Release v1.0.0-beta.4: install OOM fix + reduce-effects + self-hosted CI

[jaylfc]

Promotes this session's work to master and cuts beta.4.

User-facing

• Install no longer builds the SPA locally (Projects feature: scoped agent workspaces with per-project memory #117): CI publishes a prebuilt bundle keyed by the desktop/ tree SHA; the installer + in-app update download it instead of running the vite build that OOMs on small machines (the 'ran the installer, still on the old version' bug). Local build is a loud-failing fallback only.
• Reduce effects / performance mode (E2E integration test: model aggregation + deploy routing on Fedora LXC worker #58): opt-in Settings -> Accessibility toggle that strips backdrop blur + heavy shadows + continuous animations for low-end devices.

Infra

• Self-hosted CI runners (VPS + Fedora pool) for fast test runs (mDNS / zero-conf discovery for workers on the local network #119)
• Desktop vitest suite now gated in CI (qmd fork drift: upgrade path from pinned feat/remote-llm-provider branch #114)
• Version bump to 1.0.0-beta.4

NOTE: the first master CI run after merge publishes the rolling bundle-latest prebuilt bundle for the first time -- watching it. The beta.4 GitHub release is cut from master tip after that's verified.

[qodo-code-review]

Qodo reviews are paused for this user.

Troubleshooting steps vary by plan Learn more →

On a Teams plan?
Reviews resume once this user has a paid seat and their Git account is linked in Qodo.
Link Git account →

Using GitHub Enterprise Server, GitLab Self-Managed, or Bitbucket Data Center?
These require an Enterprise plan - Contact us
Contact us →

[coderabbitai]

Warning

Review limit reached

@jaylfc, we couldn't start this review because you've reached your PR review rate limit.

More reviews will be available in 8 minutes and 28 seconds. Learn how PR review limits work.

Your organization has used up its prepaid credits, and credit purchases are no longer available. Enable the review add-on in the billing tab to keep reviews running — you're only billed for reviews past your plan's rate limits ($0.25/file).

⌛ How to resolve this issue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based credits.

🚦 How do rate limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan refill rate.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, the refill rate gradually slows as usage increases. The highest same-day bursts are limited more strictly.

Please see our Fair Usage Limits Policy for further information.

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: 6b99f7b7-2b3d-4cc0-a753-7840bf784c59

📥 Commits

Reviewing files that changed from the base of the PR and between 8996ac2 and 96eb555.

📒 Files selected for processing (14)

• .github/workflows/ci.yml
• desktop/src/App.tsx
• desktop/src/apps/SettingsApp.tsx
• desktop/src/lib/__tests__/projects-tasks.test.ts
• desktop/src/lib/__tests__/userspace-apps.test.ts
• desktop/src/stores/__tests__/reduce-effects.test.ts
• desktop/src/stores/theme-store.ts
• desktop/src/theme/tokens.css
• desktop/vite.config.ts
• docs/STATUS.md
• scripts/install-server.sh
• tests/test_desktop_rebuild.py
• tinyagentos/__init__.py
• tinyagentos/desktop_rebuild.py

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch dev

---

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.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

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

[github-actions]

👋 Thanks for the PR! This one targets master, which is our
stable branch (it's what live installs track). Please retarget it to
dev — click Edit next to the PR title and change the base
branch dropdown from master to dev. Your commits and any review
carry over, nothing is lost.

See CONTRIBUTING.md for the branch model.

[gitar-bot]

Note

Your trial team has used its Gitar budget, so automatic reviews are paused. Upgrade now to unlock full capacity. Comment "Gitar review" to trigger a review manually.
Learn more about usage limits

Code Review 👍 Approved with suggestions 0 resolved / 5 findings

Introduces prebuilt SPA bundles and a low-end device performance mode to improve installation reliability. Address several security and robustness concerns regarding prebuilt bundle integrity, insecure temporary file paths, and potential race conditions during the installation swap.

