extension-data-viewer

The data viewer is an admin-only, paginated inspection extension for the stable canister state of Caffeine AI apps. Every Caffeine app already ships with the caffeineai-data-viewer mops package and the moc --generate-view-queries flag enabled, so the feature is pre-installed rather than something you add. When the actor body contains include MixinViews(), the compiler auto-exposes a controller-only __<var> query for every stable variable of a supported type, giving operators a way to browse users, items, orders, logs, or any backing collection without writing bespoke endpoints.

Supported types map to specific query signatures: Map.Map<K, V> exposes (?K, ?Nat) -> [(K, V)], Set.Set<K> exposes (?K, ?Nat) -> [K], and the sequence types [V], [var V], List.List<V>, Stack.Stack<V>, and Queue.Queue<V> expose (?Nat, ?Nat) -> [V]. A null cursor starts from the beginning and a null count returns everything from the cursor onward, supporting simple pagination. Each generated query traps on any non-controller caller, so the queries are strictly for admin dashboards and debug viewers and never for user-facing endpoints. The Lintoko rule include-mixin-views errors if the actor is missing include MixinViews(), since removing the include disables every auto-generated viewer.

Use this extension whenever a request calls for a viewer, dashboard, debug panel, or admin browse over backend data. The doc sets out firm rules: never use the __<var> queries as a substitute for public list/feed/search methods (those still need normal public query func definitions), never declare an actor member whose name begins with __ because it collides with generated queries or a reserved prefix, and remember that pure/immutable collections (pure/Map, pure/Set, pure/List, pure/Queue) are not supported because the viewer is mutable-only by design. The intended workflow is simply to declare a stable variable of a supported type and let the matching __<var> query appear automatically.