In #9099, we introduced the chat_generator init parameter in LLMMetadataExtractor as the primary way to configure the underlying LLM.
We also deprecated existing init parameters generator_api, generator_api_params.
In docs, we should show the new way to configure this component, to help users migrate.