Hello, I've a question about how can the LLM prioritize llama index? So I can prevent hallucinations

Hi, I'm currently working with GPTPineconeIndex. I was wondering if it's possible to provide JSON files as a parameter to create new documents for an index?