💡 Security: Prebuilt bundle installed with no content integrity verification

📄 tinyagentos/desktop_rebuild.py:101-115 📄 scripts/install-server.sh:1170-1184 📄 .github/workflows/ci.yml:152-166

_try_prebuilt_desktop_bundle (and the mirrored logic in install-server.sh) downloads desktop-bundle.tar.gz from the public bundle-latest release and serves its contents as the desktop UI after only checking that desktop-tree.txt equals the local git rev-parse HEAD:desktop. But desktop-tree.txt is published in the same release asset set as the tarball, so it provides no integrity guarantee — anyone who can write to the rolling release (or any future asset-tampering) can ship a malicious bundle alongside a matching tree string and it will be extracted and served verbatim. The 'it's our own CI output' assumption in the docstring rests entirely on GitHub release/account security, with no independent check.

Suggested hardening: after extraction, recompute a git-tree-style hash of the staged desktop/ (or, more simply, publish and verify a SHA-256 of the tarball that is generated independently of the tree key) before swapping it into static/desktop/. At minimum, document that the trust boundary is the GitHub release.

💡 Security: tar extracted without path-safe filter on older Python

📄 tinyagentos/desktop_rebuild.py:152-156 🔗 CWE-22

In _try_prebuilt_desktop_bundle, when tar.extractall(stage, filter="data") raises TypeError (Python without the filter kwarg) the code falls back to tar.extractall(stage) with no filtering. That fallback is vulnerable to path traversal / absolute-path / symlink entries (CVE-2007-4559 class) if the archive is ever malicious — which compounds the lack of integrity verification above. The filter="data" kwarg only landed in 3.12 (and 3.11.4+/3.10.12+ backports), so installs on older patch levels silently take the unsafe path.

Suggested fix: on the TypeError branch, either refuse to extract (return False and fall back to a local build) or manually validate each member's resolved path stays within stage before extracting.

Fall back to local build instead of unsafe extraction on old Python.

with tarfile.open(fileobj=io.BytesIO(blob), mode="r:gz") as tar:
    try:
        tar.extractall(stage, filter="data")  # py>=3.12 path-safe filter
    except TypeError:
        # No data filter on this Python: refuse rather than extract
        # an unvalidated archive (path-traversal risk). Build locally.
        logger.warning("tar 'data' filter unavailable; building locally.")
        return False

💡 Edge Case: Prebuilt helper errors abort rebuild instead of falling back to build

📄 tinyagentos/desktop_rebuild.py:112-122

The docstring states _try_prebuilt_desktop_bundle returns False (falling back to a local build) on "any mismatch, missing git, or network/extract failure." However the git-probe block only catches FileNotFoundError. If asyncio.create_subprocess_exec("git", ...) or proc.communicate() raises any other exception (e.g. OSError/PermissionError when spawning git), it propagates out of the helper to the outer except Exception in rebuild_desktop_bundle_if_stale, which returns RebuildResult(rebuilt=True, success=False). That reports the whole rebuild as failed and never attempts the local build — the opposite of the intended graceful fallback.

Suggested fix: broaden the probe's exception handling to return False on any exception, so the prebuilt path can never harden into a failure that blocks the local build.

Catch any exception from the git probe and fall back.

try:
    proc = await asyncio.create_subprocess_exec(
        "git", "-C", str(project_root), "rev-parse", "HEAD:desktop",
        stdout=asyncio.subprocess.PIPE, stderr=asyncio.subprocess.DEVNULL,
    )
    out, _ = await proc.communicate()
except Exception:
    return False  # git missing or unspawnable -> fall back to local build

💡 Security: install-server.sh uses predictable /tmp paths during root install

📄 scripts/install-server.sh:1177-1191

The installer stages the prebuilt bundle in fixed, world-predictable paths. Since the installer commonly runs as root, on a multi-user host a local attacker can pre-create these as symlinks or pre-seed files: removing the symlink itself, but the subsequent directory creation, download, and extraction operate on attacker-influenceable predictable names, enabling symlink/redirection style attacks. Use a randomized directory instead (e.g. mktemp -d) and extract/download inside it.

