Verification-first conversion of PDFs to clean, compilable LaTeX (.tex). Use when correctness matters — medical, scientific, legal, or regulatory contexts. Runs multiple extractors at every step (PyMuPDF, pdfplumber, pdfminer.six, OCR, pandoc), scores per-page agreement, and escalates to the user on divergence rather than silently picking a winner. Closes the loop with round-trip compile-and-diff plus numeric preservation checks. PDF only — Word documents are out of scope.