Watsonx.ai settings

Configure the watsonx.ai settings to enable API Agent and other AI-powered features in IBM® API Connect. API Agent uses Granite and Llama foundation models on IBM Cloud.

About this task

Add your watsonx.ai credentials to enable the API Agent.

If you do not provide credentials, the IBM-managed credentials are used.

Within API Agent, watsonx.ai provides access to a language models. It is essential for enabling the agent to complete the tasks such as tool selection, planning, and summarization, and for powering tools that generate APIs and support design and development workflows.

Procedure

  1. Log in to API Manager.
  2. On the Settings page, select the AI tab.
  3. If you are configuring for the first time, click Add. If you previously configured, click Edit to modify the existing configuration.
    Note: For new Provider organizations created, the watsonx.ai configuration is enabled by default. You do not need to manually add a watsonx.ai configuration unless you want to modify the default settings.
  4. In the Add watsonx.ai configuration dialog, enter the details and then click Done:
    Note: You can enable watsonx.ai from the IBM Cloud and Software tabs.
    1. In the Name field, enter the configuration name.
    2. From the URL list, select the URL.
    3. In the Authentication field, enter an IBM Cloud API key or bearer token that has the permissions to use the foundation models that are provided by watsonx.ai.
    4. In the Environment field, enter your project ID or deployment space ID that is associated with your watsonx.ai service.
  5. Optional: To verify the connection, click Test.
  6. To use your watsonx.ai credentials as the default one for API Agent, select Set as default.
    Important:

    By default, both the Granite and Llama models are selected. Granite model is mandatory and cannot be deselected. You can deselect the Llama model if it is not required.

    When Llama models are enabled, API Agent supports extended capabilities such as adaptive questioning and clarification, dynamic plan updates, recursion, and improved conversational flow. However, some API Agent generators are fine-tuned for Granite and continue to perform best with Granite, regardless of the model that you select.