Configuring resource path for your custom embedding model

  • Release version: Zurich
  • Updated November 10, 2025
  • 1 minute to read

    • The bring your own model (BYOM) configuration uses the shared alias from Generative AI Controller to connect the LLM with your custom embedding model. To avoid configuration issues, ensure the connection URL is set separately and any additional paths are correctly set in the resource path.

    Review the guidelines for configuring the resource path for custom embedding models in a BYOM setup.

    Configuration syntax

    The resource_path should include any additional URL paths beyond the base URL specified in the Connection URL. The general syntax is:
    /<API Version>/<specific-path-to-resource>
    Base URL vs Resource Path
    • The Connection URL in the alias contains only the base URL. For example, https://example.com.
      Note:
      Do not add any additional path, such as (/api/v1/model) in the connection URL, as it may break the LLM configuration.
    • Specify any additional paths in the resource_path value.

    Example Configuration

    For a Google Gemini model, the configuration consists:
    • Connection URL: 
      https://us-central1-aiplatform.googleapis.com
    • resource_path: 
      /v1/projects/sn-63151000.63151000-byom-poc-x6jx/locations/us-central1/publishers/google/models/gemini-embedding-001:predict
    Parameters and Variables of resource path includes:
    API Version
    Specify the version of the API you are using, for example, /v1).
    Specific Path to Resource
    Include the detailed path to the specific resource or model endpoint.