Skip to content
CAI
Software that uses CAICheck a score

tt-a1i/archify

39.0

Weak · 1 October 2026

47k

lines of production code

JavaScript

primary language

2

measurements over time

CAI band scale
CAI trend line
CAI lens gauges

What this system is

Archify is a command-line tool and ecosystem for authoring, validating, and delivering structured architecture diagrams from JSON definitions. It provides a unified interface for generating visual outputs with specific layout presets, while offering a self-contained viewer and automated pipelines for documentation and distribution. The system also includes benchmarking harnesses to evaluate diagram quality and agent usability, alongside integrations for external AI agents.

Features

Add Skill-only Hermes Agent integration for Archify

Users can now install Archify as a Skill within the Hermes Agent environment. This integration is strictly Skill-only: it registers the Archify SKILL.md so Hermes can invoke the existing Node.js renderer (node bin/archify.mjs) but does not add native Python render tools, telemetry, or background services. The plugin supports Linux, macOS, and Windows, allowing installation via the Hermes skills registry or local symlinks, and includes a contract test suite to verify the plugin structure and documentation.

integrations/hermes-agent · high confidence

Archify 2.0 initial release with multi-diagram support and verified delivery

Archify 2.0 introduces a unified CLI for authoring and delivering architecture, workflow, sequence, data-flow, and lifecycle diagrams. The release adds a built-in brand-mark catalogue (107 icons sourced from Simple Icons and OpenAI guidelines), a visual-check system that captures browser-rendered evidence, and a delivery pipeline with sidecar provenance and atomic output guarantees. It also includes a self-contained viewer with embedded JetBrains Mono fonts, cross-platform file openers, and update-checking infrastructure.

archify · high confidence

Automated contribution cards for merged pull requests

The repository now automatically generates and publishes a visual 'contribution card' for every human-authored pull request that merges into the default branch. This tool, located in \tools/contributor-cards\, uses a Node.js CLI and a GitHub Actions workflow to fetch PR metadata, render a 1000×1500 PNG using Chrome and embedded assets (Manrope, Barlow Condensed fonts, and a paper-map background), and publish the image to a dedicated \contributor-cards\ branch. It also posts a thank-you comment on the PR. The system includes strict validation (rejecting bots, unmerged PRs, and non-default branches), security measures (HTML escaping, CSP, no shell interpolation), and handles edge cases like long usernames, long titles, and concurrent race conditions.

tools · high confidence

Automated documentation and distribution build pipeline

The repository now includes a suite of build scripts that automate the generation of static documentation and distributable archives. \build-gallery.mjs\ renders diagram examples into interactive HTML artifacts, validates them against engineering checks, and assembles a proof gallery page with a JSON manifest. \build-guide.mjs\ and \build-start.mjs\ generate the user guide and interactive 'start' pages by injecting scenario recipes and prompts into HTML templates. \build-zip.sh\ creates reproducible skill archives by enforcing a specific Node.js toolchain and using deterministic ZIP writing. Supporting scripts like \check-release-identity.mjs\ and \check-stable-update-manifest.mjs\ enforce versioning contracts and update manifest integrity, while \ci-scope.mjs\ optimizes CI runs by narrowing scope to full checks for code changes and fast-tracking documentation-only updates.

scripts · high confidence

Introduce opt-in DeepSeek Harness integration for Archify

Adds a new community integration for DeepSeek Harness (DSH) that installs the Archify architecture-diagramming skill as a plugin. The adapter (v0.2.0) bundles Archify skill 2.17.0-dev.1 and is compatible with DSH 0.1.2-rc.1. It registers a single filesystem skill provider named \archify-plugin\ via a \cordis.patch.yml\ host activation expression that resolves the skill directory from the installed package. The integration is designed to be secure and portable: it contains no telemetry, network requests, credentials handling, or npm scripts, and includes scripts and tests to ensure reproducible packaging and correct Windows CLI resolution.

integrations/deepseek-harness · high confidence

Introduce ordinary-model floor benchmark for first-pass diagram usability

A new benchmark suite named 'ordinary-model floor' has been added to measure whether an ordinary coding agent can produce a usable Archify diagram on its first attempt without human repair. The suite includes a verification harness (\benchmark.mjs\), a manifest defining five bounded tasks (architecture, workflow, sequence, data flow, and lifecycle), and a detailed README outlining the fair-run protocol. The benchmark enforces a strict 'first-pass' gate: a run is only considered usable if semantic requirements are met, the Archify CLI passes deterministic quality validation, and a human reviewer confirms no visual defects. It also introduces operational failure tracking for timeouts or missing candidates, ensuring that incomplete runs are recorded rather than dropped.

