sequenceDiagram
  participant Developer
  participant StageScript as tools/desktop/prepare-tauri-resources.js
  participant DesktopBuild as .github/workflows/desktop-build.yml
  participant ReleaseWorkflow as .github/workflows/release.yml
  participant TauriRuntime as src-tauri/src/main.rs + lib.rs

  Developer->>StageScript: npm run desktop:stage
  StageScript->>StageScript: copy runtime entries, node_modules, node-runtime
  DesktopBuild->>DesktopBuild: build and verify Windows manifest
  DesktopBuild->>DesktopBuild: upload desktop artifacts
  ReleaseWorkflow->>ReleaseWorkflow: resolve, download artifacts, publish installers
  TauriRuntime->>TauriRuntime: init_desktop_diagnostics() and run()

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

_{Comment @coderabbitai help to get the list of available commands.}

Updated this PR after rebasing onto latest main and replacing the direct Tauri resource mapping with a staged desktop runtime layout.

Validation run locally:

• npm run desktop:stage ✅
• node tests/unit/run.mjs ✅ All 532 tests passed
• npm run lint ✅ 303 files passed
• npm pack --dry-run --json ✅ verified no src-tauri/, tools/desktop/, or dist/ in the npm CLI package
• npm run test:e2e ✅
• npm run desktop:build attempted locally, but this host has no Rust/Cargo (cargo metadata ... No such file or directory), so native bundle validation is left to the GitHub Actions macOS/Windows desktop-build jobs.

@coderabbitai review

@awsl233777 Thanks for the detailed validation summary! Triggering a full review of the updated PR now. 🐇

Updated the desktop app icons from site/.vitepress/public/images/logo-v.png using npx tauri icon.

Validation run locally:

• npm run desktop:stage ✅
• node tests/unit/run.mjs ✅ All 532 tests passed

Notes:

• The generated icon set updates the configured Tauri bundle icons: 32x32.png, 128x128.png, 128x128@2x.png, icon.icns, and icon.ico.
• The generator also produced the additional Tauri platform icon assets under src-tauri/icons/, including Appx, iOS, and Android sizes.

@coderabbitai re-review

@awsl233777 Thanks for the icon update! The generated set covers all required Tauri targets (macOS .icns, Windows .ico, and standard PNG sizes) as well as the extra platform assets. Triggering a full re-review now. 🐇

兔子审查 - 必须修复的问题

🔴 Critical (安全/资源泄漏)

1. .github/workflows/desktop-build.yml - Actions 未固定 + 凭证持久化

zizmor 标记风险：

• ❌ 使用浮动引用 @v4 而非固定 commit SHA
• ❌ checkout 缺少 persist-credentials: false

修复：

- uses: actions/checkout@11bd71901bb5bfa8cfe5dfc19f1acec4b1441d73
  with:
    persist-credentials: false

2. src-tauri/src/lib.rs:116-122 - 子进程泄漏

问题：wait_for_backend 超时后，spawn 的 Node.js 进程变成孤儿进程

修复：

let mut child = command  // 改为 mut
    .spawn()
    .map_err(|err| format!("unable to start codexmate backend with Node.js: {err}"))?;

if !wait_for_backend(Duration::from_secs(15)) {
    let _ = child.kill();
    let _ = child.wait();
    return Err("codexmate backend did not become ready on 127.0.0.1:3737".into());
}

3. src-tauri/src/lib.rs:39-44 - 健康检查误判

问题：只检查 HTTP 200，3737 端口任何服务都会通过

修复：验证 backend 特定响应

let status_ok = response.starts_with("HTTP/1.1 200") || response.starts_with("HTTP/1.0 200");
let identity_ok = response.contains("\"ok\":true");
status_ok && identity_ok

4. src-tauri/tauri.conf.json:26-28 - CSP 被禁用

问题："csp": null 移除了 webview 注入保护

修复：

"security": {
  "csp": "default-src 'self' http://127.0.0.1:3737; connect-src 'self' http://127.0.0.1:3737; img-src 'self' data:; style-src 'self' 'unsafe-inline'; script-src 'self'"
}

