83079Vermont/dsh-mineru-parse-plugin

README excerpt

mineru-parse-dsh-tool English 简体中文 A self-contained DeepSeek Harness (DSH) plugin that registers a global model tool parse docs : parse local PDF / DOCX / PPTX / XLSX / TIFF and image batches into Markdown with local MinerU. Features - Self-contained — bundles its own wrapper script assets/scripts/parse.ps1 ; no skill package dependency - Zero runtime dependencies — Node builtins only - Global scope — mounted at the profile/host layer, available to every session - Batched + cached — multiple files in one MinerU call; fresh .md outputs are reused - Sandbox-aware (Windows) — fail-fast with a one-shot escalation hint when MinerU cannot run parse docs handles: pdf / docx / pptx / xlsx / tif / tiff / jpg / jpeg / png / bmp . Plain text ( txt / md / json / csv / code) should be read with the built-in read tool, and single images with read image — parse docs returns a hint pointing to the right tool. Requirements - DSH with a profile (tested on the web profile) - Windows + PowerShell (the wrapper script is .ps1 ) - Node.js = 18 - MinerU installed — the parser engine is not bundled. parse.ps1 locates it in this order: 1. Env var MINERU BAT (recommended) 2. mineru.config.json — in \assets\s…

Category