I can try to help you for the first query
Nodes is basically a fixed size collection of document information along with other details related to nodes like prev/next node infor, metadata etc
IndexNode works the same way but in addition to this it can have a reference to a indices, query engines, retrievers.
https://github.com/run-llama/llama_index/blob/cb8d1aa5ec767edc7f5be30e91122710eb08547d/llama_index/schema.py#L400Maybe having this reference helps the agent to query better.
For better explanation on the second question we will have to wait for Logan 😅
