Releases: Elitelord/TaskManagerPlus
Release list
v2.7.2
TaskManager+ v2.7.2 — Faster, more reliable folder sizing
This release makes the Storage page's folder inspector faster and fixes a caching bug that could leave folder sizes blank on large drives.
Storage inspector
- Instant breakdowns after a scan — when you run a storage scan, the app now records each folder's subfolder sizes during the same pass it already makes to measure totals. Clicking a big slice (Downloads, Steam, a game folder) shows its breakdown immediately instead of re-scanning it.
- Sizes keep loading for folders a scan didn't cover — folders outside the scan still fill in progressively, one at a time, so you always see steady progress rather than a long blank wait.
Bug fixes
- Cached folder sizes no longer stop saving — on drives with many folders, the size cache could quietly hit its storage limit and silently stop caching anything new, so sizes appeared to "not stick." The cache now trims its oldest entries to stay within budget and saves progressively as it works.
Full Changelog: v2.7.1...v2.7.2
v2.7.1
TaskManager+ v2.7.1 — Storage accounting overhaul
The Storage page now measures and shows disk usage accurately. The "What's using space" ring is a true breakdown of the drive — every slice adds up to the space actually in use.
What's using space on C:
- Your profile is broken out — Downloads, Documents, AppData, OneDrive and the rest appear as their own slices instead of one opaque "Users" lump.
- The ring sums to real used space — slices, the Recycle Bin, system files, and a labelled remainder add up to what the drive reports as used. A caption appears if logical file sizes (compression, cloud files) exceed that.
Correct sizes
- Fixed several double-counting bugs where a folder and its own subfolders (or an app's install folder and its game library) were both counted, inflating totals.
- App sizes are more honest: sizes that came from a shallow or interrupted scan are now labelled as estimates rather than shown as exact.
- "Biggest files" lists now actually show the biggest files, and a cleanup card only claims space it can list.
Folder inspector
- Folder size caching is now reliable — sizes persist between visits instead of silently stopping once the cache filled up, and a slow folder keeps whatever finished if you navigate away.
One-time note
The first Storage scan after updating recomputes from scratch (the cached breakdown from the old version is discarded). Subsequent scans are fast.
Full Changelog: v2.7.0...v2.7.1
v2.7.0
TaskManager+ v2.7.0 — TaskManager+ Lite
A standalone native version of the app, shipped as a separate download alongside the usual installer.
TaskManager+ Lite
taskmanagerplus-lite-v2.7.0-win64.zip on this release. Unzip it and run taskmanagerplus-lite.exe — there is no installer, and it does not interfere with a normal TaskManager+ install.
- Processes — the full process table with sorting and search.
- Performance — live CPU, memory, disk, network, and GPU graphs.
- Battery — charge, health, and power draw.
- Settings — refresh rate (0.5s to 5s), which page opens on launch, and an always-on-top toggle. Stored separately from the main app's settings.
It opens immediately and uses a fraction of the memory of the full app, because it is a native window rather than a browser engine. The heavier pages — Storage, Insights, and the AI features — stay in the main app.
Which one to use
The main installer is unchanged and remains the complete app. Lite is worth it if you want a task manager that starts instantly and stays out of the way, or if you would rather not have a WebView-based app resident.
The zip holds two files and needs both in the same folder: the executable and taskmanager_native.dll. Nothing else is required — no WebView2, no runtime install.
Windows will show a SmartScreen "unknown publisher" prompt on first run, because the executable is not code-signed. Choose "More info" then "Run anyway".
Under the hood
- Lite is a separate process built on a native GUI toolkit, sharing the same telemetry backend as the main app rather than reimplementing it.
- The main app does not link any of the lite UI code, so its binary is unaffected.
- CI now builds the lite app on every commit and verifies the bundle layout before a release can be cut.
Full Changelog: v2.6.8...v2.7.0
v2.6.8
TaskManager+ v2.6.8 — Native tray widget
Left-clicking the tray icon now opens a small native widget instead of a second app window.
The widget
- Live system metrics — CPU, memory (used / total), GPU, disk throughput, network throughput, and battery, each with a colour-coded bar that shifts from green to orange to red as usage climbs. A footer line shows the running process count and current power draw.
- Anchored to the tray — the popup opens directly above the tray icon, clamped so it stays on screen. It handles scaled displays correctly, so it lands in the right place at 125% or 150% zoom.
- Click to toggle — clicking the tray icon again closes it. Double-clicking still opens the main window.
