Optimization of large model websites

I am currently working on a website for a large model application and I have a feature that I want to optimize. This function specifically includes: a. The basic information input is simplified into: [Industry] (filled in by the user), [Recommended indicator categories] (users can add or delete them themselves), and [Core indicators] (filled in by the user, multiple are allowed) b. The returned mind map: the root node is [industry], the primary category is [recommended indicator category]+[core indicator], and the lower level nodes are the given data and the model's own judgment of drilling down results. c. The user inputs the analysis case enterprise, the system searches for relevant content based on the brain map structure, and generates an analysis report. The user can download the report. Note: This step can refer to calling the dify interface. What I want to solve is step c above. It's about searching for some metrics from Google and specific websites, and then generating reports for the big model.