The browser is the workhorse of the system. It renders documents natively and sovereignly — your files never leave your hardware, and every viewer is a focused Nishi organ, not third-party code.
Live now — click to open inline
Word document
Headings, bold/italic, and tables rendered as a real page — with a Download .docx button.
Open the venue agreement →Markdown
Headings, lists, code blocks, quotes, and links — the format your READMEs and notes live in.
Open a markdown doc →Spreadsheet (CSV)
Tabular data as a clean, sortable-looking table with a sticky header row.
Open a budget sheet →PDF (text-layer)
Sovereign extractor (FlateDecode + per-font ToUnicode) → filtered, readable text preview of a real 634KB research paper.
Open a Cochrane review →Excel workbook
SpreadsheetML → chart + table; formulas compute (=SUM() evaluated sovereignly) and pivots group-by with SUM/COUNT/AVG.
Slide deck
PresentationML → 16:9 slide cards, inline. Authored, read, and rendered by nx_pptx (v1: titles + bullets).
Open the reunion deck →On the roadmap — one registry row each
& more
Calendars, structured data, images — each a small organ.
Also on NishiOS — the OS opens documents natively
The same capability runs on the sovereign NishiOS: boot the OS image to long mode on the sovereign emulator, and the OS app layer opens report.md through its on-OS viewer registry (renderers link as libraries — no process fork on the OS yet) into a NISHI DOCS desktop window. Boot-gated, 12/12 KATs with negative controls. See the boot-to-document proof →
Why it doesn't bloat the browser
The browser core never grows. A data-driven registry — knowledge/doc_viewers.conf — maps each file extension to a focused sovereign renderer organ. The dispatcher nx_doc_view detects the format, routes to the renderer, and hands the browser plain HTML it already knows how to paint. Supporting a new format is one registry row plus one small organ — the dispatcher and the browser never change.