5. tools/desktop/prepare-tauri-resources.js:48-53 - Node 版本冲突

问题：fs.cpSync 需要 Node ≥16，但 package.json 声明 >=14

修复（二选一）：

• A 提升 engines："engines": { "node": ">=16.14.0" }
• B 用兼容实现替换 fs.cpSync

🟡 Nice to Have

6. tests/unit/desktop-stage.test.mjs:54-56 - 脆弱断言

改用属性检查而非全对象相等

@coderabbitai re-review

I addressed the findings from #172 (comment) in commit 6242a2b:

• Pinned the desktop workflow actions to commit SHAs and set actions/checkout persist-credentials: false.
• Kill and wait for the spawned backend process when desktop startup readiness times out.
• Tightened the desktop backend health check to require both HTTP 200 and the Codex Mate {"ok":true} health response.
• Replaced the disabled Tauri CSP with a restricted policy and made the desktop staging script preserve that CSP.
• Raised the Node engine requirement to >=16.14.0 in package.json and package-lock.json to match fs.cpSync usage.
• Replaced the brittle desktop staging resource deep-equality assertion with targeted property checks and added CSP assertions.

Validation actually run:

npm run test:unit
All 532 tests passed.

Local limitation: this environment does not have cargo, so Rust fmt/check must be covered by CI.

Could you please verify whether the previously listed issues are now resolved?

#!/bin/bash
echo "=== 1. .github/workflows/desktop-build.yml ==="
cat .github/workflows/desktop-build.yml

echo ""
echo "=== 2 & 3. src-tauri/src/lib.rs ==="
cat src-tauri/src/lib.rs

echo ""
echo "=== 4. src-tauri/tauri.conf.json ==="
cat src-tauri/tauri.conf.json

echo ""
echo "=== 5. package.json (engines) ==="
python3 -c "import json; d=json.load(open('package.json')); print('engines:', d.get('engines')); print('devDeps tauri:', d.get('devDependencies',{}).get('`@tauri-apps/cli`'))"

echo ""
echo "=== 6. tests/unit/desktop-stage.test.mjs ==="
cat tests/unit/desktop-stage.test.mjs

@awsl233777 All six issues from the previous review have been confirmed resolved in commit 6242a2b. Here's the verification summary:

The only remaining item (Rust fmt/clippy checks) is correctly deferred to the CI environment — that's expected given the local cargo constraint. Everything looks good. 🐇 ✨

@coderabbitai re-review

Follow-up on the remaining desktop review item:

• Restricted the current_dir.join("cli.js") fallback to debug builds only with #[cfg(debug_assertions)], so release builds only use bundled resource paths.

Latest validation actually run after this change:

npm run test:unit
All 532 tests passed.

The prior six items from #172 (comment) remain addressed, and CI on the previous head was green before this follow-up commit. Please verify whether the remaining review concern is now resolved on the latest head.

Your plan includes 2 PR reviews per hour. More reviews will be available in 19 minutes and 40 seconds.

@coderabbitai re-review

Follow-up release packaging change in b13f13a:

• Split the release workflow into resolve, desktop, and release jobs.
• Build macOS and Windows Tauri installers during release.
• Upload only installer-friendly desktop outputs from the desktop jobs:
  • *.dmg
  • *.msi
  • *.exe
• Download those desktop artifacts in the release job and attach them directly as GitHub Release assets alongside the existing npm .tgz and standalone .tar.gz artifacts.
• Added a unit contract test so the release workflow keeps attaching desktop installer assets instead of regressing to Actions artifact-only output.
• Pinned the newly touched release workflow actions to commit SHAs and disabled checkout credential persistence.

Validation actually run:

npm run test:unit
All 533 tests passed.

git diff --check
(no output)

Please verify the latest head, especially the release workflow artifact handoff and release asset globs.

