The RAG pipeline executes only after a PDF is uploaded. This avoids accidental indexing, ghost data, and interview red flags.