Skip to content
Draft
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 12 additions & 10 deletions .gitignore
Original file line number Diff line number Diff line change
@@ -1,19 +1,13 @@
node_modules/

# Action runtime bundle must stay versioned.
dist/*
!dist/index.js

# Generated build artifacts should not be tracked.
packages/*/dist
workers/*/dist
workers/*/dist-cjs
workers/*/.wrangler
.vercel
dist/
out/
.next/
target/
*.tsbuildinfo
.build/
default.profraw

# Desktop app build
apps/desktop/src-tauri/target/
Expand All @@ -35,6 +29,11 @@ apps/desktop/native/AgentIsland/.build/
# Secrets
.claude/settings.local.json

# Local agent/tool state and output — not shared repo content.
.claude/projects/
.fleet/
.impeccable/

# IDE
.vscode/
.idea/
Expand All @@ -50,7 +49,10 @@ apps/desktop/synthetic-qa-artifacts/
.playwright-mcp/

# Public benchmark reviewer output (drop tool outputs here, not tracked)
benchmark/reviews/
benchmarks/public-catch-rate/reviews/

# Scratch output from CLI runs (--out artifacts/...). Committed evidence lives in evidence/.
artifacts/

# Coverage
coverage/
Expand Down
15 changes: 15 additions & 0 deletions .xcodebuildmcp/config.yaml
Original file line number Diff line number Diff line change
@@ -0,0 +1,15 @@
schemaVersion: 1
enabledWorkflows:
- doctor
- macos
- project-discovery
- project-scaffolding
- swift-package
- coverage
- utilities
sessionDefaults:
workspacePath: ./apps/macos/CodeVetter.xcworkspace
scheme: CodeVetter
filePathRenderStyle: tree
showTestTiming: true
sentryDisabled: true
39 changes: 26 additions & 13 deletions DESIGN.md
Original file line number Diff line number Diff line change
Expand Up @@ -2,10 +2,10 @@
name: CodeVetter
description: An evidence-first verification workbench in ink, warm amber, and explicit semantic state.
colors:
canvas-ink: "#060708"
surface-ink: "#0c0d0f"
raised-ink: "#111316"
elevated-ink: "#17191d"
canvas-ink: "#000000"
surface-ink: "#050506"
raised-ink: "#09090b"
elevated-ink: "#0d0d10"
evidence-white: "#f4f4f5"
secondary-gray: "#a1a1aa"
muted-gray: "#8a8a93"
Expand Down Expand Up @@ -100,16 +100,29 @@ oversized presentation typography on operating surfaces.

**Key Characteristics:**

- Ink surfaces separated by restrained tonal steps and hairline borders.
- A true-black canvas with near-black working planes separated by restrained
hairline borders.
- Compact native-feeling controls with generous focus treatment.
- Warm amber used sparingly for action, selection, and verification emphasis.
- Monospace reserved for paths, revisions, commands, and evidence identities.
- Every state remains understandable without color alone.

## Colors

The palette is near-black ink with a single warm action voice and explicit
semantic evidence colors.
The palette is true-black ink with a single warm action voice and explicit
semantic evidence colors. Chrome remains true black; working planes rise only
1–5% above it so hierarchy survives without turning the product charcoal.
Hairline borders carry the remaining separation instead of grey fill.

Light appearance is a native counterpart, not an inverted dark skin. It uses a
warm `#f7f6f3` canvas, quieter `#f1f0ed` chrome, white evidence planes, and a
soft `#f3f2ef` inspector. Amber keeps the same action and selection meaning;
amber text and icons deepen independently of the bright action fill, and
success, warning, and failure colors use darker light-mode counterparts so
normal-size evidence text retains at least 4.5:1 contrast. Semantic status
colors retain their written labels. Review, Testing,
Performance, and Runs must preserve the same evidence hierarchy and control
priority in both appearances.

### Primary

Expand All @@ -119,10 +132,10 @@ semantic evidence colors.

### Neutral

- **Canvas Ink:** the application background and deepest visual plane.
- **Surface Ink:** the default card and panel plane.
- **Raised Ink:** controls and nested evidence tiles.
- **Elevated Ink:** overlays or deliberately elevated sub-surfaces.
- **Canvas Ink:** the true-black application background and deepest plane.
- **Surface Ink:** the almost-black default card and panel plane.
- **Raised Ink:** a subtly lighter control and nested-evidence plane.
- **Elevated Ink:** the highest near-black overlay or inspector plane.
- **Evidence White:** primary text and decisive result labels.
- **Secondary Gray:** supporting explanations and metadata.
- **Muted Gray:** placeholders and low-priority context.
Expand Down Expand Up @@ -159,8 +172,8 @@ feed to another tool; prose and actions stay in the system sans.
## Layout

The desktop shell has a persistent top navigation and repository sidebar, with
a configured minimum window width of 900px. Operating pages use a centered,
wide workbench column and stack compact bordered panels vertically.
a configured minimum window size of 980 x 640 points. Operating pages use a
centered, wide workbench column and stack compact bordered panels vertically.

