PM — Settings

Cloud models

Model roster

Only models listed here appear in surfaces. Local models are managed in State. Grouped by provider.

Chat defaults

Default chat mode

Default models (pre-selected in chat)

Panel policies

Interactive (chat panels)

Reserved hot slots Max active support models Max selected participants Require family diversity

Production (pipeline panels)

Quality over speed Max loaded total Minimum distinct families Allow long-running jobs Panel rounds

Reasoning layers

Set reasoning layers independently per surface — Home Chat, LoreCore, and Pipeline each have their own KRL configuration per model.