A small home server can be doing several jobs at once. Opening a collection of administration pages makes it harder to spot the one thing that needs attention.
The build
What makes it work.
A compact monitoring website brings server health, application memory, playback activity, backup age and compression progress into one consistent view. A separate authenticated control layer handles selected actions.
Health with context. Memory use is compared with each application's actual limit. Backup ages and stale desktop data become visible warnings rather than unexplained numbers.
See the work happening. Playback and transcoding activity sit alongside compression progress, waiting work and failures, so the relationship between services is easier to understand.
Small screen, useful overview. Grouped cards, clear status colours and responsive sizing were designed to keep the everyday overview compact.
Controls with boundaries. Protected actions use an allowlist, confirmation where appropriate and an audit trail. Desktop commands have a waiting list, cancellation and expiry.
What it achieves
Built for the everyday.
A practical operations view for this particular system, with both monitoring and guarded controls. It is not an enterprise monitoring service or a promise that every fault is detected.
Monitoring capture only. Private identifiers and operational controls are removed from portfolio images.
The real server health panel, with private identifiers removed. Actual interface · privacy reviewedThe real storage overview, with private paths and identifiers removed. Actual interface · privacy reviewed
How it worksAnimated explanation
01Collect signals
02Check health
03Show context
04Review safely
Illustrated sequence · no live connection or performance figures
02 / Featured projectInfrastructure & engineering
Media Production Controller
A queue-driven compression workflow that treats verification and recovery as part of the job.
Long-running media compression needs more than an encoder command. Interrupted work, poor savings, storage pressure and an unsafe replacement can matter as much as the conversion itself.
The build
What makes it work.
A Windows controller uses a persistent SQLite queue to coordinate bounded FFmpeg/HEVC work. Jobs pass through explicit preparation, encoding, validation, staging and verification states before completion.
Persistent ownership. Transactional claims and leases give a job one owner. State changes leave a history, and completed work is not automatically claimed again.
One encode at a time. The workflow limits parallel work and checks storage headroom. Night windows and pause-after-current controls let it coexist with everyday use.
Useful savings, not smaller at any cost. The production contract requires meaningful storage savings as well as successful output checks. Unsuitable or insufficiently beneficial candidates can be skipped.
Verification before replacement. Duration, stream and decode checks precede staged hash checks and a guarded replacement. The media server is checked again before a job becomes complete.
What it achieves
Built for the everyday.
A repeatable processing system with explicit checkpoints, failure handling and operator control. Savings and speed vary with the source; no universal performance or quality percentage is claimed.
Savings vary by source. There is no invented percentage or promise that every file will be smaller.
How it worksAnimated explanation
01Queue
02Encode
03Validate
04Safe replace
Illustrated sequence · no live connection or performance figures
03 / Featured projectApplication
CluckyLens
A technology briefing that makes room for context and keeps the original reporting close at hand.
A headline feed is easy to scan and easy to forget. The aim was a calmer way to understand consequential technology developments without filling the page with routine stories.
The build
What makes it work.
A shared briefing service supplies Windows, Android and web interfaces. Short summaries, why-it-matters context, topic filters and original-source links turn the same edition into a useful reading experience across screens.
A rolling reading window. Only stories from the latest 24 hours remain visible, including cached offline stories. Reading progress is stored on each device.
Thirteen ways into the news. Topic filters cover AI, software, security, chips, cloud, robotics and more. Twenty is a ceiling, not a filler target.
Checks without needless rewriting. Normal feeds are checked hourly. Changed, relevant input can trigger a new brief, within a daily editing budget; unchanged input does not need another AI pass.
Fast events, cautious handling. A lightweight critical-source monitor checks more frequently without AI. Verified high-impact events can reach the owner through the existing notification channel; routine reports do not trigger alerts.
What it achieves
Built for the everyday.
Released Windows and Android clients with a responsive shared interface, device-local reading state and signed update metadata. Editorial selection is still being refined, and generated summaries can miss context.
Descriptions reflect documented source and release evidence. Illustrations explain behaviour; they do not show live measurements.
The real connection settings, shown without changing configuration. Actual interface · privacy reviewedThe responsive web interface at a phone width, not a physical-device photo. Actual interface · privacy reviewed
How it worksAnimated explanation
01Source feeds
02Significance filter
03Editorial brief
04Reader
Illustrated sequence · no live connection or performance figures
04 / Featured projectApplication
CluckyMedia Desktop
A personal-media experience with a consistent identity from the first launch to the next update.
A useful media interface should feel coherent across a browser, a Windows desktop and an Android device. Installation and update screens should feel like part of that experience too.
The build
What makes it work.
A customised Jellyfin interface is paired with Windows and Android clients. The work combines product styling, playback integration, startup behaviour and a shared approach to release notes and updates.
