I have two sets of data. One is standard rules and regulation and another is customer data. I want to combine information from the two knowledge bases to answer user query. What would be an efficient way to combine data from the two sources?
Should I go for rag based retrieval from both the sources?
Should I create a knowledge graph for the two sources and do the retrieval?
If you have any other ideas to work with this type of data it will be helpful as well.
Honestly, I would just do RAG for the rules and regulations (or perhaps even as a system instruction), and use the Knowledge graph for customer data.
I’m kinda torn whether or not a knowledge graph might be overkill, but it all depends on what that data is and whether or not the connections between the data matters. If the connections don’t matter, and its essentially storing a giant list, there are other vector databases out there that would be a better fit, so you’re better off with regular RAG.