Be able to override the configuration in prompt API - like change the used model version, max_tokens, etc