Integrated Databricks AI data

After the synchronization, the assets are available in the specified domain.

Warning Do not move the assets to another domain. Doing so may lead to errors during future synchronizations.

Asset status

By default, the assets are shown in a plain list, but you can enable a multi-path hierarchy to show it in a tree structure.

The Default Asset Status field in the capability determines the status of synchronized assets.

  • If you select No Status, newly created assets receive the first status listed in your Operating Model statuses, and existing assets keep their assigned status.
  • If you select Implemented, all assets receive the "Implemented" status.
Note For more information about AI Model Version asset statuses, go to Configure the lifecycle stages of AI Governance assets.

Synchronized metadata per asset type

Important considerations:

  • We always integrate the latest version of a model.
  • The integrated attributes depend on training run type.
  • If you don't see the listed synchronized metadata, the attribute type may not be included in the out-of-the-box asset page layout. In that case, you can manually add the attribute type to the layout to make it visible on the asset page.
  • The Source Tags attribute is currently not ingested.

Asset type

Synchronized metadata

 Public ID
AI Base Model Description From Source System DescriptionFromSourceSystem
Initiating User in Source InitiatingUserInSource
Creation Date in Source CreationDateInSource
Modification Date in Source ModificationDateInSource
Owner in Source OwnerInSource
Source Tags SourceTags
 Databricks AI Model Version Description From Source System DescriptionFromSourceSystem
Version Version
Initiating User in Source InitiatingUserInSource
Creation Date in Source CreationDateInSource
Modification Date in Source ModificationDateInSource
Source Tags SourceTags
Model Accuracy ModelAccuracy
Model Precision ModelPrecision
Mean Squared Error MeanSquaredError
Mean Absolute Error MeanAbsoluteError

Any custom metrics defined via the configuration. If you use custom metrics, ensure that you add them to the assignment and layout on the asset type page.

  • accuracy_score,
  • exact_match
  • example_count
  • f1_score
  • f1_score_micro
  • f1_score_macro
  • false_negatives
  • false_positives
  • log_loss
  • max_error
  • mean_absolute_error
  • mean_absolute_percentage_error
  • mean_on_target
  • mean_squared_error
  • precision
  • precision_recall_auc
  • r2_score
  • recall
  • roc_auc
  • root_mean_squared_error
  • sum_on_target
  • token_count
  • true_negatives
  • true_positives
  
AI Model Deployment Description from source system DescriptionFromSourceSystem
Initiating User in Source InitiatingUserInSource
Creation Date in Source CreationDateInSource
Modification Date in Source ModificationDateInSource
Retirement Date in Source RetirementDateInSource
Implemented Content Filtering ImplementedContentFiltering
Compute Configuration ComputeConfiguration
AI Agent Description Description
Initiating User in Source InitiatingUserInSource
Description from source system DescriptionFromSourceSystem
Instructions Instructions
Creation Date in Source CreationDateInSource
Modification Date in Source ModificationDateInSource
Tool Usage ToolUsage

Any custom metrics defined via the configuration. If you use custom metrics, ensure that you add them to the assignment and layout on the asset type page.

  
AI Agent has version / is version of AI Agent Version AgentHasVersion
AI Agent Tool AI Agent Tool can consult data source / can be consulted via Inference Data AIAgentToolCanReadInferenceData
AI Agent Tool delegates to / can receive requests from AI Agent Version AIAgentToolCallsAIAgentVersion
AI Agent Tool delegates to / can receive requests from AI Endpoint AIAgentToolCallsAIAgentEndpoint
AI Agent Version Version Version
Initiating User in Source InitiatingUserInSource
AI Agent Version can call / can be called by AI Agent Tool AgentVersionCallsTool
Databricks Volume Volume ID VolumeId
Volume Type VolumeType
Catalog Name CatalogName
Schema Name SchemaName
File Location FileLocation
Owner in Source OwnerInSource
Description from source system DescriptionFromSourceSystem
AI Endpoint Access Method AccessMethod
Access Instructions AccessInstructions
Traffic Split TrafficSplit
AI Monitor Data Drift Detection Enabled DataDriftDetection
Prediction Drift Detection Enabled PredictionDriftDetection
Schedule Schedule
Alert Configuration AlertConfiguration
URL Url
File Description Description
URL Url
Document size DocumentSize
File Type FileType
Document creation date DocumentCreationDate
Document modification date DocumentModificationDate
Access Type AccessType
Storage Container Description Description
URL Url
Location Location
External System Label ExternalSystemLabel
Database Description from source system DescriptionFromSourceSystem
Owner in source OwnerInSource
Data Source Type DataSourceType
Schema

