PARTIAL SAVE: writes ONLY the 'AI' field of TitlePublicConfigurationModel; all other v1/v2 fields stay untouched. InApp AI generation features of the title - what a PLAYER can generate at runtime through the engine (text/NPC chat, images, video, speech, music, 3D). Enabled is the master switch; Features maps a stable FeatureKey (no '.' or '$') to one feature: its Modality, a Model = the id of a MODE (Text: an AI Coder mode such as "standard") or a TIER (media: an asset generation tier such as "image-balanced") - list the usable ones, with what the publisher's plan allows, with the dashboard action GetInAppAIModels (Text features that AllowImages need a mode that AcceptsImages), Behavior defaults per modality (no platform caps: MaxTokens is bounded only by what the mode's model can produce), Safety prompt checks, an audience Gate (SegmentGate), a Schedule (ScheduleSpec), per-player Limits (LimitSpec) and the in-game PriceOptions the PLAYER pays (in-game resources and/or ONE store purchase per generation - a Purchase entry in the option's base price; never inside PremiumTiers, never two per option). Each generation is charged to the PUBLISHER's credits by the engine at the model's price; per-title spend caps (in credits) live in the PRIVATE config (InAppAIBudget), not here. Server-only section: never sent to the game client or the CDN.