Replace fixed /tmp paths with mktemp-created locations.

_stage="$(mktemp -d)"
_tarball="$(mktemp)"
if curl -fsSL --max-time 120 "$_bundle_base/desktop-bundle.tar.gz" -o "$_tarball" \
   && tar -C "$_stage" -xzf "$_tarball" \
   && [[ -f "$_stage/desktop/index.html" ]]; then
    rm -rf "$INSTALL_DIR/static/desktop"
    mkdir -p "$INSTALL_DIR/static"
    mv "$_stage/desktop" "$INSTALL_DIR/static/desktop"
    # ... (chown unchanged) ...
fi
rm -rf "$_stage" "$_tarball"

💡 Bug: Prebuilt install deletes existing bundle before move can fail

📄 tinyagentos/desktop_rebuild.py:161-171

In _try_prebuilt_desktop_bundle the swap does shutil.rmtree(target) (removing the existing, working static/desktop/) and only then shutil.move(staged, target). stage is created under the system temp dir, which is frequently a different filesystem from the project, so shutil.move degrades to copy-then-delete and can fail partway (disk full, permissions, etc.). If the move fails after the rmtree, the install is left with no desktop bundle at all — the very 'silently stuck with no UI' outcome this feature was meant to prevent. The verify-before-swap (index.html check) guards against bad archives but not against a failed move. Consider moving the staged tree into a sibling of target (same filesystem) and doing an atomic-ish rename, or copy-into-temp-then-rename, so the old bundle survives a failed swap.

🤖 Prompt for agents

Code Review: Introduces prebuilt SPA bundles and a low-end device performance mode to improve installation reliability. Address several security and robustness concerns regarding prebuilt bundle integrity, insecure temporary file paths, and potential race conditions during the installation swap.

1. 💡 Security: Prebuilt bundle installed with no content integrity verification
   Files: tinyagentos/desktop_rebuild.py:101-115, scripts/install-server.sh:1170-1184, .github/workflows/ci.yml:152-166

   `_try_prebuilt_desktop_bundle` (and the mirrored logic in install-server.sh) downloads `desktop-bundle.tar.gz` from the public `bundle-latest` release and serves its contents as the desktop UI after only checking that `desktop-tree.txt` equals the local `git rev-parse HEAD:desktop`. But `desktop-tree.txt` is published in the *same* release asset set as the tarball, so it provides no integrity guarantee — anyone who can write to the rolling release (or any future asset-tampering) can ship a malicious bundle alongside a matching tree string and it will be extracted and served verbatim. The 'it's our own CI output' assumption in the docstring rests entirely on GitHub release/account security, with no independent check.
   
   Suggested hardening: after extraction, recompute a git-tree-style hash of the staged `desktop/` (or, more simply, publish and verify a SHA-256 of the tarball that is generated independently of the tree key) before swapping it into `static/desktop/`. At minimum, document that the trust boundary is the GitHub release.

2. 💡 Security: tar extracted without path-safe filter on older Python
   Files: tinyagentos/desktop_rebuild.py:152-156

   In `_try_prebuilt_desktop_bundle`, when `tar.extractall(stage, filter="data")` raises `TypeError` (Python without the `filter` kwarg) the code falls back to `tar.extractall(stage)` with no filtering. That fallback is vulnerable to path traversal / absolute-path / symlink entries (CVE-2007-4559 class) if the archive is ever malicious — which compounds the lack of integrity verification above. The `filter="data"` kwarg only landed in 3.12 (and 3.11.4+/3.10.12+ backports), so installs on older patch levels silently take the unsafe path.
   
   Suggested fix: on the `TypeError` branch, either refuse to extract (return False and fall back to a local build) or manually validate each member's resolved path stays within `stage` before extracting.