- Movable and resizable — drag anywhere in the body to reposition it, or drag an edge or corner to resize. It stays on top of other windows.
- Open button — jumps to the full app.
Why it is faster
The widget is its own lightweight process built on a native GUI toolkit rather than a second browser window. It launches immediately, uses a small fraction of the memory a WebView-based popup would, and reads telemetry directly instead of routing it through the web layer. The main app is unaffected — it does not load any of the widget's code.
Under the hood
- The widget ships alongside the app as a bundled resource and is resolved by its installed name first, so a leftover copy from an older build cannot shadow the shipped one.
- CI now builds and verifies the widget on every commit.
Full Changelog: v2.6.7...v2.6.8
v2.6.7
TaskManager+ v2.6.7 — App icons fixed in light mode
App icons no longer render with black corners in light mode.
The step that shrinks each icon to display size was discarding transparency, turning see-through pixels opaque. The app now performs that resize itself, reading the icon's own pixel data and preserving transparency exactly. Icons with no transparency to begin with are unaffected.
This is the change that resolves the black corners; the icon-related fixes in v2.6.5 and v2.6.6 addressed different problems.
Full Changelog: v2.6.6...v2.6.7
v2.6.6
TaskManager+ v2.6.6 — App icons fixed in light mode, and recent fixes actually reaching you
Two bugs, one of which was quietly hiding the other. App icons no longer render with black corners in light mode, and a stale component left behind by older installs is no longer loaded in place of the current one — which means the under-the-hood improvements from the last few releases are, for the first time, actually running on upgraded installs.
App icons in light mode
Every app icon in the Processes list sat on a black square in light mode. On the dark theme it was invisible, so it only ever showed for light-theme users.
The icons themselves were never the problem — they were correct, with proper transparency, the whole time. The rows they sit in were the issue. To keep long process lists fast the list only draws the rows you can see, and that technique hands each row to the graphics layer as its own surface. Those surfaces were set up with see-through backgrounds, and a see-through surface with nothing behind it renders as black. Anywhere an icon was transparent — the corners of a rounded logo — that black showed through.
Rows now have a solid background, with their striping and hover shading layered on top, so there is never an empty surface behind an icon.
The v2.6.5 notes described an icon fix that, in hindsight, addressed a different and much subtler issue and did not resolve this. This release is the actual fix.
Recent improvements now actually taking effect
A more consequential bug sat underneath that one.
The app loads a native component that does the heavy system-monitoring work. That file was renamed in an earlier release, but the app still looked for the old name first — and Windows installers only remove files they currently know about, so anyone who had upgraded rather than installed fresh still had the old file sitting next to the app. The app found it and loaded it, ignoring the current one entirely.
In practice that meant an install could be running a native component from May while reporting the latest version. Everything shipped in that component since then quietly did nothing on upgraded machines, including v2.6.5's two memory-leak fixes, the more accurate installed-app sizes, and the icon work.
The app now prefers the current file, and falls back to the old name only when that is genuinely what is installed. It also records which component it loaded, so this class of problem is diagnosable from a normal install rather than invisible.
If you have been on an upgraded install, this release is the first one where the last several rounds of fixes are actually active.
Faster startup
The interface is now split into separate pieces that load on demand, so opening the app no longer parses the code for pages you have not visited. The Storage, Insights and Settings pages load the first time you open them.
Full Changelog: v2.6.5...v2.6.6
v2.6.5
TaskManager+ v2.6.5 — A much lighter app, and app icons fixed in light mode
The headline here is memory. TaskManager+ used to reserve several hundred megabytes at launch for AI features you might never open in that session — it now starts at a fraction of that and loads them only when you actually reach for one. This release also fixes two slow leaks in long-running sessions, and includes the light-mode icon fix that was prepared for 2.6.2 but never shipped.
Starting up light
Previously, launching the app loaded the on-device AI models straight away: the search embedder and, on the Enhanced tier, the 379 MB writing model. Memory climbed from about 7 MB to roughly 820 MB within seven seconds of opening the app, whether or not you ever used an AI feature.
Those models now load on intent — when you open the command palette to search, or when you deliberately turn an AI tier on in Settings. A cold start now settles around 44 MB.
The first AI action in a session pays the load time that launch used to absorb. If you would rather have it ready up front, Settings → "Prewarm AI at startup" restores the old behaviour; it is off by default.
A few other paths were quietly loading the embedder as a side effect of unrelated work. Those have been closed too, so the models load when you ask for them and not before.