Description from source system

 DescriptionFromSourceSystem
Owner in source OwnerInSource
Data Source Type DataSourceType
Table

Description from source system

 DescriptionFromSourceSystem
Owner in source OwnerInSource
Column

Description from source system

 DescriptionFromSourceSystem
Column Position ColumnPosition
Is Nullable IsNullable
Is Primary Key IsPrimaryKey
Primary Key Name (if the column is the primary key) PrimaryKeyName
Original Name OriginalName

Technical Data Type

Tip 

You see the technical data type in the Technical Data Type field in the At a glance sidebar of the Column asset. If the At a glance sidebar is hidden, click Info icon. For columns that have a structured technical data type, Array or Struct, click the hyperlink to see the structure of the data in a dialog box. In other locations, for example, in Table assets, click the View Array or View Struct button to open the dialog box.



 TechnicalDataType

Synchronized metadata per agent type

The Databricks AI integration catalogs several types of Databricks AI agents: Knowledge Assistants, Supervisor Agents, Information Extraction tiles, Text Classification tiles, and Genie Agents. They expose different metadata to Databricks, so what is synchronized in Collibra Platform differs by type. If an agent is missing metadata you expected, it is likely due to one of these type-specific cases rather than a synchronization failure.

The following table summarizes what is synchronized for each type, grouped by category:

Synchronized metadata

Knowledge Assistant

Supervisor Agent

Genie Agent

Information Extraction

Text Classification
Agent structure:
AI Agent Version and AI Agent Tool Yes Yes Conditional Yes Yes
Traceability:
Tool links to a Table asset Yes No Yes No Yes
Tool links to a Volume asset Yes No No Yes Yes
Serving and monitoring:
AI Endpoint Yes Yes No Conditional Conditional
AI Monitor Yes Yes No Yes Yes
Basic agent information:
Instructions on AI Agent Yes Yes Conditional Conditional Conditional
Instructions on AI Agent Version No No No No Conditional
Creation Date in Source Yes Yes No Yes Yes
Initiating User in Source Yes Yes No Yes Yes
Modification Date in Source No No No Yes Yes

Conditions:

  • For Genie Agents, AI Agent Version and AI Agent Tool assets are created only after at least one queried table is already ingested as a Table asset. Until then, only the AI Agent asset appears.
  • The AI Endpoint asset connects to the AI Agent Version asset through an AI Agent Deployment complex relation for Knowledge Assistants, Supervisor Agents, and Information Extraction and Text Classification tiles. That relation is what surfaces the AI Monitor asset on the version. For Information Extraction and Text Classification tiles, this AI Endpoint asset is a shared placeholder, populated only when the tile has scheduled scorers, and not a real serving endpoint.
  • For Genie Agents, the Instructions attribute is populated only if the Genie space is configured with text instructions. For Information Extraction and Text Classification tiles, it's populated only if the task specification defines instructions. Text Classification tiles can also get Instructions on the AI Agent Version asset under that same condition.
  • Modification Date in Source is unavailable for Knowledge Assistants, Supervisor Agents, and Genie Agents because the source APIs for these agent types do not expose a last updated timestamp.

Operational trust metrics

For AI agents in Databricks, the Databricks AI integration also collects operational trust metrics from agent interaction data. This includes metrics on LLM judge passes and fails, and token consumption metrics.

  • Like attributes, these metrics are stored internally. However, unlike attributes, these aren't accessible through the Collibra Platform APIs.
  • These metrics power the Agents dashboard, the Quality tab on AI Agent assets, and the Operational Health theme on the AI Trust Score.
  • The metrics run on their own monitoring schedule, separate from and more frequently than the regular synchronization schedule. Since Databricks does not guarantee how quickly LLM judge results become available after an agent interaction, there may be a lag before these results appear in Collibra.

Support varies by agent type:

Agent type Operational trust metrics Token consumption metrics
Knowledge Assistant Judge pass, fail, and no-assessment counts Not available. Databricks doesn't expose token usage data for this agent type.
Supervisor Agent Judge pass, fail, and no-assessment counts Input and output token counts
Information Extraction Not available Not available
Text Classification Not available Not available
Genie Agent Not available Not available

To explore these metrics, in Settings, under AI Command Center, switch on Operational trust for AI agents.

For more information, go to Operational trust for AI agents and AI monitoring for AI agents.