   Fix (Fall back to local build instead of unsafe extraction on old Python.):
   with tarfile.open(fileobj=io.BytesIO(blob), mode="r:gz") as tar:
       try:
           tar.extractall(stage, filter="data")  # py>=3.12 path-safe filter
       except TypeError:
           # No data filter on this Python: refuse rather than extract
           # an unvalidated archive (path-traversal risk). Build locally.
           logger.warning("tar 'data' filter unavailable; building locally.")
           return False

3. 💡 Edge Case: Prebuilt helper errors abort rebuild instead of falling back to build
   Files: tinyagentos/desktop_rebuild.py:112-122

   The docstring states `_try_prebuilt_desktop_bundle` returns False (falling back to a local build) on "any mismatch, missing git, or network/extract failure." However the git-probe block only catches `FileNotFoundError`. If `asyncio.create_subprocess_exec("git", ...)` or `proc.communicate()` raises any other exception (e.g. `OSError`/`PermissionError` when spawning git), it propagates out of the helper to the outer `except Exception` in `rebuild_desktop_bundle_if_stale`, which returns `RebuildResult(rebuilt=True, success=False)`. That reports the whole rebuild as failed and never attempts the local build — the opposite of the intended graceful fallback.
   
   Suggested fix: broaden the probe's exception handling to return False on any exception, so the prebuilt path can never harden into a failure that blocks the local build.

   Fix (Catch any exception from the git probe and fall back.):
   try:
       proc = await asyncio.create_subprocess_exec(
           "git", "-C", str(project_root), "rev-parse", "HEAD:desktop",
           stdout=asyncio.subprocess.PIPE, stderr=asyncio.subprocess.DEVNULL,
       )
       out, _ = await proc.communicate()
   except Exception:
       return False  # git missing or unspawnable -> fall back to local build

4. 💡 Security: install-server.sh uses predictable /tmp paths during root install
   Files: scripts/install-server.sh:1177-1191

   The installer stages the prebuilt bundle in fixed, world-predictable paths. Since the installer commonly runs as root, on a multi-user host a local attacker can pre-create these as symlinks or pre-seed files: removing the symlink itself, but the subsequent directory creation, download, and extraction operate on attacker-influenceable predictable names, enabling symlink/redirection style attacks. Use a randomized directory instead (e.g. `mktemp -d`) and extract/download inside it.

   Fix (Replace fixed /tmp paths with mktemp-created locations.):
   _stage="$(mktemp -d)"
   _tarball="$(mktemp)"
   if curl -fsSL --max-time 120 "$_bundle_base/desktop-bundle.tar.gz" -o "$_tarball" \
      && tar -C "$_stage" -xzf "$_tarball" \
      && [[ -f "$_stage/desktop/index.html" ]]; then
       rm -rf "$INSTALL_DIR/static/desktop"
       mkdir -p "$INSTALL_DIR/static"
       mv "$_stage/desktop" "$INSTALL_DIR/static/desktop"
       # ... (chown unchanged) ...
   fi
   rm -rf "$_stage" "$_tarball"

5. 💡 Bug: Prebuilt install deletes existing bundle before move can fail
   Files: tinyagentos/desktop_rebuild.py:161-171

   In `_try_prebuilt_desktop_bundle` the swap does `shutil.rmtree(target)` (removing the existing, working `static/desktop/`) and only then `shutil.move(staged, target)`. `stage` is created under the system temp dir, which is frequently a different filesystem from the project, so `shutil.move` degrades to copy-then-delete and can fail partway (disk full, permissions, etc.). If the move fails after the rmtree, the install is left with no desktop bundle at all — the very 'silently stuck with no UI' outcome this feature was meant to prevent. The verify-before-swap (index.html check) guards against bad archives but not against a failed move. Consider moving the staged tree into a sibling of `target` (same filesystem) and doing an atomic-ish rename, or copy-into-temp-then-rename, so the old bundle survives a failed swap.

Options

Display: compact → Showing less information.

Comment with these commands to change:

Compact

gitar display:verbose

---

Important

Your trial ends in 7 days — upgrade now to keep code review, CI analysis, auto-apply, custom automations, and more.

_{Was this helpful? React with 👍 / 👎 | Gitar}