I'm building an automation in n8n, which basically is an agent that generates answers in another language as if it were me, in its knowledge base in pinecone I add pictures of conversations I had with different people and how I interact with them, and I'm trying that for example I pass a screenshot of a person who speaks to me, and I need the agent to first translate what he says and then generate an answer as if it were me. The screenshots of conversations that I am using for his knowledge are in a google doc and that goes directly to pinecone. Any recommendations on how to improve this?