Nico0713520/dsh-doc-to-markdown

README 摘要

dsh-doc-to-markdown Convert PDF & Word documents to clean, LLM-friendly Markdown. Deterministic rules — no AI calls, no tokens, no cost. 中文用户请看 下方速览。 Why this one markitdown raw pymupdf4llm dsh-doc-to-markdown Chinese PDF font glitches ( ⼯ → 工 , � ) ❌ ❌ ✅ NFKC normalization + rescue Windows Chinese/space paths ⚠️ ⚠️ argv encoding bugs ✅ temp-file arg passing Scanned PDF detection ❌ silent garbage ❌ ✅ explicit "needs OCR" report Path traversal safety ❌ n/a ✅ built-in guard Agent-friendly output ❌ human text ❌ ✅ JSON stdout contract DOCX without Python ❌ needs Python n/a ✅ pure Node (mammoth) One-line pitch: it's the converter that survives real-world Chinese PDFs and Windows paths, and speaks JSON to your agent. Demo Input: a real-world Chinese PDF (exported notes with tables) Output Markdown (actual conversion, see examples/): Headings, tables, bold, lists — all preserved. Zero garbled characters. Quick start JSON result contract (parse this in your agent): Failure is also structured — show error to the user, never retry blindly: Engine routing Format Engine Fallback .docx mammoth + turndown (pure Node) — .pdf (text layer) PyMuPDF4LLM via Python subprocess clear setup hint if Pytho…

分类