This code uses the implementation provided in the link above, and the result is not that good as the one I was getting using the basic implementation, as such, I wanted to add an agent to improve the result by analyzing the input based on the context of the question
When you say the result was not as good, what does that mean?
You probably will want to add extra descriptions for each table, in addition to the schema. These object classes are really new, I'd have to look up where that goes 😅 but I assume that will help improve the table retrieval.
You could also increase the top k slightly.
Although, you might be better off using the basic approach if all the tables schema fit into the LLM input, since then there's no need for the retrieval code
