An example would be:
- AI consolidates a very long conversation to: “Tony would like to understand how question consolidation works and how examples are provided”
- RAG is performed on the text “Tony would like to understand how question consolidation works and how examples are provided” looking for fragments that are similar (for example it could be “question consolidation works by XYZ”)
- Then the chunks that are close to the question asked are added into the conversation as pre-existing context