Skip to content

Release 9.2.45.3

This release lets one runtime plane scan many Databricks workspaces and Microsoft tenants as separate connections, gives custom agents a collector of their own, and adds saved credentials so a secret is entered once and rotated in one place.

AI Assets Collector

Collector connections

Collector connections

  • Add each Databricks workspace, or each Microsoft tenant service, as its own connection from the new Databricks, Azure AI Foundry, Azure Copilot Studio and Azure M365 tiles.
  • A Microsoft connection reads only its own service, so a service principal with access to one service gets no errors for the others.
  • Choose Shared (the default) to run up to 10 connections in one pod, or Dedicated to give a connection its own pod. A runtime plane can have several shared hosts, each with its own scan schedule and settings.
  • Pause a Shared connection with its Enable/Disable toggle.
  • Each workspace and tenant service has one connection per runtime plane, so the same source is never scanned twice.
  • See Collector Connections.
Custom agents collector

Custom agents collector

  • The new Custom agents tile creates a collector of its own for agents you build yourself, with the Custom Agents, Trace Ingester and Monitoring sections. It always runs in its own pod.
  • Each Custom agents collector writes and reads its traces under its own prefix by default, so several can share one bucket.
  • See Collect Traces (OTLP).
Saved credentials

Saved credentials

  • A runtime plane's new Credentials tab stores Azure Entra service principals, Databricks tokens and service principals, and OpenAI and Anthropic API keys.
  • In a collector, choose Saved credential instead of typing a secret, or save the values you typed with Save this token?.
  • Editing a credential shows which collectors will restart before you save, then updates every one of them and shows each one's status, with Retry for any that were not updated.
  • Works with runtime planes that keep secrets in AWS Secrets Manager, Azure Key Vault, Google Secret Manager or a Kubernetes Secret.
  • See Saved Credentials and the Saved Credentials API.

AI Governance

Source account on asset details

Source account on asset details

  • An asset's details page shows the account the asset comes from: the Databricks account ID for a Databricks asset, or the Azure tenant ID for a Microsoft Copilot Studio, Azure AI Foundry or M365 agent.
  • When several workspaces or tenants are connected, this shows which one an asset belongs to.

Upgrade Notes

  • Connections, the Custom agents tile and saved credentials need a runtime plane on 9.2.45.3 or later, or with Auto Upgrade on.
  • Existing collectors are unchanged and nothing is migrated. They appear as AI Assets Collector (Legacy) and keep their own pod and full form.
  • To move a workspace into a connection, add the connection, then remove the workspace from the old collector. Do not leave a workspace in both: each would mark the other's assets inactive.

Known Issues

  • Two Azure AI Foundry connections on the same Foundry resource both report its model deployments. When one of them is deleted or loses access, those deployments can show as inactive even though the other connection still sees them, until a deployment changes.
  • The Credentials tab does not yet flag a credential whose secret was encrypted with a runtime plane key that has since been rotated. List the credentials through the API to find them (keyRevoked).