shixiliya1/dsh-rich-file-reader
Local image, Office, and PDF reader plugin for DeepSeek Harness
DSH Rich File Reader provides unified read_rich_file and ocr_pdf tools for DeepSeek Harness and adds an Import document button in the DSH Web composer. Supported formats: images (PNG/JPG/JPEG/WebP/GIF — passed to the model via DSH's official attachment flow), Word (.doc/.docx/.docm/.dotx/.dotm), Excel (.xls/.xlsx/.xlsm/.xltx/.xltm), PowerPoint (.ppt/.pptx/.pptm/.potx/.potm/.ppsx/.ppsm), PDF (text layers page-by-page, scanned pages via local OCR), and forced OCR (ocr_pdf for scans with incomplete text layers). The Import document button opens the browser's native file picker (Office + PDF), extracts up to 24,000 characters in local memory, and appends detected content to the current draft — no server copy, avoids the 403 issue with host.pickDirectory in Edge. All processing is local: modern Office files are unpacked locally, legacy binary Office uses the bundled office-oxide parser, PDFs use local PDF.js, and OCR uses bundled Tesseract.js.
Install
dsh plugin --profile web add github:shixiliya1/dsh-rich-file-readerThe README documents two install forms: a release tarball ('pnpm dsh plugin --profile web add https://github.com/shixiliya1/dsh-rich-file-reader/releases/download/...') and a git form ('pnpm dsh plugin --profile web add github:shixiliya1/dsh-rich-file-reader'). No npm package named dsh-rich-file-reader exists on the public registry as of 2026-10-01; the README explicitly states 'dsh-v0.1.3-alpha.1 is not published to npm'. Chinese or other language OCR data can be provided as a Tesseract language key (e.g. chi_sim) when calling ocr_pdf; non-bundled language data is downloaded by Tesseract on first use (document is never uploaded).
Compatibility
DSH dsh-v0.1.3-alpha.1 from source tag; Node.js ^22.19.0 || >=24.0.0. No Python, Microsoft Office, or API key required. No document content is uploaded to any third-party service.
Details
- Repo: shixiliya1/dsh-rich-file-reader
- Category: Files & Data
- Stars: 0
- Version: GitHub repo shixiliya1/dsh-rich-file-reader (no public npm registry entry as of 2026-10-01); MIT
- Last push: 2026-08-15
- First seen: 2026-08-14
Recent updates
The README documents the full format table, the Import button behaviour (24k char limit, no empty-page placeholders, Edge 403 avoidance), the offline processing approach, the Tesseract language key for non-English OCR, and the compatibility constraints (specific DSH source tag, Node version).
FAQ
- How do I install DSH Rich File Reader?
- From GitHub: dsh plugin --profile web add github:shixiliya1/dsh-rich-file-reader; or use the release tarball. No npm entry exists as of 2026-10-01.
- What file types can it read?
- Images (PNG/JPG/JPEG/WebP/GIF), Word, Excel, PowerPoint, and PDF (including scanned pages via local Tesseract.js OCR).
- Are my documents uploaded anywhere?
- No — the README says no document content is uploaded to a third-party service; all processing runs locally.
Alternatives
jiavenzhong/dsh-tool-markitdown · joejojoking-cloud/dsh-file-explorer · HongMing-Huang/dsh-file-upload