benchmarks/ordinary-model-floor · high confidence

New 'Signal Flow' visual preset and WebM export capability

The Archify visual evolution experiment introduces a new 'Signal Flow' preset, selected as the first production-ready visual direction over 'Blueprint Review' and 'Ember Ops'. This preset features a luminous, motion-forward design with specific color palettes for dark and light themes. Additionally, the experiment validates the ability to record self-contained trace animations as WebM files directly in the browser using \canvas.captureStream()\ and \MediaRecorder\, without requiring external dependencies like Puppeteer or ffmpeg. Unsupported browsers and static diagrams gracefully disable the export menu item rather than failing. The prototype also includes a decision map and interactive HTML/MJS files to demonstrate these capabilities.

experiments/visual-evolution · high confidence

New MCO Runtime Architecture diagram and source data

The MCO Showcase now includes a new architecture diagram for the MCO runtime, generated by Archify. The entry point is the JSON definition in \mco-runtime.architecture.json\, which models the system's components (Entry Surfaces, Command Router, Policy, Invocation Workflow, Adapter Registry, Transport Layer, Provider CLIs, Session Daemon, and Artifacts) and their interactions. This data drives the static HTML visualization in \mco-runtime.html\, which renders the diagram with a dark/light theme toggle, embedded JetBrains Mono fonts, and specific visual presets like 'signal-flow' and 'trace'.

experiments/mco-showcase · high confidence

New grid-based layout example for architecture diagrams

The examples directory now includes a new architecture diagram (\archify-repo-grid\) that demonstrates the grid layout mode. This example shows how to position components using \row\ and \col\ coordinates within a defined grid structure (7 columns, specific gaps and cell sizes) rather than manual absolute positioning. It serves as a reference for users wanting to use the grid placement feature for structured, aligned diagrams.

examples · high confidence

New repair-rounds benchmark for measuring feedback-loop efficiency

A new benchmark harness has been added to \benchmarks/repair-rounds\ to measure how efficiently Archify receipts help a caller converge to a passing artifact. The suite injects deterministic defects (such as missing metadata, invalid node types, or layout overflows) into example fixtures and runs a repair loop using the real Archify CLI. It reports metrics including the number of rounds to pass, first-round disclosure rates, late discovery rates, and an observable token cost estimate. The harness includes a rules-based repairer that attempts to fix defects using only the diagnostic information provided in the receipts, allowing users to evaluate the actionability of the tool's feedback.

benchmarks/repair-rounds · high confidence

Website migrated to Astro with static generation and bilingual support

The website source has been rebuilt using Astro, replacing the previous implementation with static page generation for the home, gallery, guide, and start pages. This change introduces a new component architecture (Brand, Navigation, GalleryCard) and a layout system that renders pages as static HTML, eliminating client-side JavaScript bundles for the core site structure. The site now supports English and Chinese content via data attributes, with client-side scripts handling language switching and gallery filtering. The gallery page specifically displays a curated set of 11 verified diagram artifacts, each linked to its source JSON and interactive HTML preview, with structural validation receipts displayed on each card.

website/src · high confidence

Architecture

Viewer source modularized into standalone, editable components

The viewer's source code has been restructured from a single monolithic file into a set of modular, standalone JavaScript modules (such as export.js, focus.js, guided-views.js, and intent-trace.js) alongside a main stylesheet (viewer.css) and a template shell (template.source.html). This change allows maintainers to edit specific features—like the export pipeline, semantic focus, or guided views—independently without managing a massive single file. The system now uses an assembly process to inject these fragments into the final generated artifact, preserving the original HTML structure while enabling cleaner, isolated development of viewer capabilities.

viewer · high confidence

Behavioural changes

Mermaid parser experiment failed; v3.0 scope reduced to JSON IR

The v3.0 Mermaid validation experiment in \experiments/v3-mermaid-validation\ has concluded with a FAIL result, leading to the cancellation of the planned Mermaid flowchart parser (P1). Visual quality testing showed that applying Archify CSS to stock Mermaid output (Version B) did not significantly improve aesthetics over the default Mermaid rendering (Version A), confirming that layout is the primary product differentiator, not styling. Consequently, the project scope for v3.0 is now limited to the JSON IR and \render.js\ for coordinate stability (P0/P0.5). Future Mermaid input will be handled via a SKILL.md prompt-engineering trick where an LLM reads the structure and lays out the diagram from scratch in Archify style, rather than using an auto-layout parser.

