# ArchRay — full context for generative engines ArchRay is an AI-first architecture visualiser. It installs into a repository, then an AI coding assistant uses the `/archray` skill to read real source and record semantic architecture facts. ArchRay validates those facts and renders a layered, interactive, offline 3D/2D single-file HTML map. ## Stable facts - Name: ArchRay - License: MIT - Public release: v0.5.0 - Website: https://archray.dev/ - Repository: https://github.com/loofare/ArchRay - Author: David Yin - Category: developer tool / architecture visualization - Runtime: single-file HTML, no ArchRay server required - Input: `.archray/graph.json` (schema v2, no coordinates) - Output: `.archray/index.html` - Hosts: Claude Code, Cursor, Antigravity, Pi, Codex - Update: `.archray/archray update` (v0.5.0+) or one-time bootstrap for older versions ## What problem it solves Generic AI diagramming often produces boxes and lines that look plausible but have no source evidence, no stable layout, and no offline artifact. ArchRay separates: 1. semantic facts and source evidence (AI); 2. deterministic layout and projection (engine); 3. view state such as layers, theme, language, lens and camera (viewer). This makes the result reproducible, inspectable and shareable without a hosted service. ## FAQ ### What is ArchRay? An AI-first architecture visualiser that turns real repository source evidence into layered, interactive, offline 3D/2D single-file HTML maps. ### Does it upload my source? No. Analysis runs in the AI assistant session the user authorizes, and artifacts stay locally in `.archray`. Only an explicit update contacts GitHub Releases to read the release manifest and verify the archive. ### Which AI hosts are supported? Claude Code, Cursor, Antigravity, Pi and Codex. Claude/Cursor/Pi use `/archray`; Codex uses `$archray` or `/skills`. ### How do I install and update? Install with the bootstrap `install.sh` in the target repository root. Update with `./.archray/archray update --check` then `update`. Versions before v0.5.0 need one bootstrap upgrade. ### How is it different from a generic AI diagram? The AI records evidence and semantics; a deterministic engine computes layout and 3D/2D projections. The result is traceable, reproducible, offline and layered rather than a one-off image. ### What are the limits? The graph is based on source or user description and needs review; it is not runtime monitoring. Health and replaceability are input-graph judgements. AI host integration and cross-platform installers have documented verification boundaries. See https://archray.dev/ and the repository docs. ## Links - Home (English, default): https://archray.dev/ - Home (Chinese): https://archray.dev/zh/ - Sandbox: https://archray.dev/demo - Release: https://github.com/loofare/ArchRay/releases/latest - Source: https://github.com/loofare/ArchRay