I am interested in learning about the capabilities of LangChain as an extension of Large Language Models (LLMs). Specifically in the capability to create customized side effects from a conversation as well as configure embeddings for long term memory.