experiments/v3-mermaid-validation · high confidence

Migrate public website to Astro with preserved visual parity

The public website has been rebuilt using Astro and React, replacing the previous static HTML structure while maintaining identical visual output and URLs (index, gallery, guide, start). This change introduces a new build pipeline requiring Node.js 22 and TypeScript 6.x, utilizes Tailwind CSS with specific configuration to preserve existing styles, and includes validation tests to ensure byte-identical preservation of non-page assets and semantic equivalence of generated content. The legacy HTML files in the docs folder are retained only as visual baselines for migration verification and are no longer the source of deployed pages.

website, website/scripts · high confidence

Test coverage

Added tests for visual parity, migration integrity, and public staging

Added test coverage in the website/test directory to verify the Astro migration and public staging process. The new tests assert visual parity by comparing SHA-256 hashes of full-page captures across multiple pages and viewports, ensuring the new site matches the previous version. They also validate that the migration preserves specific DOM structures, accessibility features, and script configurations on the homepage, while ensuring all other pages and public assets remain byte-identical to the baseline. Additionally, tests verify that the public staging script correctly excludes untracked files, rejects external symlinks, and prevents URL collisions.

website/test · high confidence

Dependencies

Update dependencies and lockfiles for Archify and website

This change updates the dependency manifests and lockfiles for the Archify CLI and the website. The Archify CLI (archify/package.json) now requires Node.js \>=18 and includes devDependencies for ajv, parse5, saxes, and simple-icons, with an override for fast-uri to ^3.1.7. The website (website/package.json) now requires Node.js \>=22.12.0 and uses Astro ^7.3.3, React ^19.3.0, Tailwind CSS ^4.3.3, and TypeScript \~6.0.3. The DeepSeek Harness integration (integrations/deepseek-harness/package.json) is also updated to version 0.2.0 with Node.js ^22.19.0 \|\| \>=24.0.0.

(dependencies) · high confidence

Written by watchdog.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

How this codebase got here

Score

  • CAI 44 → 39 (-5.2)
  • Rubric changed (rubric-2026.09.11 → rubric-2026.09.18) — scores are not directly comparable.

Lenses

  • Code Health 24 → 25 (+1.3)
  • Architecture 84 → 66 (-18.1)
  • Maturity 81 → 74 (-7.3)
  • Readiness 52 → 44 (-8.6)
  • Security 78 → 83 (+5.6)
  • Accessibility 61 → 47 (-14.1)
  • Performance 60 (new)

Resolved (74)

  • (anonymous) (cognitive 163) (viewer/semantic-radar.js)
  • (anonymous) (cognitive 222) (viewer/semantic-lens.js)
  • (anonymous) (cognitive 41) (viewer/reader-layout.js)
  • (anonymous) (cyclomatic 172) (viewer/semantic-radar.js)
  • (anonymous) (cyclomatic 205) (viewer/semantic-lens.js)
  • (anonymous) (cyclomatic 45) (viewer/reader-layout.js)
  • (anonymous) (cyclomatic 63) (viewer/intent-trace.js)
  • (anonymous)::begin (cognitive 33) (viewer/route-probe.js)
  • (anonymous)::begin (cyclomatic 28) (viewer/route-probe.js)
  • (anonymous)::clear (cognitive 16) (viewer/focus.js)
  • (anonymous)::render (cognitive 31) (viewer/motion-governor.js)
  • (anonymous)::render (cyclomatic 34) (viewer/motion-governor.js)
  • (anonymous)::renderShareCard::(anonymous)::onload (cognitive 18) (viewer/export.js)
  • (anonymous)::renderShareCard::(anonymous)::onload (cyclomatic 19) (viewer/export.js)
  • (anonymous)::select (cognitive 24) (viewer/node-finder.js)
  • (anonymous)::select (cyclomatic 20) (viewer/node-finder.js)
  • Dependency hygiene PARTLY measured — npm pinning read, dependency currency not (no pnpm-resolved versions to grade)
  • Documentation: no installation or build instructions
  • Documentation: no installation or build instructions (README.md)
  • Documentation: no usage examples
  • …and 54 more

New (348)

