Skip to content

Burrow 0.7.0

Choose a tag to compare

View all tags
@github-actions github-actions released this 13 Jun 13:10
v0.7.0
0142edd

Burrow 0.7.0

The big one: a top-to-bottom redesign of every tool โ€” onboarding, a real Clean
review pipeline, a new Software tab, and rebuilt Status, popover, Analyze, and
Settings โ€” plus a wave of polish on top.

A redesigned app

  • First-run onboarding. A short, skippable intro: grant Full Disk Access
    (with a one-click relaunch), confirm the mo engine is installed and see its
    version, and a "free & open source" card. No more cold start.
  • Clean, reviewable before it runs. Scan now streams a live count-up โ€” the
    animation is the scan โ€” into a per-item review screen: tri-state category
    cards, honest "what this frees" lines, open-app badges, and a "Permanently
    clean ยท N GB" pill. Deselected items ride a fenced whitelist session for
    exactly one run and are restored byte-for-byte after. New Move-to-Trash
    mode     recycles only the reviewed items (recoverable) instead of permanent
    deletion.
  • New Software tab. Uninstall with an expandable per-app leftover review
    (auto-selects app/support/prefs/containers; flags caches/logs for review) and
    two-path removal. Updates โ€” one list with Sparkle / App Store / Electron /
    Homebrew badges (network only on click). Startup โ€” a read-only Launch
    Agents/Daemons inventory with problem rows.
  • Rebuilt Status dashboard. Corner chips, battery ring gauges (Bluetooth
    folded in), a low-space gradient bar, a read-only fan tile, a power column
    (honest "โ€”" where the kernel won't say), and an independently scrolling process
    table with a per-row Quit/Force-Kill menu.
  • Rebuilt menu-bar popover. A one-line health header + hardware chips, six
    metric tiles, a battery card with "โšก Top drain", a Stay-Awake / Wipe / Eject
    strip, and a Clean Watch footer.
  • Analyze: real scan progress. A true per-child counter ("โ— ~/Downloads ยท
    3/12") โ€” Burrow drives the loop, so the number is real, never invented.
  • Settings, reorganized. Tabbed panel โ€” General / Maintenance / Menu Bar /
    Advanced โ€” with a whitelist manager (Protected Items), Permanent|Trash removal
    mode, status-item Icon|Metrics mode, and global shortcut recorders.
  • Menu-bar tools. Keep Screen On (with durations), Clean Screen (Esc always
    exits), an About panel, and manual Check for Updates.

Charts & metrics

  • History charts as bars โ€” CPU usage, GPU usage, and health score render as
    clean, evenly-spaced bars at every range.
  • Real GPU usage on Apple Silicon, read natively and persisted (no more flat
    zero).
  • Tighter live tiles โ€” network sparkline windowed to the last couple of
    minutes, GPU as bars like CPU, and a new fan RPM-over-time sparkline.

Notifications

  • Finish-line notices when a real clean, optimize, or uninstall completes โ€”
    with what it freed.
  • Opt-in smart reminders โ€” low disk, full Trash, or "it's been a while since
    your last clean." Off by default and throttled so they never get chatty.

Polish & hardening

  • One-click Relaunch when Full Disk Access is granted mid-session.
  • Truthful Touch ID copy (it covers terminal sudo, not Burrow's own admin
    prompts).
  • Clean-review safety: whitelist session paths glob-escaped, unreadable
    whitelist aborts instead of overwriting, session always restored when a run
    ends.
  • Popover height tracks content; Wipe shows an armed state; deduped the doubled
    "macOS" version label.
  • ~230 new strings localized in ็ฎ€ไฝ“ไธญๆ–‡ and ็น้ซ”ไธญๆ–‡; accessibility labels and
    Reduce Motion on every new surface.

For agents & under the hood

  • Steadier metrics. A single malformed sample no longer blanks the
    dashboard โ€” bad rows are skipped, counted, and the latest good reading is
    used. GET /info and the burrow_info MCP tool now report decode-skip counts,
    so you can tell at a glance whether data is flowing.
  • One gated policy for actions. The app and the MCP server now decide "can
    this run?" through a single shared gate, and the agent-facing action tools
    return a structured result summary (space freed, items, categories) alongside
    the raw output, on a frozen, golden-tested JSON contract.
  • Lighter History. The History view's auto-refresh is demand-driven now โ€” it
    stops the moment you leave the view instead of leaving a timer running.
  • Deeper, more testable core. The mo runner (honest timeouts,
    runner-classified auth-cancel), the metrics reader, and the actions core were
    consolidated into deep modules with a large new test suite.
Assets 3
Loading