High-fidelity tool to convert PDF files and images into fully editable PowerPoint presentations (.pptx). Unlike basic converters, it uses AI-powered structure extraction (MinerU) to accurately reconstruct text, layouts, and images, ensuring the output is not just a screenshot but a professional, modifiable document.
Agent skill for reconstructing scanned documents into searchable, typeset PDFs with vector diagrams, provenance, and visual QA.
Agent skill for reconstructing scanned books with visible text, preserved layouts, illustrations, LaTeX math, and PDF hierarchy.
Tear PDF pages into non-overlapping fragments with exact ground truth, plus dm-bench: a torn-document reassembly benchmark with an evaluation harness and baseline solver.
Reconstruct clean, searchable documents from scrambled, degraded, bilingual scanned bundles — offline OCR core + optional Claude-vision smart layer, measured against synthetic ground truth.
Agent skill for reconstructing scanned exams and forms as editable Word documents with layout verification.
High-fidelity PDF to Word (.docx) reconstruction skill for AI agents -- rebuilds layout instead of transcoding it, with OCR+vision fusion for scanned PDFs and a render-diff verification loop.
The ability to edit scanned documents
Hybrid-retrieval CLI (vector + SQL + refusing local LLM) and whole-document reconstruction over a sovereign local archive. Part of A Mirror of My Becoming™ — Suite.
To associate your repository with the document-reconstruction topic, visit your repo's landing page and select "manage topics."