Multimodal RAG Implementation Prompt

Implement the multimodal RAG pipeline using LlamaIndex. Focus on handling various document types (text, PDF, simulated image descriptions). How will you generate and store embeddings for these diverse inputs, and how will your RAG retrieve context most relevant to a given query, potentially spanning multiple modalities?