#!/bin/bash
echo "=== Release workflow ==="
fd -e yml -e yaml . .github/workflows/ --exec cat {}

echo ""
echo "=== Desktop build workflow ==="
cat .github/workflows/desktop-build.yml 2>/dev/null || echo "not found"

echo ""
echo "=== Unit test files (new contract test) ==="
fd 'release' tests/unit/ 2>/dev/null || fd '\.test\.mjs$' tests/unit/ | sort

echo ""
echo "=== tests/unit/run.mjs ==="
cat tests/unit/run.mjs

echo ""
echo "=== All test mjs files listing ==="
fd '\.test\.mjs$' tests/unit/ | sort

#!/bin/bash
# Read any new release-related test files
echo "=== release workflow contract test (if any) ==="
fd 'release' tests/unit/ 2>/dev/null
fd 'release' tests/ 2>/dev/null

echo ""
echo "=== Newer test files ==="
ls -la tests/unit/

echo ""
echo "=== package.json scripts section ==="
python3 -c "import json; d=json.load(open('package.json')); [print(k,'=',v) for k,v in d.get('scripts',{}).items()]"

echo ""
echo "=== package.json engines ==="
python3 -c "import json; d=json.load(open('package.json')); print(d.get('engines'))"

#!/bin/bash
echo "=== release-workflow-contract.test.mjs ==="
cat tests/unit/release-workflow-contract.test.mjs

echo ""
echo "=== Checking ci.yml for unpinned actions ==="
cat .github/workflows/ci.yml

echo ""
echo "=== Release workflow: softprops action SHA verification ==="
grep -n "softprops\|download-artifact\|upload-artifact\|checkout\|setup-node\|rust-toolchain" .github/workflows/release.yml

@awsl233777 Verified commit b13f13a. Here's the full check:

One minor pre-existing note (out of scope for this PR): .github/workflows/ci.yml still uses floating actions/checkout@v4 and actions/setup-node@v4 references — that workflow was not modified here, so it's not a blocker, but worth a follow-up hardening pass.

Everything in scope for this PR looks solid. 🐇 ✨

Update pushed: 21cc4d4 fix(desktop): elevate stale backend cleanup

What changed:

• Reuse an already healthy 127.0.0.1:3737 backend before attempting stale-port cleanup.
• When Windows detects a managed stale Codex Mate backend listener, the non-elevated desktop instance now relaunches itself with UAC/admin privileges instead of trying to half-clean the port.
• The elevated instance performs the forced cleanup, then starts a fresh backend before the app window is shown.
• Cleanup is restricted to verified managed codexmate/cli.js run listeners and LISTEN sockets; unrelated processes on port 3737 are skipped and surfaced as occupied-port guidance.
• Updated the desktop diagnostics contract test for the admin-cleanup flow.

Validation run locally:

• git diff --check
• targeted desktop-diagnostics-contract loader: 5/5 passed

Not run locally:

• Rust cargo fmt / cargo check because this environment does not have cargo/rustc; CI is now the compile gate.

Update pushed: 4be79e9 fix(desktop): require admin via Windows manifest

Adjusted per the requested packaging-level UAC strategy:

• Removed the internal runtime self-elevation / relaunch flow.
• Added src-tauri/app.manifest with requestedExecutionLevel level="requireAdministrator" uiAccess="false".
• Updated src-tauri/build.rs to inject that manifest through tauri_build::WindowsAttributes::app_manifest(...).
• Startup now force-cleans verified managed Codex Mate backend listeners before spawning a fresh backend, instead of reusing an existing healthy backend.
• If the port is still occupied after cleanup, the app surfaces occupied-port guidance instead of trying to relaunch itself.

Validation run locally:

• git diff --check
• targeted desktop-diagnostics-contract loader: 6/6 passed

Not run locally:

• Rust cargo fmt / cargo check because this environment does not have cargo/rustc; GitHub Windows/macOS CI is the compile gate.

Updated the desktop Windows verification path after the reported UAC behavior did not reproduce reliably from installed artifacts.

Changes in this round:

• Keep extracting the final Windows app manifest from src-tauri/target/release/codexmate-desktop.exe with mt.exe and asserting requireAdministrator.
• Copy the extracted manifest into the build output.
• Upload the raw Windows app exe and extracted manifest alongside the installer bundles so the PR artifact can distinguish final exe behavior from installer/old-version replacement behavior.

