Set when the at-risk load fails — the workspace shows a "couldn't load" banner with retry instead of the misleading no-results empty state.
Whether the at-risk list is loading for the open prediction.
The ranked at-risk rows for the open prediction's latest run (empty until loaded / when no run yet).
ReadonlycapabilityCapability cards for the first-run intro (what PS can do), shown when the catalog is empty.
Protected ReadonlycdrEmbedded co-pilot conversation state. The chat-area starts in "new conversation" mode; on the first
send it creates the conversation and emits conversationCreated, at which point we capture the live
conversation and flip out of new-mode so subsequent sends append to the same thread. Without this
wiring the suppressed empty-state input has no valid conversation to write into and the first send
silently no-ops (matches the proven Form Builder co-pilot pattern).
"Send to a list" in-flight guard + last-result message (P1 #4).
Protecteddestroy$Plain-language "what's driving this" drivers for the open prediction (global feature importance).
Provider-scoped Predictive Studio engine. Resolved in loadSection via
PredictiveStudioEngine.GetProviderInstance so a multi-provider client (or an embedded
non-default provider) reads the right server's PS reference data. Bound into the hosted panel's
[engine] input once loading completes.
True until the engine resolves; gates the panel host so [engine] is never bound before it's ready.
Set when loadSection fails to resolve the engine. The section templates render a "couldn't load" banner with a Try-again button (retryLoad) instead of silently showing empty panels, so a failed load is honest rather than indistinguishable from a genuinely-empty studio.
ProtectednavigationTab ID for query param notification scoping. Set by resource wrappers that render child dashboards, so the child knows which tab it belongs to. If not set, falls back to Data.Configuration.tabId.
If specified, this provider will be used for communication and for all metadata purposes. By default, if not provided, the Metadata and RunView classes are used for this and the default GraphQLDataProvider is used which is connected to the same back-end MJAPI instance as the Metadata and RunView classes. If you want to have this component connect to a different MJAPI back-end, create an instance of a ProviderBase sub-class like GraphQLDataProvider/etc, and configure it as appropriate to connect to the MJAPI back-end you want to use, and then pass it in here.
Protected ReadonlySectionFont Awesome icon for the section (chrome + shell tab).
Protected ReadonlySectionStable section identifier for agent context + (subclass) titling.
Protected ReadonlySectionHuman-readable section label used in chrome + agent context.
catalog (the home grid) ↔ workspace (a selected prediction).
StaticɵdirStaticɵfacThe business catalog cards, most-trustworthy first, derived from the engine's published models.
Returns either the default Metadata provider or the one specified in the Provider property, if it was specified
Returns either the default RunQuery provider or the one specified in the Provider property, if it was specified
Returns either the default RunReport provider or the one specified in the Provider property, if it was specified
Returns either the default RunView provider or the one specified in the Provider property, if it was specified
Public door label for templates (the error banner title). Mirrors the protected SectionLabel.
"Save scores" — still routed conversationally through the co-pilot (needs a designated write-back column).
"Share / export" — download the at-risk list as a CSV (dependency-free).
ProtectedextraDeep agent context for the Predictions door: catalog counts + names, and (in workspace) the selection + at-risk breakdown.
ProtectedGetRead current query params from tab configuration. Use in initDashboard() / ngOnInit() to get initial URL state.
Get this component's tab ID. Checks ParentTabId input first (set by resource wrappers for child dashboards), then falls back to Data.Configuration.tabId.
"Improve this" on a blocked (not-trustworthy-enough) prediction — turns the trust dead-end into a next step by opening the co-pilot seeded with the model's name + why it's held, so the agent proposes concrete ways to make it reliable and can rebuild it.
ProtectedloadResolve the provider-scoped engine, then notify the shell. Always calls NotifyLoadComplete — even on failure — so the loading screen can never hang on a direct-URL section load.
"+ New prediction" — open the co-pilot to a clean chat (Sonar-style); the user describes their goal and the agent drives the build.
