, configurable context-window sizes, and reasoning-effort levels from runtime metadata. Press Enter to accept each default, or enter a displayed number or value.
The catalog and execution sessions are separate so the interactive selection
happens outside resolveModel. The callback remains fast and simply returns
the cached policy for each LLM request.