| # pdfsys-parser-vlm | |
| Vision-language model backend for scanned PDFs with complex content (tables, formulas). **Stub only — not yet implemented.** | |
| Will handle regions flagged as TABLE or FORMULA by the layout analyser, sending them to a VLM (MinerU 2.5 / PaddleOCR-VL) that can produce structured output (HTML tables, LaTeX formulas). Simple text regions in the same document may still be handled by the pipeline backend. | |