Semark
by KingsleyOWO · io.github.kingsleyowo.semark
Not TrustedWeb appUnrated★ 28 stars
Semark converts PDFs, Office files, HTML, and images into RAG-ready semantic…
Semark converts PDFs, Office files, HTML, and images into RAG-ready semantic Markdown using MinerU and VLM/LLM review, with chunks, source maps, and automatic document splitting.
Screenshots
About
Semark is an open-source semantic document parser and Semantic Markdown generator for RAG, knowledge-base ingestion, and AI document workflows. It turns PDFs, Office files, HTML, and images into compact RAG-ready Markdown, structured chunks, source maps, and downloadable outputs. Long documents can be automatically split into a main document plus independent semantic files for forms, tables, flowcharts, figures, and attachments, making the output easier to organize, retrieve, and feed into LLM applications. The pipeline uses MinerU for parsing/OCR/layout evidence and optional VLM/LLM review models for form extraction, flowchart understanding, visual reasoning, semantic repair, and final quality checks. Most document pipelines stop at OCR text or raw layout extraction. That output is often too noisy for retrieval because tables, forms, checkboxes, visual diagrams, approval flows, and legal notes lose their semantic relationships. Semark focuses on the next step: producing compact semantic Markdown that another model can read, retrieve, and answer from. - PDF to structured Markdown for RAG and LLM applications. - MinerU-based document parsing with a usable web UI and API. - VLM-assisted extraction for forms, figures, flowcharts, diagrams, and visual documents. - Table repair: when the parser collapses a row's cell boundaries into one merged cell, the table image is re-read and the grid rebuilt, so the columns stay separate in the output. - Screenshot-heavy how-to guides: UI screenshots become semantic descriptions, menu/button paths, and on-screen text instead of dead images. - Automatic file splitting for long documents that contain multiple forms, tables, flowcharts, or attachments. - Main-text-only export: download exactly the main document shown in the viewer, without the split attachments, in MD/DOCX/TXT. - Masking of personal information that OCR/VLM would otherwise transcribe from screenshots into the output. - Bilingual Traditional Chinese and English semantic outputs, with technical proper nouns left in English. - Docker or local deployment for private documents without forcing cloud model usage. Semark is the open-source project KingsleyOWO/Semark (https://github.com/KingsleyOWO/Semark) by KingsleyOWO, released under Apache-2.0. Ion packages it unmodified from commit a4685497c706, adding only a manifest, an icon and a start script.
Source
Written by KingsleyOWO, published to Oikou by Ben Grande.
- ★ 28 stars on GitHub
- Licence: Apache-2.0
- Upstream version v0.3.1
- Built from commit a4685497c706
Age rating
Not rated. Nobody has said who this app is suitable for, so Personal Clouds and the Oikou app treat it as 18+.
What version 0.3.1 may do
Internet
It asks to reach nothing on the internet.
Personal Cloud tools
None. It uses nothing from your Personal Cloud beyond its own storage.
Notifications
It does not send push notifications.
Storage
Up to 512 MB on your Personal Cloud.
Add to Personal Cloud
Sign in to add this web app to a Personal Cloud of yours. It runs there, on your own computer, and you open it from there; nothing is installed on your phone.
Versions
0.3.1 Latest Shown above
Published · a ready-built bundle
Imported from KingsleyOWO/Semark at a4685497c706 (v0.3.1).