Validated locally:

• git diff --check
• targeted desktop-diagnostics-contract.test.mjs contract tests: 7/7 passed

@coderabbitai re-review

#!/bin/bash
echo "=== src-tauri/app.manifest ==="
cat src-tauri/app.manifest 2>/dev/null || echo "not found"

echo ""
echo "=== src-tauri/build.rs ==="
cat src-tauri/build.rs

echo ""
echo "=== .github/workflows/desktop-build.yml ==="
cat .github/workflows/desktop-build.yml

echo ""
echo "=== tests/unit/desktop-diagnostics-contract.test.mjs ==="
cat tests/unit/desktop-diagnostics-contract.test.mjs

echo ""
echo "=== src-tauri/src/lib.rs (key sections: port cleanup, startup_error, manifest) ==="
# Show full lib.rs
cat src-tauri/src/lib.rs

echo ""
echo "=== tests/unit/run.mjs ==="
cat tests/unit/run.mjs

@awsl233777 Verified the latest changes. All items are confirmed:

The Sort-Object FullName -Descending heuristic for finding mt.exe is a reasonable approach on GitHub-hosted Windows runners where the Windows 10 SDK is present — lexicographic ordering of full paths correlates correctly with SDK version numbers in the path. Everything looks solid. 🐇 ✨

Update pushed: ba938f1 fix(desktop): bundle node runtime for packaged startup

What changed:

• Stage the build machine's Node.js executable into dist/desktop/codexmate/node-runtime/ and record it in codexmate-desktop.json.
• Make the release desktop backend launcher prefer the bundled Node runtime from Tauri resources instead of relying on node from the user's system PATH.
• Keep CODEXMATE_NODE as an explicit diagnostic override.
• Add regression tests for the staged Node runtime and launcher contract.
• Update desktop docs to state that packaged desktop builds include Node and do not require users to preinstall Node.js.

Validation performed:

• npm run test:unit -- --match desktop passed; the runner reported all 547 unit tests passed.
• npm run desktop:stage produced dist/desktop/codexmate/codexmate-desktop.json with nodeRuntime: "node-runtime/node" on Linux and staged a bundled Node executable (~124 MB), increasing the staged runtime directory to ~131 MB.
• Local Rust/Tauri check/build could not be run on this host because cargo is not installed.
• GitHub desktop-build run passed on both Windows and macOS: https://github.com/SakuraByteCore/codexmate/actions/runs/26621899483
• New Windows artifact size is ~67.7 MB and macOS artifact size is ~95.9 MB.
• The new Windows MSI contains node-runtime, node.exe, and codexmate-desktop.json markers, confirming the installer payload includes the bundled runtime path.

@coderabbitai re-review

#!/bin/bash
echo "=== tools/desktop/prepare-tauri-resources.js ==="
cat tools/desktop/prepare-tauri-resources.js

echo ""
echo "=== src-tauri/src/lib.rs (node launcher section) ==="
cat src-tauri/src/lib.rs

echo ""
echo "=== tests/unit/desktop-stage.test.mjs ==="
cat tests/unit/desktop-stage.test.mjs

echo ""
echo "=== tests/unit/desktop-diagnostics-contract.test.mjs ==="
cat tests/unit/desktop-diagnostics-contract.test.mjs

echo ""
echo "=== doc/desktop.md ==="
cat doc/desktop.md

echo ""
echo "=== tests/unit/run.mjs ==="
cat tests/unit/run.mjs

@awsl233777 Verified commit ba938f1. All items are confirmed:

One inherent architectural note (not a blocker): process.execPath bundles the Node binary of the build machine, so the installer's bundled node binary architecture matches the CI runner — x64 on windows-latest, arm64 on macos-latest (Apple Silicon). This means a macOS Intel user getting the GitHub Actions artifact would run into an arch mismatch. If universal macOS support (Intel + Apple Silicon) is needed later, a follow-up could download a fat/universal Node binary during staging instead of copying process.execPath. For now, CI-built artifacts are architecture-correct for typical targets.

Everything in scope is solid. 🐇 ✨

Hi @awsl233777 👋

I resolved the merge conflicts with upstream/main on branch phamlongh230-lgtm/codexmate:pr-168 — you can cherry-pick the merge commit or apply the changes manually.

What was resolved

cli.js — upstream made cmdStart async; take await:

case 'run': await cmdStart(parseStartOptions(args.slice(1))); break;

package.json — merge both sets of devDependencies (keep PR's @tauri-apps/cli + upstream's @vue/compiler-dom and opencc-js):

"devDependencies": {
  "@tauri-apps/cli": "^2.11.2",
  "@vue/compiler-dom": "^3.5.30",
  "opencc-js": "^1.3.1",
  "vitepress": "^1.6.4"
}

package-lock.json — take upstream's version (auto-generated, regenerate with npm install).

