Configure an LLM provider integration

Connect the automation orchestrator to an OpenAI API-compatible large language model (LLM) endpoint so that agent nodes can use LLMs in workflows.

Before you begin

  • You must have an administrator role in the automation orchestrator. For more information about roles, see Role-based access control.
  • Your LLM endpoint must implement the OpenAI v1 API.
    Note:

    Compatible providers include OpenAI, Red Hat AI (through OpenAI-compatible endpoint), and other OpenAI API-compatible services.

  • You must have the following information from your LLM provider:
    • Base URL
    • A connection credential for health checks
  • The LLM endpoint must be accessible from the automation orchestrator network.

About this task

The Configure integration wizard guides you through three steps: integration details, connection credential, and model enablement.

Procedure

  1. Complete the integration details.
    1. Log in to the automation orchestrator.
    2. Navigate to Configuration > Integrations.
    3. Click Configure integration.
    4. From the Integration type list, select LLM Provider.
    5. Enter the integration details.
      Field Description
      Name A display name for this integration. Use a descriptive name that identifies the provider and purpose, such as Red Hat AI - Granite or OpenAI Production.
      Description An optional description of the integration and its purpose.
      Provider type Select the LLM provider from the list (for example, Red Hat AI).
      Base URL The base URL for the LLM API endpoint. This must be an OpenAI API-compatible endpoint (for example, https://api.example.com/v1).
      Scope When Global is enabled, this integration is available to all projects. Disable to scope the integration to specific projects.
    6. Expand the Security section and configure the following optional fields.
      Field Description
      Allow HTTP connections Permits unencrypted HTTP URLs for this integration.
      Disable TLS certificate verification Skips validation of the server's TLS certificate on connections.
      CA certificate PEM-encoded CA certificate to trust for this integration's TLS connections.
    7. Click Next.
  2. Configure the connection credential.
    1. Select a Health check credential from the list. This credential is used to verify the connection to the integration and perform periodic health checks.
      Note:

      This credential is not the credential used by the integration at runtime in a workflow. Workflow credentials are configured separately in the workflow builder.

    2. Click Test connection to verify that the endpoint is reachable and to discover available resources.
      • If the test succeeds, a confirmation message is displayed.
      • If the test fails, review the error message and verify your Base URL and health check credential. For more information, see Test and monitor integration health.
    3. Click Next.
  3. Enable models.
    1. Review the models discovered from the LLM provider endpoint. All discovered models are enabled by default.
    2. Clear the checkbox for any model that you do not want available for use in workflows.
    3. Optional: Select a model to set as the default for this integration.
    4. Click Save.

      The integration is created with an enabled status. You can disable the integration or individual models later from the Integrations list or detail page.

Results

On the Integrations list page, verify that the new integration shows an Available status and the expected number of enabled resources.

Note:

The LLM provider integration is available for workflow designers to select when configuring agent nodes. Workflow designers can only select integrations that are globally scoped or project-scoped to their current project. After a workflow designer assigns this integration to an agent node, that node uses the configured LLM endpoint for AI-powered reasoning at runtime.