Process one or more PDFs into a traceable analysis entry. The workflow first classifies the request, preserves file/page evidence, synthesizes across documents when needed, and stores a structured memory index.
LLM should manually run pdf-toolkit scripts then summarize and
memory_save.
