Using LangChain ReAct Agents for Answering Multi-hop Questions in RAG Systems