ProtectedNotifyAsk the host shell to close/dismiss this resource (typically: close the tab). Called by subclasses that hosted a form that emitted a 'dismiss' navigation event — most often, a brand-new record where the user clicked Discard. The record was never saved, the form is empty, and leaving it open serves no purpose. The tab-container listens for this and closes the workspace tab.
ProtectedNotifyCall this to notify the tab system that the resource's display name has changed. The tab container will update the tab title and browser title accordingly.
ProtectedNotifyProtectedNotifyThe chat-area created its backing conversation after the user's first message — capture the live conversation, re-feed the pending message in the same change-detection cycle, and leave new-mode so the thread renders (mirrors the Form Builder co-pilot's atomic state-flip).
The chat-area delivered the seeded prompt — clear the buffer so a re-render doesn't resend it.
ProtectedOnCalled by the framework when query params change from an external source (browser back/forward, deep link navigation). Override in subclasses to react to query param changes.
The new query params from the URL
'popstate' for back/forward, 'deeplink' for external URL entry
Open the "+ New prediction" co-pilot (clean chat) — used by the read-only OpenNewPredictionCopilot agent tool.
ProtectedpublishReport the section's state to the AI agent (async chat + realtime co-agent share this snapshot). Subclasses may override extraAgentContext to add section-specific fields. No client tools are registered — Predictive Studio surfaces are read/navigate-only to the agent; model creation, training, and promotion run through PS's own dialogs and the Model Development Agent.
ProtectedRefreshMarks this component's view (and its ancestors) as needing change detection and schedules a tick — the reliable way for a resource component to re-render after updating state from async work (RunView/RunQuery continuations, timers, websocket pushes, RxJS subscriptions).
WHY THIS EXISTS: the Explorer shell hosts resource components dynamically — createComponent()
ApplicationRef.attachView() — because tab content mounts into Golden Layout / cached DOM
containers, not into an Angular template. Since Angular 18's change-detection scheduler
rework (Explorer is on Angular 21), ApplicationRef.tick() only refreshes attached views
that are FLAGGED dirty; a root-level attached view whose component mutates plain fields from
an async continuation is never flagged, so the view silently never re-renders — even though
the component uses default (non-OnPush) change detection (issue #3106).The framework calls this automatically from NotifyLoadStarted() / NotifyLoadComplete(), so the standard load lifecycle re-renders without any subclass action. Call it yourself after any LATER async state change that must reach the DOM outside those signals.
Protectedregister🔒 Read/navigate-only agent tools for the Predictions door: open a (trust-cleared) prediction, go back to the catalog, review the call list, export the at-risk list (CSV — read-only, mirrors Data Explorer's ExportView), and open the "+ New prediction" co-pilot. DELIBERATELY NOT exposed: "Save scores to records", "Send to a list" (both WRITE records), and any train/publish/delete — those stay behind the user's clicks and the approve-gated builder.
ProtectedResourceRetry loading the open prediction's at-risk list after a failure — bound to the at-risk error banner.
Retry the engine load after a failure — bound to the error banner's "Try again" button.
"Review the call list" — scroll the ranked at-risk list into view.
P1 #4 — "Send to a list" as a one-click server action (no agent sentence to parse): create an MJ List over the model's target entity and add the at-risk members (high/medium band, capped) so a team can act on it immediately (outreach campaign, tasks, etc.).
ProtectedUpdatePush query param changes to the URL. Creates a browser history entry. Safe to call during OnQueryParamsChanged — auto-suppressed to prevent loops.
Predictions — the business-user home for Predictive Studio (
option-b-refinedmockup). Reframes published ML models as plain-language "predictions" in a catalog, each with a trust badge; opening one goes to a workspace whose trust verdict GATES the actions (a Poor/unmeasured prediction can't be acted on). "+ New prediction" opens the Model Development Agent (the deterministic builder) as a docked co-pilot. Zero ML jargon — the analyst surfaces stay under the other nav items as "Advanced".