C3 AI Documentation Home

Configure Data Models from Agent Workbench

Use this guide to configure a data model for any instances Dynamic Agent in the Agent Workbench. A well-configured data model helps your agent run structured queries against the right types and fields while keeping prompt context focused and accurate.

What a data model is

In Agent Workbench, a data model is a structured representation of the application schema that you would like to have on your Dynamic Agent for effective querying. It identifies which C3 Types and fields are in scope for the agent, and can include:

  • Type and field hierarchy
  • Type documentation
  • Optional example values for selected fields

This representation is used to generate Data Model Documentation that is injected into the agent context for structured query tasks.

What a data model is used for

Data models help Dynamic agents perform structured retrieval and planning tasks with better reliability.

Use a data model to:

  • Limit agent access to relevant schema instead of the entire application model
  • Improve structured query quality by providing clear type and field context
  • Reduce noise and token usage by excluding unrelated types and fields
  • Keep generated documentation aligned with selected data sources

Before you begin

  • You have created a Dynamic agent draft and opened it in Agent Workbench.
  • You can edit agent configurations in your environment.
  • Your application contains the C3 Types you want the agent to query.

Create a data model

  1. In the Data Model section, select Create new data model.
  2. Enter a unique Data Model Name.
  3. Optional: Select an existing data model as a template. This populates the new data model with the fields and types from the template so you don't have to start from scratch.
  4. Select Create.

After creation, the workbench displays your data model name and enables data model configuration fields.

Select data sources and fields

  1. In Data Sources, add one or more C3 Types.
  2. Then, on the right of the Data Sources tab, select Select Fields.
  3. In the fields modal, select the fields you want the agent to query.
  4. Select Save.

The selected types and fields define the schema slice your agent can use for structured queries.

Configure documentation generation options

  1. Set Include Type Documentation to Yes or No.
  2. Optional: Enter Number of Examples.
  3. Select Generate Documentation.

The system generates Data Model Documentation from your selected data sources and options.

Review and edit Data Model Documentation

After generation, review the documentation in Data Model Documentation.

  • You can edit generated text manually for clarity and project-specific guidance.
  • If you regenerate documentation later, your manual edits can be overwritten.

Resolve out-of-sync warnings

If you modify data sources after generating documentation, the workbench can block key actions until documentation is regenerated.

You may see this warning when you:

  • Save configuration
  • Finish configuration
  • Chat with the agent in preview

To proceed, regenerate Data Model Documentation so it matches current data sources. If you made any manual changes to the documentation make sure you copy them elsewhere so you can paste them back in after the updated documentation generation otherwise they will be overwritten.

Remove a data model

  1. In Data Model, select Remove data model.
  2. Confirm the removal.

This removes the data model association from the agent, deletes the data model from the application, and returns the section to its initial create state.

Read-only behavior

When an agent is no longer editable in workbench state, data model controls become read-only.

  • Field selection switches to View Fields
  • Data model editing and documentation generation are disabled

See also

Was this page helpful?