A recognisable family. The C-family mark, navy and indigo palette, and readable controls carry through the app, installer and updater.
More than a browser tab. The Windows wrapper includes desktop lifecycle handling and a playback hand-off path. Android adds a dedicated app container and update flow.
Care around the edges. Connection recovery, failed-update notices, startup behaviour and clear release notes are treated as user-facing work.
Built on an existing foundation. Jellyfin supplies the underlying media platform. This project customises and integrates it rather than claiming to have invented a media server.
What it achieves
Built for the everyday.
A privately released set of clients and a custom web presentation. Playback depends on the source, client and server; this portfolio does not claim every device or codec combination is verified.
A customised Jellyfin client experience. Jellyfin remains the underlying media platform.
The actual installation progress screen during the same upgrade test. Actual interface · privacy reviewed
How it worksAnimated explanation
01Shared design
02Windows
03Android
04Web
Illustrated sequence · no live connection or performance figures
05 / Featured projectApplication
PixelRemote
A remote desktop shaped around the realities of using a large screen from a small one.
A phone and a desktop have very different shapes and input methods. Monitor changes, zoom, reconnects and pointer releases need to stay predictable when the user moves between them.
The build
What makes it work.
An Android remote desktop app derived from RustDesk, with targeted changes to input, display switching, reconnect behaviour, file-transfer access and the frame-rendering path.
Monitor changes that make sense. Work on aspect handling and stale-monitor input guards helps the phone stay aligned with the selected desktop display.
Zoom that survives a reconnect. Zoom behaviour and persistence were adjusted so returning to a session does not unnecessarily discard the chosen view.
A lower-copy rendering path. The project includes work that sends frames into a native texture for Flutter to display, with the older rendering route kept as a fallback.
File access with a purpose. A file-transfer entry point makes moving between phone and PC a visible part of the interface. Platform permission handling remains part of the flow.
What it achieves
Built for the everyday.
Version 1.7.12 was released privately with selected physical-device checks recorded. RustDesk remains the underlying remote-access foundation; not every inherited menu feature is proven end to end.
Built on RustDesk. Selected device checks are documented; this is not a claim that every feature has been tested.
How it worksAnimated explanation
01Connect
02Choose a display
03Send input
04Reconnect
Illustrated sequence · no live connection or performance figures
02 / Beneath the surface
The quieter systems doing the heavy lifting.
The things that help the main experience stay safe, recoverable and easy to use.
An encoder exiting successfully does not establish that a long media file is complete, that its streams survived, or that a copied candidate is the exact file that was checked.
The build
What makes it work.
Validation and recovery tools sit around the compression workflow. They compare identity, duration and stream structure, perform bounded decode checks on the Windows PC, and verify staged content before replacement.
Several checks, different failure modes. A readable header, a plausible duration and a clean decode answer different questions. The workflow combines checks rather than treating one success signal as sufficient.
Know which file passed. Hashes connect the validated candidate to the staged and installed copies. A mismatched or ambiguous file is held for review.
Recover the checkpoint. Retained completed candidates can be reconsidered through validation and guarded replacement. Recovery does not blindly trust an old output or silently overwrite a conflicting stage.
Protect the original during change. The replacement path keeps an original available during verification. Its later retention follows the production batch policy; this is not an indefinite backup promise.
What it achieves
Built for the everyday.
A supporting safety system, not a separate consumer app or a guarantee of perceptual losslessness. Historical incident analysis informed its checks; no universal quality score is asserted.
Descriptions reflect documented source and release evidence. Illustrations explain behaviour; they do not show live measurements.
How it worksAnimated explanation
01Inspect
02Check integrity
03Pass or retain
04Recover
Illustrated sequence · no live connection or performance figures
A download button is only the beginning. People also need to understand what changed, see progress, verify the update and retain a working path when installation fails.
The build
What makes it work.
Branded Windows installers and matching in-app update screens combine grouped release notes, version checks, verified metadata and download integrity checks. Android hands installation to the operating system.
Readable release notes. Short New, Improved and Fixed groups explain the user-facing change without exposing implementation details.
Verify before running. Signed update metadata and a file hash are checked before installation. CluckyLens binds its product, platform and version into the signed offer.
Keep the choice visible. Update now and Later, progress views and clear failure messages make the process understandable without forcing an immediate install.
Shared style, separate identity. Apps reuse the same presentation language while keeping their own icon, app identity and signing continuity.
What it achieves
Built for the everyday.
A shared delivery pattern used by the released applications. Signed update metadata is not the same as a Windows publisher certificate, and the draft should not imply the installers have one.
Descriptions reflect documented source and release evidence. Illustrations explain behaviour; they do not show live measurements.
How it worksAnimated explanation
01Read release
02Verify signature
03Check file
04Confirm install
Illustrated sequence · no live connection or performance figures
Source control cannot preserve every local working detail, and a backup is hard to trust if an incomplete archive looks finished or a failed run disappears quietly.
