Hey there, my LLM application uses Pinecone as RAG context to generate outputs. I was wondering if there's anyway to pass Pinecone embeddings as context in promptfoo so that the score I get gives a more accurate representation of my app's performance as opposed to directly putting text from my documents.