Set up Data Governance Agents

Administrators configure models, DAM connectivity, user identity, agents, and templates so users can start conversations and create spaces. Users with an available agent can go directly to the quick start.

Configuration sequence

Order	Task	Verify
1	Configure a model and set the default	Connection test and sample response succeed
2	Configure global baseUrl	The agent reaches the correct DAM environment
3	Configure user identity	Calls use the current user's authorized identity
4	Import governance and asset query agents	Agents are listed with correct models and tools
5	Prepare saved prompts	Users can select prompts and replace example scope
6	Publish both template types	Governance and query templates are available to spaces

Configure a model

1. Obtain the protocol, base URL, model identifier, and credentials from the model service owner.
2. Open Configuration Management > Model Configuration.
3. Add the model using the service's parameters.
4. Test the connection and set the default model.
5. Verify responses with a small, non-sensitive sample before governance use.

The provider selects a communication protocol; do not infer it from the model brand alone. Enter keys in authentication fields. See model services for discovery, LLM and embedding parameters, and testing.

Set the DAM root URL

In Configuration Management > Parameter Configuration, add global parameter baseUrl with the DAM root URL accessible to the agent.

Parameter: baseUrl
Example value: https://dam.example.com

Replace the example with your confirmed environment address. Use the DAM root rather than the DIG page, a complete business page route, or another environment.

Configure user identity

Obtain a valid user token in DAM under Personal Workspace > Token Management. Configure the corresponding credentials in DIG's personal parameters under Parameter Configuration. Use the identity of the user executing the task.

When custom agents use official Skills, verify the runtime identity reference required by the Skill prompt: ${session.token}. This literal reference is resolved at runtime; it is not a real token to copy into examples.

Static personal credentials and runtime session parameters have different purposes. If authentication fails, check token validity, user permissions, and whether the agent actually uses the parameter. Keep real tokens out of conversations and screenshots.

Import agents

1. Open Configuration Management > Agents > Import Agent.
2. Import the governance or asset query agent package supplied by your implementation team.
3. Give it a name that describes its responsibility.
4. Review system prompt, model, MCP, Skills, and dependent tools.
5. Set a default agent if required for embedded conversations.
6. Start a test session to verify object queries and proposal generation before user release.

Check imported addresses and paths for environment differences. Successful import alone does not validate dependencies.

Prepare prompts and templates

Administrators can import saved prompts for Chinese business name enrichment, classification and security levels, standard creation, standards mapping, rule creation, and asset inventory. Users must replace example scope and constraints before submitting prompts from My Prompts.

Follow agent templates to publish a Data Governance Agent template and a Business Query Agent template. An available agent does not mean a template has been published.

Verify setup

• DAM users can open AI and use the intended default or selected agent.
• The agent can query a test object within the user's permissions.
• A small governance proposal can be reviewed and verified in DAM.
• A published governance template is available when creating a space.
• An appropriate business query template is available for the query phase.
• User identity, files, and instance settings belong to the intended user and environment.

Diagnose failures

Check license and permissions, model connection, baseUrl, user identity, current agent, then tool and Skill dependencies. A successful model connection only validates the model service.