Reltio AgentFlow™ Quality overview
Learn about Reltio AgentFlow™ Quality and how it helps you assess and prepare Databricks source data for Reltio.
Reltio AgentFlow™ Quality connects to a Databricks source, assesses the quality of a selected table, and publishes validated data to Reltio. Based on the table's structure and a sample of its data, AgentFlow Quality suggests data quality rules tailored to the table's content. It validates the selected rules against the full table, and if the results are acceptable, you can confirm the publish to Reltio. AgentFlow Quality maps the table to your Reltio tenant schema and loads the data using Data Loader. The entire workflow takes place within a single conversation in the AgentFlow Quality workspace.
AgentFlow Quality is designed for the following user roles who prepare warehouse data for Reltio:
To use AgentFlow Quality, you need the following system roles:
-
ROLE_AGENT_FLOW_QUALITY -
ROLE_EXECUTE_AGENTS -
ROLE_EXECUTE_MCP -
ROLE_DATALOADER
Key capabilities
AgentFlow Quality supports the following capabilities within a single guided conversation:
-
Connects to a Databricks source.
Suggests data quality rules, such as uniqueness or format checks, based on the table's structure and a sample of its data.
-
Runs the selected rules against the full table and returns pass/fail results for each rule.
-
Answers follow-up questions about failed checks and suggests ways to resolve the issues.
-
Maps the source table to your Reltio tenant schema and publishes validated data to Reltio using Data Loader.
When to use AgentFlow Quality
Use AgentFlow Quality in the following scenarios:
-
Assess a Databricks table against data quality rules before loading it into Reltio, instead of discovering issues after a failed or incomplete load.
-
Get data quality rule suggestions based on your table data instead of drafting it manually.
-
Investigate why specific rows or columns failed a quality check and troubleshoot the issues using natural language.
-
Run the assessment and publish workflow in a single conversation instead of switching between multiple configuration screens.