LLMManagerMixin

Run on new output token.

Only available when streaming is enabled.

For both chat models and non-chat models (legacy text completion LLMs).

Run when LLM ends running.

Run when LLM errors.