Cards use 20px internal padding and 24px section rhythm by default. Dense form
rows may align horizontally when space permits, then stack without changing
Expand Down
6 changes: 4 additions & 2 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -106,8 +106,10 @@ apps/
desktop/ Tauri 2 + React 19 + Vite desktop app — the core product
landing-page-astro/ Astro marketing site (static export, deployed to Cloudflare Pages — codevetter.com)
docs/ Canonical knowledge system — see docs/index.md
benchmark/ Public catch-rate benchmark cases + harness
openspec/ Spec-driven workflow (specs + changes/archive)
docs-site/ Blume presentation layer for docs/ (generated output is gitignored)
benchmarks/ Evaluation corpora (public catch-rate, agent PRs, runtime challenges)
evidence/ Committed run evidence (design, performance, reviews)
scripts/ Benchmark + corpus + deploy + doc-validation scripts
```

> The legacy Next.js `apps/landing-page/` was removed on 2026-07-03. The
Expand Down
21 changes: 17 additions & 4 deletions agents.md
Original file line number Diff line number Diff line change
Expand Up @@ -7,6 +7,14 @@ commands are authoritative; no sibling Fleet checkout is required. Protect
production stability, keep changes scoped, verify work with repo-local checks,
and record durable follow-up in this repository's GitHub Issues.

Do not run XCUITest, desktop runtime comparison, AppleScript, accessibility
control, or any automation that can launch apps, take focus, type, click, or
manage windows while the operator is using the Mac. Foreground automation
requires fresh authorization for an idle-screen window on that invocation;
never infer or persist approval. Prefer the background-safe native lane and
offscreen render gates. Use a dedicated graphical macOS runner for unattended
interaction qualification.

## Purpose
CodeVetter is an execution-backed verification and evaluation system for coding
agents. It determines whether an agent completed a software task correctly
Expand Down Expand Up @@ -56,14 +64,19 @@ apps/
tests/ # Playwright e2e tests
landing-page-astro/ # Astro marketing site → Cloudflare Pages (codevetter.com)
docs/ # Canonical knowledge system — see docs/index.md
benchmark/ # Public catch-rate benchmark cases + harness
scripts/ # Benchmark + deploy + doc-validation scripts
docs-site/ # Blume presentation layer for docs/ (NOT the source of truth)
benchmarks/ # Evaluation corpora — public-catch-rate/, agent-prs/,
# agent-tasks/, runtime-challenges/, context-*/, performance-lab/
evidence/ # Committed run evidence: design/, performance/, reviews/
scripts/ # Benchmark + corpus + deploy + doc-validation scripts
.github/workflows/ # ci, auto-release, release, deploy-landing, weekly, docs
blume.config.ts # Blume presentation layer for docs/ (NOT the source of truth)
STATUS.md # Compatibility pointer
PROJECT_STATUS.md # Current/shipped product truth (fleet source of truth)
```

`artifacts/` is gitignored scratch for CLI runs (`--out artifacts/...`).
Committed evidence belongs in `evidence/`.

## Key commands
```bash
# From apps/desktop/
Expand Down Expand Up @@ -121,7 +134,7 @@ node scripts/check-docs.mjs # Validate docs (links, frontmatter, structure)

The committed Markdown under `docs/` is the **source of truth** for product
knowledge, architecture, decisions, workflows, operations, learnings, and
failed approaches. Blume (`blume.config.ts`) is only the presentation/search
failed approaches. Blume (`docs-site/blume.config.ts`) is only the presentation/search
layer — generated output (`.blume/`) is gitignored.

- **Navigation hub**: `docs/index.md`
Expand Down
1 change: 0 additions & 1 deletion apps/landing-page-astro/src/layouts/Layout.astro
Original file line number Diff line number Diff line change
Expand Up @@ -107,7 +107,6 @@ const OG_IMAGE = `${SITE_URL}/og-image.png`;
is:inline
src="https://sassmaker.com/ai-chat-footer.js"
data-name="CodeVetter"
data-compose="false"
defer
></script>
</body>
Expand Down
4 changes: 3 additions & 1 deletion docs-site/blume.config.ts
Original file line number Diff line number Diff line change
Expand Up @@ -29,7 +29,9 @@ export default defineConfig({
},

theme: {
accent: 'amber', // matches the product's warm amber accent (#d4a039)
// Use the product token directly. Blume 1.5 no longer ships an "amber"
// preset, and treating the unknown slug as a CSS color breaks OG rendering.
accent: '#d4a039',
radius: 'md',
mode: 'system',
fonts: {
Expand Down
2 changes: 1 addition & 1 deletion docs-site/package.json
Original file line number Diff line number Diff line change
Expand Up @@ -8,6 +8,6 @@
"validate": "blume validate"
},
"dependencies": {
"blume": "1.0.4"
"blume": "1.5.3"
}
}
Loading
Loading