A new standard dataset, Dataset Export History, is now available in NinjaCat. It surfaces export configurations and run history for dataset exports across an organization, including current status, schedule, destination, and per-run results — rows, bytes, files, errors, and timing. Use it to monitor export health, troubleshoot failed runs, and build reports or agent workflows on top of export activity without leaving the platform.

Dataset lineage now appears throughout NinjaCat. Open the Lineage tab on a dataset to see an interactive upstream/downstream graph, and check the new lineage health column on dataset and connector lists to spot issues at a glance. Transformation pages flag unhealthy source datasets directly where you build them.

When something looks off — or when a dataset has no lineage yet — click Reconcile Lineage from the dataset detail, the dataset list row menu, or a connector's datasets table. Reconciling re-walks the dataset's full upstream and downstream dependency chain, revalidates every node, and refreshes the health status everywhere lineage is shown.

The Forge API v1.14 budgets query now returns a cycle_active field that mirrors the runtime status shown on the Shape dashboard: 0 = inactive, 1 = active, 2 = pending. Add cycle_active to your query selection to use it — existing integrations that don't request the field see no change in their responses.

The description on the existing active field has also been corrected. active indicates only that the budget record exists (it isn't a runtime status), so integrations that need to know whether a budget is currently spending should read cycle_active instead. Older Forge versions (v1.6–v1.13) are unchanged; integrate against v1.14 to use the new field.

Webhook and scheduled triggers now appear as message bubbles in the assistant chat, with a clear visual hierarchy showing the trigger type, name, and timestamp. Webhook triggers display a human-readable message parsed from the incoming payload instead of raw data, making conversations easier to scan at a glance.

A new "Show Details" toggle reveals the full raw payload in a copyable code block when you need it, and long messages clamp with a fade so the conversation stays tidy. Webhook and time-based triggers now share a consistent style end-to-end.

AI assistant responses now appear with a progressive reveal animation — words, list items, and table rows surface gradually instead of dropping in as a wall of text all at once. The chat view auto-scrolls to follow new content as it appears, making longer answers easier to read and follow.

The animation respects the operating system's reduced-motion setting, and previously sent messages still load instantly when reopening a conversation.

When you create a new AI assistant, the "Choose a Starter Agent" library now includes 340 pre-built agents across 15 categories, so you can start from a template that's much closer to what you actually need. The library adds a new Paid Social Optimization category, and Paid Search has been renamed to Paid Search Optimization. To make the larger selection easy to navigate, the picker now supports search, category filtering, and pagination (50 agents per page, with a "Show All" option).

The transformation editor now protects your work when more than one person is editing the same SQL, View, or Data Cleaning transformation. If someone else has saved a newer version since you opened the editor, clicking Save will surface a warning showing who made the more recent change and when, and let you either cancel to review their update or explicitly overwrite it. This prevents the silent loss of edits that could occur when two people saved a transformation in quick succession.

The transformation version history view has also been cleaned up: each entry now shows its version number, and the "edited by" column displays just the user name without the email domain.

Admins can now control agent access at three levels instead of a single on/off toggle. The prior all-or-nothing "Agent Access" toggle has been replaced with three user-level toggles — Agents View/Chat, Agents Create/Edit, and Agents Super Admin — giving you more granular control at the User level before Agent-level permission rules are applied.

• Agents View/Chat lets a user browse the agent library and chat with agents they have access to, with no creation or editing controls visible.
• Agents Create/Edit adds the ability to build new agents and edit or duplicate ones they already have access to.
• Agents Super Admin extends Create/Edit with visibility into every agent (including private ones) and the ability to manage permissions across agents — intended for team managers.

User-level permissions act as a hard cap, so agent-level settings can never elevate a user above their assigned tier. This means admins can confidently roll out the agent library to broad audiences in View/Chat mode, give individual contributors the ability to build agents, and designate team managers as Super Admins — all without the previous all-or-nothing trade-off.

Migration note: With this release, any user who already had the prior "Agent Access" toggle enabled automatically has "Agents Create/Edit" access. Any user who had a System Admin role (with access to ALL permissions) automatically has all three toggles enabled — including Agents Super Admin. Please review and adjust these permissions as needed for your users.

Admins can now control agent access at three levels instead of a single on/off toggle. The prior all-or-nothing Agent Access toggle has been replaced with three user-level toggles — Agents View/Chat, Agents Create/Edit, and Agents Super Admin — giving you more granular control at the User level before Agent-level permission rules are applied.

• Agents View/Chat — Lets a user browse the agent library and chat with agents they have access to, with no creation or editing controls visible.
• Agents Create/Edit — Adds the ability to build new agents and edit or duplicate ones they already have access to.
• Agents Super Admin — Extends Create/Edit with visibility into every agent (including private ones) and the ability to manage permissions across agents. Intended for team managers.

User-level permissions act as a hard cap, so agent-level settings can never elevate a user above their assigned tier. This means admins can confidently roll out the agent library to broad audiences in View/Chat mode, give individual contributors the ability to build agents, and designate team managers as Super Admins — all without the previous all-or-nothing trade-off.

Migration note: With this release, any user who already had the prior Agent Access toggle enabled automatically has Agents Create/Edit access. Any user who had a System Admin role (with access to ALL permissions) automatically has all three toggles enabled — including Agents Super Admin. Please review and adjust these permissions as needed for your users.

Everyone can now see the Sync Now button, view current sync status, and check error details including last successful and last attempted sync timestamps.

What does Reporting Sync do? The Sync Now button pushes already-ingested dataset data into NinjaCat's reporting layer (that drives templates, dashboards, and report exports). It doesn't pull new data from the source — it makes data that's already been ingested visible in your reports.

When should you use it?

• New accounts or advertisers were added to the data source but aren't showing up in report templates
• Data looks correct in the dataset preview but isn't appearing in dashboards or reports
• A dataset sync just completed, but reports are still showing stale data

Status indicators at a glance:

• In Sync — data is visible in reports
• Syncing — a sync is in progress (button is temporarily disabled)
• Error — last sync failed; click View for details
• Never Synced — no RDW mapping configured yet (contact support)
• Unknown — no status data available

Note: Reporting Sync runs automatically after every scheduled dataset sync. The manual button is there for on-demand refreshes when something isn't showing up as expected.