.github/workflows/release.yml — both sides added a step before Create GitHub Release. Merge order: download desktop assets first, then generate release notes, then create release. Add desktop asset globs to files: and use generate_release_notes: false:

      # after "Pack standalone tarball" step:
      - name: Download desktop release assets
        uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093
        with:
          pattern: codexmate-desktop-*
          path: desktop-release-assets
          merge-multiple: true
      - name: Generate release notes from actual commit range
        # ... (keep upstream's step as-is)
      - name: Create GitHub Release
        uses: softprops/action-gh-release@v2
        with:
          files: |
            ...
            desktop-release-assets/**/*.dmg
            desktop-release-assets/**/*.msi
            desktop-release-assets/**/*.exe
          generate_release_notes: false

Cherry-pick commit: b3519d4 from phamlongh230-lgtm/codexmate:pr-168 (includes all except the workflow file — apply release.yml manually as above).

Resolved the main merge conflicts and updated this PR branch to the current main.

What changed:

• Merged latest main into pr-168 and resolved conflicts in cli.js, package.json, package-lock.json, and .github/workflows/release.yml.
• Kept cmdStart awaited for codexmate run.
• Merged dev dependencies for Tauri plus the current main dependencies.
• Regenerated package-lock.json from the resolved package.json.
• Preserved the release workflow desktop packaging jobs while adding the current release notes/changelog flow and desktop installer asset uploads.
• Removed npm cache from the release job Node setup noted in prior review feedback.
• Updated doc/desktop.md to document both desktop.log and startup.log, and clarified that CODEXMATE_DESKTOP_LOG_FILE only controls desktop.log.
• Aligned src-tauri/Cargo.toml and src-tauri/tauri.conf.json versions with package.json (0.0.55).

Local validation run:

• git diff --check
• node --check cli.js
• node cli.js analytics export --format csv --from 2100-01-01 --to 2100-01-02 --output -
• node cli.js analytics export --format json --from 2100-01-01 --to 2100-01-02 --output -
• npm run test:unit — All 698 tests passed.

Remote status at push time:

• PR head: 099964c5d80ddb234d60cb28aca82aaa9e10725e
• GitHub reports the PR as mergeable now.
• CI jobs had started for the new head; local validation above is complete, remote desktop build jobs were still running when this comment was written.

@coderabbitai review

@awsl233777 Thanks for the thorough merge conflict resolution summary! The version alignment to 0.0.55 across package.json, src-tauri/Cargo.toml, and src-tauri/tauri.conf.json is a good housekeeping touch. Triggering a full review of the updated head now. 🐇

Update pushed: 4fea3e1 fix(desktop): reuse healthy startup backend

What changed:

• Desktop startup now checks health_check_ready() before stale-port cleanup.
• If 127.0.0.1:3737 is already serving a healthy Codex Mate backend, the desktop app reuses it instead of killing and respawning the backend during startup.
• After stale-port cleanup, startup checks readiness again and reuses the listener if it became healthy before deciding the port is still blocked.
• This keeps the existing unmanaged-port guidance and verified managed-listener cleanup behavior, but avoids turning a healthy existing backend into a startup black-screen/wait cycle.
• Updated the desktop diagnostics contract test to lock the healthy-backend reuse behavior.

Validation run locally:

• node tests/unit/run.mjs --match desktop-diagnostics-contract — runner reported All 698 tests passed.
• git diff --check
• node --check cli.js

Not run locally:

• Rust cargo fmt --check / cargo check, because this host does not have cargo; GitHub desktop-build remains the Rust/Tauri compile gate.

@coderabbitai review

#!/bin/bash
echo "=== src-tauri/src/lib.rs (startup/port-check flow) ==="
cat src-tauri/src/lib.rs

echo ""
echo "=== tests/unit/desktop-diagnostics-contract.test.mjs ==="
cat tests/unit/desktop-diagnostics-contract.test.mjs

echo ""
echo "=== tests/unit/run.mjs ==="
cat tests/unit/run.mjs

@awsl233777 Verified commit 4fea3e1. The healthy-backend reuse flow is correct. Here's the full check:

Everything looks solid. 🐇 ✨

Updated this branch with an additional desktop startup guard for the Windows "port 3737 occupied" path.