Leaks in long-running sessions
Two fixes that matter if you leave the app running in the tray for days rather than hours:
- Performance counter tables. The app watches GPU, network and disk activity through Windows performance counters using wildcard queries. Those queries quietly accumulate an entry for every instance they have ever seen — every GPU engine, every network interface — and never discard retired ones, so the table grows without limit inside a Windows system library. The app now rebuilds the query periodically, which drops the stale entries.
- Process detail cache. The cache holding per-application details (icons, publisher information) had no upper bound and grew with every distinct process seen over a session. It is now capped.
App icons in light mode
Every app icon — in the Processes list, in Frequent Apps, and in Startup apps — was drawn with its transparent areas filled in solid black. On the dark theme that is invisible, because black against a near-black page reads as transparent. In light mode each icon sat in an obvious black square.
The icons were being requested from Windows in a mode that flattens them against a mask instead of keeping their transparency. They now come back with their alpha channel intact and sit cleanly on either theme.
This dates back to v1.2.0 and was only ever visible to light-theme users.
Build and dependency cleanup
sharp, an image-processing library, has been removed. It was a leftover from Tauri v1 — whose command-line tool used it to generate app icons — and nothing in the app ever imported it. It carried four known vulnerabilities in the native image library it bundles. Two other build-time dependencies were updated to patched versions. None of these ever shipped inside the app, but the project now reports no known vulnerabilities.
Full Changelog: v2.6.1...v2.6.5
v2.6.1
TaskManager+ v2.6.1 — Crash fix for GPU-accelerated AI
A patch release fixing a crash introduced by the GPU generative backend. If you use the Enhanced AI tier with GPU acceleration, v2.6.0 could close the app without warning when opening the file inspector. Updating fixes it; no settings need changing.
The crash
Opening a file or folder in the inspector on the Enhanced tier asks the on-device writing model for a summary and name suggestions. With GPU acceleration active, that request could terminate the app instantly.
The cause was a mismatch in how the app described one of llama.cpp's configuration structures to the GPU library. The app's definition was 16 bytes shorter than the library's, so when the library filled it in, it wrote past the end of the space reserved for it and corrupted memory that belonged to something else. The next call then failed with an access violation.
Two details made this awkward to find:
- It only happened in release builds. Development builds lay out memory with enough slack that the overspill landed in unused space and nothing appeared to go wrong. Optimised builds pack memory tightly, so the same write hit live data. The shipped app crashed; the same code under a debugger did not.
- It looked like a timing problem. The crash arrived after a burst of activity, which pointed at concurrent requests. It wasn't — it was the very first request every time, and the burst was unrelated.
A second, quieter version of the same bug was fixed alongside it: the model-loading structure was missing a field in the middle, which shifted several later options by one position and meant the app was passing the wrong values for some of them.
Both structures are now checked against the library's real layout when the app is compiled, so this class of mismatch fails the build instead of shipping.
Logging
The app now writes its log to a file:
%LOCALAPPDATA%\com.taskmanagerplus.app\taskmanagerplus.log
Previously the log went only to a console, which a released build doesn't have — so when the app crashed, it left nothing behind to diagnose from. The log is flushed continuously so the final entry survives even an abrupt termination, and it starts a fresh file once it passes 5 MB. If you hit a problem worth reporting, this file is the thing to attach.
Build maintenance
The project's CI now runs on Node 24. GitHub is retiring Node 20 for Actions in September 2026, and the build was still pinned to it.
Full Changelog: v2.6.0...v2.6.1
v2.6.0
TaskManager+ v2.6.0 — Interface Revamp, Folder Inspector & Steadier On-Device AI
The biggest visual update since the app's first release: a pass over every page that replaces scattered one-off styling with a consistent design system. Alongside it, the file inspector becomes something you can actually browse, the Smart Organizer gets a set of new cleanup heuristics, and the on-device AI stops tripping over itself.
A redesigned interface
Every page has been reworked against a single set of design tokens rather than the ad-hoc values that had accumulated across them.
- A real data-visualization palette. Charts previously drew from two divergent hardcoded colour arrays plus around twenty one-off hex values scattered through the pages. There's now one source of truth, and it follows a rule: colour is only spent where it encodes something. Severity, thermal bands, fan profiles and memory cache tiers keep their hues because there the colour is the information; a file category or one app in a stack gets a lightness ramp instead. The result is charts that read as one system instead of a jumble of competing colours.
- Light theme actually retuned. The chart palette has its own light-theme values rather than reusing dark ones at lower contrast, so graphs stay legible on a white background.
