Skip to main content

AI Agents & Automations

This is one of the most disruptive changes in Collate 2.0. The three agent applications are deleted during the upgrade and replaced by AI Automations. The migration preserves your configuration, but the replacements arrive switched off.

The agent apps are removed

Breaking. Affects every deployment that installed the Documentation, Tier or Quality agent.
These three applications no longer exist: Their agentType values (CollateAI, CollateAITierAgent, CollateAIQualityAgent) are removed from the schema, their marketplace definitions are deleted, and their Quartz jobs are removed.
Only Metadata remains as an agentType. Automation that filters ingestion pipelines by agentType should move to the new values metadata and application.

What the migration does

The 2.0.0 migration runs these steps in order:
1

Strip the dead agentType

The retired agentType values are removed from the schema enum, so a persisted app row carrying one would fail strict deserialization before any field is read. The migration clears the value first (an update, not a delete), so the apps survive long enough to be converted.
2

Seed the Data Access Request workflow

DataAccessRequestTaskWorkflow is seeded so Data Access Requests can route through governance workflows.
3

Reseed the AutoPilot workflow

AutoPilot’s legacy runAppTask nodes (which called the three agent apps) are replaced with createAndRunAIAutomationTask nodes. Tier now runs before the profiler. Documentation and Quality run after it.
4

Convert each app into an AI Automation

Each installed agent app is deep-copied from its matching packaged template into a real automation that carries the app’s Elasticsearch filter, links the template’s agent, and is named with a Migrated suffix and a (migrated) display suffix. Provider is set to USER so it surfaces in your review list.
5

Delete the apps

The apps are hard-deleted. Each owns its bot, so the bot cascades away with it, and the backing ingestion pipeline is undeployed from the runner rather than orphaned. Marketplace definitions are removed separately, because deleting an app doesn’t remove its definition.
6

Repair AI charts

Multi-parent AI charts are split, and chart and dashboard data is repaired for the 1.13 → 2.0 shape.
Each step is fault-isolated and idempotent. A failure on one app is logged and doesn’t block startup or the remaining steps.

What you must do after upgrading

The migrated automations are created disabled and unscheduled. Your Documentation, Tier and Quality agents stop running at the upgrade and stay stopped until you turn them back on.
1

Open AI Automations

In AI Mode, go to AI Automations. Look for entries whose display name ends in (migrated).
2

Review the carried-over filter

The automation carries the asset filter from the app it replaced. Confirm the scope is still what you want. Asset counts are shown in the wizard.
3

Set a schedule and enable it

The migration deliberately doesn’t schedule the automation. Pick a cadence and toggle it on.
4

Add destinations if you had notifications

The migration never sets destinations. If the original agent notified a channel, reattach it.
Fresh 2.0 installs never had the agent apps, so none of this applies. The templates are simply available in the AI Automations catalog.

The agent app bots are removed with them

Breaking for API and SDK clients. Each removed app owned its bot through a containment relationship, so the recursive delete removes the bot too. CollateAIApplicationBot (which was attached to the Documentation agent) no longer exists after the upgrade, along with the Tier and Quality agent bots.
If any script, SDK client or integration authenticates with a token issued to CollateAIApplicationBot, it will start failing authorization after the upgrade. Reissue those tokens against a bot that still exists before you upgrade. Collate AI impersonation is handled by the automation bot going forward.
The AI Automations listing in Collate 2.0 AI Mode

AI Automations: what you get instead

AI Automations are a first-class entity at /v1/ai/automations with a considerably larger surface than the apps they replace.
Run records are mapped onto the same shape the app run history used, so dashboards built on agent run records keep working, but they must read from the automations endpoints, not GET /v1/collate/apps/name/{name}/agentRuns, which is removed.

Abilities are now Skills

Breaking. The /v1/agents/abilities namespace is renamed to /v1/agents/skills.
All fifteen endpoints move:
The request and response shapes are unchanged: this is a path rename. Update the base path in any client and you are done.

Knowledge Center and Context Memory moved into the platform

Breaking. Knowledge Center is removed because the capability moved into the platform’s own Context Center. Context Memory keeps its paths but is reimplemented by the platform, dropping three operations.
The Knowledge Center UI is removed and replaced by Context Center, which adds articles, documents, uploaded files, memories and an archive. Memory storage is delegated to the AI platform via a user memory context, while the entity itself is now a platform entity.
The memory endpoints (/v1/contextCenter/memories/**, 10 endpoints) keep their paths but are served by the platform implementation, and three operations are gone: check-duplicates, promote-chat and remember. Rework any client that called them.
Collate 2.0 Context Center showing articles, documents, uploaded files, memories and the archive

Context Plugins

New. /v1/ai/contextPlugins lets you connect external knowledge sources so the assistant can read from them. Notion, Slack, GitHub and Google Drive ship seeded, and each plugin has a testConnection endpoint.

Context memories enforce visibility

Behavioural. Memories are now visibility-scoped in both AI answers and semantic search. You see org-wide memories, memories you own, and memories shared with you or your team, not everything.
No action required. Search indexing applies the visibility automatically once the instance is upgraded. If a memory seems to have disappeared from results, check who it is shared with.

AI Studio labelling follows the Skills rename

Behavioural. Because abilities became skills across the backend and UI, AI Studio’s terminology, labels and navigation change to match. Enablement material and screenshots that say “abilities” need updating.

AI Analytics: dashboards and charts

Collate 2.0 AI Analytics showing AI Dashboards and AICharts

AI Dashboards now enforce visibility

Breaking behaviour. Dashboard visibility (Private, Shared, Public) was stored in 1.13 but never applied, so any user with view rights could open any dashboard. 2.0 enforces it, and new dashboards default to Private.
To preserve existing access, the upgrade sets every existing dashboard to Public. That is deliberately permissive. It keeps 1.13 behaviour on day one rather than silently hiding content.
Owners should review their dashboards after upgrading and restrict the ones that should not be org-wide. Until they do, every dashboard remains visible to everyone who can reach AI Analytics.

An AI chart now belongs to exactly one dashboard

Breaking, and a security fix. In 1.13 a chart carried a list of dashboards and its read access was the union of them, so any user could read someone else’s chart by pinning it to a dashboard of their own. In 2.0 access comes from the single parent dashboard.
The create contract changes accordingly:
The query configuration is also split into a discriminated union:
On upgrade, charts pinned to several dashboards are split into one copy per dashboard, keeping title, owners and query but starting a fresh version history. Pinning to a second dashboard now creates an independent copy rather than sharing one chart.
Any client that posts dashboards: [...] or queryConfig must be rewritten. dashboard is now required at creation. A chart can no longer exist without a parent.

Other AI Analytics changes

Because scheduled refresh no longer runs through an ingestion pipeline, any monitoring you built on those pipelines (alerting on a failed AI dashboard refresh pipeline, for example) will see the pipelines disappear. Monitor the automation and dashboard endpoints instead.

AI Studio

Additive. AI Studio gains authoring for agents, personas and skills directly in the UI, including drafting a planner description for a dynamic agent. The UI-selected persona is forwarded to the AI platform so agent behaviour follows the persona.
AI Studio in Collate 2.0, where agents, personas and skills are authored

Agent invocation

Dynamic agents that pointed at the deleted agent-app bots are repointed by the migration.