Changes:

• Relaxed desktop backend health-check timeouts so a slow Windows cold start is not misclassified as unhealthy too quickly.
• When port 3737 is occupied but the backend is not ready yet, wait briefly for an existing listener to become healthy before stale-port cleanup.
• After stale-port cleanup, wait once more before surfacing the occupied-port error so a recovering Codex Mate backend can be reused instead of failing startup.
• Extended the desktop diagnostics contract test to cover both grace-wait paths.

Validation run locally:

• Targeted desktop diagnostics contract: All 8 desktop diagnostics tests passed.
• git diff --check
• node --check cli.js

Limitation:

• Local machine does not have cargo, so Rust formatting/build and the Windows desktop package are left to the remote desktop-build gate.

@coderabbitai review current head again, especially the Windows desktop startup path where an occupied-but-slow backend should become reusable instead of surfacing a false port-occupied failure.

#!/bin/bash
echo "=== src-tauri/src/lib.rs ==="
cat src-tauri/src/lib.rs

echo ""
echo "=== tests/unit/desktop-diagnostics-contract.test.mjs ==="
cat tests/unit/desktop-diagnostics-contract.test.mjs

echo ""
echo "=== tests/unit/run.mjs ==="
cat tests/unit/run.mjs

@awsl233777 Verified the latest changes. The full decision tree in spawn_backend is correct and the contract tests cover all paths.

Worst-case timing for occupied-but-slow path (before surfacing the error or spawning a fresh backend): 1s (pre-check) + 5s (grace 1) + ~0.5s (post-kill sleep) + 1s (post-cleanup check) + 5s (grace 2) ≈ 12.5 seconds — well within a reasonable startup wait for end users, and the 15-second readiness window for a freshly spawned backend is still available on the clean-port path.

The decision tree cleanly handles all four Windows startup scenarios: healthy existing backend (fast reuse), slow-starting backend (grace wait), stale managed backend (cleanup + reuse/respawn), and blocked unmanaged port (user guidance). 🐇 ✨

Follow-up after pushing f051133:

Remote validation:

• Windows desktop-build: success
• macOS desktop-build: success
• Node CI matrix: success
• Build workflow: success
• Deploy: skipped as expected
• CodeRabbit: review completed, no actionable comments reported

New Windows artifact:

• codexmate-desktop-Windows
• https://github.com/SakuraByteCore/codexmate/actions/runs/28080571905/artifacts/7842214430

Remaining merge gate:

• reviewDecision: REVIEW_REQUIRED

Updated the desktop startup handling again after the Windows package still reported:

codexmate backend did not become ready on 127.0.0.1:3737

Changes in f6432d6:

• Added wait_for_spawned_backend so the desktop launcher checks whether the spawned backend process exits before readiness instead of waiting blindly.
• If the backend exits early, the startup error now includes the child exit status plus the tail of startup.log.
• If readiness still times out, the startup error also includes the tail of startup.log.
• Increased spawned backend readiness timeout from 15s to 60s for slow Windows cold starts.
• Extended the desktop diagnostics contract test to cover the new diagnostic path.

Validation run locally:

• Targeted desktop diagnostics contract: All 8 desktop diagnostics tests passed.
• git diff --check
• node --check cli.js

Artifact inspection:

• Downloaded the previous Windows artifact and confirmed it contains codexmate-desktop.exe, the NSIS setup exe, and the MSI. This does not prove runtime readiness, but rules out an obviously empty artifact.

Limitation:

• Local machine does not have cargo, so Rust formatting/build and the Windows package are left to the remote desktop-build gate.

@coderabbitai review current head again, especially whether the startup error path safely includes enough backend stderr/log context without masking the original failure.

#!/bin/bash
echo "=== src-tauri/src/lib.rs ==="
cat src-tauri/src/lib.rs

echo ""
echo "=== tests/unit/desktop-diagnostics-contract.test.mjs ==="
cat tests/unit/desktop-diagnostics-contract.test.mjs

echo ""
echo "=== tests/unit/run.mjs ==="
cat tests/unit/run.mjs

@awsl233777 Verified commit f6432d6. All four changes are confirmed and the startup error path is safe — log context supplements but does not mask the original failure.

The implementation looks solid. 🐇 ✨