- Consistent corners and type. A named radius scale (from hairline chips through to modals) replaces literal values that were written out by hand in ~74% of cases — including seventy that just restated a token's own value. Monospace readouts now share one font stack.
- Reworked startup and layout. The launch sequence and overall page framing were rebuilt on the same tokens.
- Light mode fixed throughout. Several surfaces referenced colour variables that were never actually defined, so they silently fell back to hardcoded dark values and stayed dark on a light background — the command palette, the Frequent Apps and routine cards, the file inspector, and a handful of chips and list rows. One had no fallback at all and rendered with no background whatsoever. All of them now use real theme tokens.
- Chart labels no longer clipped. The y-axis gutter was a fixed width that fitted "100 MB" but not a longer throughput reading like "10.3 KB/s", so the leading digits ran off the edge of the canvas on the Disk and Network graphs. The gutter is now measured from the labels themselves.
The folder inspector
Left-click any folder and the panel now lets you explore it rather than just describe it.
- Browse inside. The biggest subfolders and files are listed together, largest first, so what's eating space is at the top. Click a subfolder to go deeper; the breadcrumb walks you back up to any ancestor.
- Instant on return. A previously-visited folder renders immediately from cache — sizes and all — instead of re-scanning and showing a spinner. The fresh listing still runs behind it, so newly added or deleted files appear without the panel flashing empty.
- Refresh when you want it. A refresh button re-scans the folder and regenerates its summary on demand, for when something changed outside the app or the summary just isn't good enough.
- Summaries that keep up. A folder's AI summary is generated from its file names, so it's now tied to those names — add or remove files and the summary regenerates instead of describing a folder that no longer exists.
- Correct in light mode. The panel previously stayed dark regardless of theme, because it referenced a colour variable that was never defined and silently fell back to a hardcoded dark value.
Smarter Smart Organizer
- Screenshot pile-ups — a Screenshots or screen-recording folder that has grown to hundreds of files is surfaced as a cleanup candidate.
- Document buckets — recognises recurring themes (class notes, career files like resumes and transcripts, financial documents, receipts) and suggests consolidating scattered files of the same kind.
- Near-duplicate filenames — files differing only by a trailing number or version suffix ("Report v1 / v2 / v3") are grouped for review.
- Earlier duplicate detection — the per-group waste floor dropped from 50 MB to 10 MB.
- Per-folder re-scan now works. The "Re-scan just this folder" button refreshed the underlying scan but the drill-down kept serving its previously cached listing, so the numbers appeared unchanged.
Steadier, faster on-device AI
- No more request pile-ups. Every folder suggestion requested an AI folder name the moment it appeared, so a page of suggestions fired a dozen model calls at once. They serialised on the backend anyway — each holding a worker thread and building its own GPU context while it waited — which is what made the Storage page sit busy for around half a minute. Generation is now queued, and work whose panel has closed is dropped before it ever reaches the model.
- GPU backends no longer start simultaneously. On the Enhanced tier the writing model (Vulkan) and the search model (DirectML) could initialise at the same time on the same GPU. They're now brought up one at a time.
- Better diagnostics. Generation is fully logged, so a failure inside the GPU runtime can be traced to the exact call rather than disappearing silently.
When telemetry can't be read
An endless "Loading processes…" spinner was the only signal when system monitoring failed. There's now a proper notice covering both an outright probe error and a silent hang, explaining likely causes in plain language — a blocked or missing monitoring component, for instance — with a copy-to-clipboard details block for reporting it.
Privacy
Unchanged. All AI runs on your device; the writing and search models are opt-in and downloaded once. No telemetry, no analytics, no accounts. The only network requests the app makes are the updater checking GitHub for a newer version and, if you opt in, the one-time model download.
Full Changelog: v2.5.1...v2.6.0
v2.5.1
TaskManager+ v2.5.1 — Fixes app hanging on startup
A patch release fixing a startup hang that affected some machines.
The fix
- No more "Loading processes" freeze — on systems where Windows' hardware-info service responds slowly, a background query for AI-accelerator details could stall indefinitely. Because all hardware readings share a queue, that one stuck query blocked everything, leaving the app stuck on "Loading processes" forever.
- Bounded queries — those lookups now give up after a few seconds and continue without the optional detail, instead of waiting forever.
- Isolated failures — a single hardware reading that fails or is unavailable no longer prevents the rest of the app from loading. The process list now appears even if an optional sensor can't be read.
This mainly affected machines without a dedicated AI accelerator (NPU), particularly some AMD laptops.
Full Changelog: v2.5.0...v2.5.1