Understanding Nested Chains and Metrics in LLM Evaluation

Another question In the LLM Evaluation notebook, I've noticed that Llama Index(?) has the retrieval chain and the synthesis chain both under the same umbrella, making it easy to see the RAG flow, as well as having the e2e metrics visible on the top of the chain and the retrieval related ones on the inside retrieval chain, under the main chain. How can you make sure that chains are going to be nested? Also, does that also work in the same way for Langchain? How can I show different metrics depending on the layer of the trace?