Load and use LLMChain with either a specific prompt key or custom prompt.
from_llm(
cls,
llm: BaseLanguageModel,
base_embeddings: Embeddings,
prompt_key: str | None = None,
custom_prompt: BasePromptTemplate | None = None,
**kwargs: Any = {}
) -> HypotheticalDocumentEmbedder