Release 9.2.39.2¶

Tag your assets at collection time¶

• When setting up a collector, you can now add your own tags — such as an environment name or any other detail — to the assets it collects.
• These tags make it easy to distinguish assets by their source or purpose, for example when the same asset name exists across multiple environments as it is promoted from test to production.

Choose how the collector authenticates to Databricks¶

The collector now offers an Authentication method selector for Databricks, so you can pick the credential type that best fits your organization's security practices:

• Personal access token
• Service principal (OAuth M2M)
• Entra ID service principal (Azure only)
• Entra ID workload identity (Azure AKS only)

Whichever credential you configure, discovery, monitoring, and identity resolution all use it consistently.

Collect Langfuse traces from cloud object storage¶

• The collector can now ingest Langfuse traces exported to cloud object storage (Amazon S3).
• These traces flow into the same collection, classification, and policy-evaluation pipeline as your other sources, so activity captured through Langfuse is governed alongside everything else — without needing a live connection to Langfuse.
• Flexible, secure cloud-storage credential options are supported to match your environment.

Filter assets by tag on the AI Assets page¶

• On the AI Assets page, a new tag filter lets you narrow the view to just the assets that carry the tags you care about.
• Combined with the tags added during collection, this makes it quick to focus on a specific environment, source, or group of assets.