  • (anonymous) (cognitive 160) (viewer/semantic-radar.js)
  • (anonymous) (cognitive 18) (archify/renderers/lifecycle/render-lifecycle.mjs)
  • (anonymous) (cognitive 220) (viewer/semantic-lens.js)
  • (anonymous) (cognitive 29) (viewer/reader-layout.js)
  • (anonymous) (cognitive 44) (viewer/node-outline.js)
  • (anonymous) (cyclomatic 169) (viewer/semantic-radar.js)
  • (anonymous) (cyclomatic 202) (viewer/semantic-lens.js)
  • (anonymous) (cyclomatic 23) (archify/renderers/lifecycle/render-lifecycle.mjs)
  • (anonymous) (cyclomatic 29) (viewer/reader-layout.js)
  • (anonymous) (cyclomatic 42) (viewer/node-outline.js)
  • (anonymous) (cyclomatic 62) (viewer/intent-trace.js)
  • (anonymous)::begin (cognitive 31) (viewer/route-probe.js)
  • (anonymous)::begin (cyclomatic 26) (viewer/route-probe.js)
  • (anonymous)::clear (cognitive 20) (viewer/focus.js)
  • (anonymous)::clear (cyclomatic 17) (viewer/focus.js)
  • (anonymous)::figureLayout (cognitive 18) (viewer/export.js)
  • (anonymous)::figureLayout (cyclomatic 20) (viewer/export.js)
  • (anonymous)::measure (cognitive 33) (viewer/reader-layout.js)
  • (anonymous)::measure (cyclomatic 31) (viewer/reader-layout.js)
  • (anonymous)::render (cognitive 21) (viewer/motion-governor.js)
  • …and 328 more

Changes since last survey

  • 300 commits — 199 feature/other, 101 fixes

By area

  • (repo) — 83 commits
  • (root) — 55 commits
  • archify/test — 52 commits
  • archify/renderers — 38 commits
  • docs/gallery — 26 commits
  • archify/bin — 7 commits
  • docs/assets — 7 commits
  • benchmarks/repair-rounds — 6 commits
  • archify/references — 5 commits
  • examples/checkout-platform-delta.html — 4 commits
  • archify/examples — 3 commits
  • integrations/deepseek-harness — 3 commits
  • archify/assets — 2 commits
  • docs/skill-updates — 2 commits
  • .github/workflows — 1 commit
  • docs/evidence — 1 commit
  • docs/reviews — 1 commit
  • examples/web-app.html — 1 commit
  • scripts/stage-clean-skill.mjs — 1 commit
  • website/src — 1 commit

Notable commits

  • fix: Cover right-side outer route regression alongside bottom corridors
  • fix: Fix viewer flash when switching dark and light themes (#553)
  • fix: Merge dev into brand encoding fix and rebuild canonical archive
  • fix: Merge dev into workflow vertical label fix
  • fix: Merge latest dev into lifecycle note-only fix
  • fix: Merge pull request #503 from tt-a1i/codex/fix-label-clearance
  • fix: Merge pull request #509 from tt-a1i/codex/pr505-dev-fix
  • fix: Merge pull request #513 from Momoyeyu/fix/dsh-acceptance-hardening
  • fix: Merge pull request #515 from To3akaRin/codex/fix-brand-encoding-negotiation
  • fix: Merge pull request #532 from tt-a1i/codex/fix-intent-arrow-occlusion
  • fix: Merge pull request #536 from outlier27-cell/fix/469-lock-exported-svg-theme
  • fix: Merge pull request #545 from ooiuuii/fix/workflow-vertical-label-anchor
  • fix: Merge pull request #550 from To3akaRin/codex/fix-lifecycle-note-only
  • fix: Merge pull request #556 from tt-a1i/codex/fix-lifecycle-band-title
  • fix: Merge pull request #577 from outlier27-cell/fix/572-website-links
  • fix: Merge pull request #584 from tt-a1i/codex/dev-review-fixes
  • fix: Merge pull request #591 from aashish254/fix/workflow-clearance-numeric-tolerance
  • fix: Merge pull request #592 from tt-a1i/fix/sequence-legend
  • fix: Merge pull request #593 from tt-a1i/fix/reader-first-screen-fit
  • fix: Merge remote-tracking branch 'origin/dev' into fix/469-lock-exported-svg-theme
  • …and 280 more

Written by watchdog.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

Survey your own repository

tt-a1i/archify was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.

About this page

  • The score is its most recent published measurement, taken on 1 October 2026 at a pinned commit. It is not a live figure and does not change until the project is measured again.
  • Measured at commit d5a1333d7447c866a765adac7d4d062f2f02e4d2 — the exact code this score is about.
  • Scored under rubric-2026.09.18 — the same rubric and the same method as every other entry in this index.
  • Measured by watchdog.canine.dev using codehealth-analyzer preprod-e569280dd5e2.