Convert PDFs and documents to markdown, index them locally for RAG retrieval, and analyze them token-efficiently.