Permissions and Guardrails
Why is a model call rejected even though authentication succeeds?
Successful authentication and model usage permission are separate things. Check your account permissions, whether the service is activated, the call path, and the usage conditions of the model·version. If a permission check is required, send the request identifier and error code to your account representative.
How do I request to have guardrails lifted?
Provide your account representative with the intended use, the call path, and the model and version, and ask whether it can be applied and what the approval process is. Changing parameters alone does not grant usage permission. Apply the settings from the relevant engine documentation within the approved account and scope.
Can I apply guardrail settings to another model as is?
The field location, data type, and supported range may differ. Do not mix the fields of VOD and MPS, and do not copy request values from another version just because they belong to the same family. Check the special settings in the engine documentation you use and the scope approved for your account.
What information should I provide when submitting a technical inquiry?
Provide the call path, Action, model·version, request time and time zone, request identifier, TaskId, and error code. Also include the parameters and file specifications needed to reproduce the issue. Remove SecretKey, API keys, signature values, and URLs that contain access permissions from public posts and general logs.