In visual knowledge processing tasks, certain AI operators utilize large models to implement corresponding data processing functions. The range of large models available for operator selection must be pre-added by the enterprise administrator in the console. This document describes how to add the large models required by operators.

It is recommended that the enterprise administrator maintain the large model invocation settings. If the enterprise administrator needs to designate another user to maintain the large model invocation settings, the user must first be granted Visual modeling, Dataset Advanced configuration edit and view permissions. For detailed authorization instructions, please refer to Manage by content.

1. Log in to the Marketing Agent console.

2. In the left navigation bar, select Enterprise Knowledge Engine.

3. In the left navigation bar of the Enterprise Knowledge Engine, select Project Center > Knowledge Processing Configuration.

4. In the left navigation bar of Knowledge Processing Configuration, select Advanced Settings > Large Model Invocation Settings.

5. On the Large Model Invocation Settings page, select the operator for which you want to add a large model.
   

6. Click Add Model on the right side of the page.
   

7. Complete the configuration items in the pop-up window and click Confirm.
   The required configuration varies by operator. Please configure according to the actual interface information.

   Configuration Scenarios	Configuration Item Description
   	Model Name: Custom name to identify the current model. Model Invocation Address: The invocation address for the application. Inference Endpoint: The typical format is ep-xxxxxxxxxx-yyyy. For creation instructions, refer to Create Custom Inference Endpoint. API Key: The typical format is xxxxxxxx-yyyy-xxxx-yyyy-xxxxxxxxxxxx. For acquisition instructions, refer to Obtain and Configure API Key.
   	Model Name: Custom name to identify the current model. AppId: The App ID under Doubao Voice Service. For information on how to obtain it, see How to Obtain Parameters such as App, Token, etc.. Access Token: The Access Token under Doubao Voice Service. For information on how to obtain it, see How to Obtain Parameters such as App, Token, etc.. Resource Information ID: Doubao Audio Document Model 1.0: volc.bigasr.auc Doubao Audio Document Model 2.0: volc.seedasr.auc
   	Model Name: Custom name to identify the current model. Access Key ID and Secret Access Key: Access credentials (AK/SK) for a BytePlus account with Image Recognition and Understanding Service enabled. For operation instructions, see API Access Key Management.

8. Verify whether the configuration is effective.

   1. After adding a model, the operator's model list supports viewing configuration information, as well as Edit or Delete operations.

      warning

      Once a model is Delete, it cannot be recovered. Please proceed with caution.

      
   2. Create a test task to preview the models available for selection in the corresponding operator configuration.
      For instructions on creating a task, please refer to Step 1: Create a New Task.