The build
What makes it work.
A focused PC-only backup path streams selected irreplaceable data to an encrypted archive on the server. Verification precedes promotion, while bounded logs and failure reporting make the job observable.
Focused rather than indiscriminate. The source script selects local-only working data and undo copies while excluding regenerable caches and large transient directories.
Stream instead of staging plaintext. The PC sends the archive stream to the protected destination without writing an intermediate plaintext archive on its own disk.
Finish only after checking. The destination writes a partial encrypted archive, checks that it can be decrypted and read, then promotes the completed artifact and records a checksum.
Visible failures and bounded history. Connection keepalives, explicit exit checks, retained logs and a retention policy help distinguish a healthy backup from a stalled or failed attempt.
What it achieves
Built for the everyday.
A source-verified backup design with separate monitoring of recent success. This portfolio does not claim an independent/offsite backup, current restore drill, or full disaster-recovery guarantee.
Descriptions reflect documented source and release evidence. Illustrations explain behaviour; they do not show live measurements.
How it worksAnimated explanation
01Inspect
02Check integrity
03Pass or retain
04Recover
Illustrated sequence · no live connection or performance figures
An installed-looking web experience can still fail on a weak connection or keep serving old code. Caching needs boundaries as well as a fast path.
The build
What makes it work.
A versioned service worker caches content-hashed static assets, keeps changing application data on the network and provides an offline navigation fallback. An update-ready view gives the user a deliberate reload path.
Cache the right things. Hashed static files can be cached; fresh navigation and changing API/configuration responses are treated differently.
Make disconnection understandable. A dedicated offline page explains the failed navigation instead of leaving the user with a blank screen.
A visible update handover. The app can check for a new release when it returns to the foreground and offer an update-ready reload.
Keep rollback practical. Versioned caches and a network-loaded entry page help prevent the frontend from being trapped on an obsolete release.
What it achieves
Built for the everyday.
A delivery component of CluckyMedia, not another standalone app. The earlier in-app Install app prompt was deliberately removed; do not present it as a current feature.
Descriptions reflect documented source and release evidence. Illustrations explain behaviour; they do not show live measurements.
How it worksAnimated explanation
01Load online
02Cache shell
03Network drops
04Offline fallback
Illustrated sequence · no live connection or performance figures
03 / A connected approach
The pieces work together.
Individual projects solve specific problems. Monitoring, automation and interfaces serve different needs. This map groups the ideas behind the work; each project has its own role.
CluckyWorksUseful software. A personal touch.
Read & discover
Watch & enjoy
See & understand
Automate & protect
Conceptual relationship map · not infrastructure topology
04 / Also in the workshop
More of what gets built.
Smaller pieces that sit inside the bigger projects. They are listed here because they are real work, not separate products.
Customised Jellyfin web client
A modified web client for a personal Jellyfin server: request and discovery rows in a Seerr-style design, a quality selector, season pills, hideable home rows and TV-remote navigation.
Built on the open-source Jellyfin web client and Seerr. Shipped as numbered versions, each with a documented rollback.
Windows and Android packaging
The same client wrapped as a Windows installer and an Android app, with a signed update check, a branded installer and an offline fallback page.
Installer upgrades are rehearsed in Windows Sandbox before release.
The icon family
Every product shares one C-ring mark on a dark rounded tile; only the symbol inside changes. Icons are generated from geometry and checked at 16, 32 and 48 pixels.
Flat two-dimensional marks, one idea per icon.
Watchdog and alerts
Scheduled checks and a watchdog raise a phone alert when something fails, instead of waiting for it to be noticed. Repeat alerts are rate-limited so a real problem is not buried.
Personal-scale monitoring, not an enterprise service.
05 / How the work gets built
Habits behind every release.
Independent projects, designed, built, released and looked after by one person. These are the working rules that keep them dependable.
Preview before publish
Anything people will see is built and reviewed locally first. Nothing visible goes out on the strength of a passing test alone.
Every change has an undo
A change log records what changed, the backup taken first and the exact steps to reverse it.
Test, then verify live
Services ship with automated tests, and changes are checked again on the running system before they count as done.
Updates wait for a quiet moment
Web releases check server load and active playback first, and a failed release can be rolled back to the previous version.
Private by default
No tracking on this site. Screenshots are privacy-reviewed, and source code lives in private repositories unless it is meant to be shared.
Toolbox
Python
PowerShell
JavaScript & TypeScript
HTML & CSS
SQLite
FFmpeg / HEVC
Electron
Capacitor (Android)
Docker on Linux
Git
06 / Behind the work
Built around real use.
CluckyWorks brings together the personal software projects of Clucky7. Some are apps you open every day. Others quietly check, organise and protect the work behind them.
Each story here focuses on a real need, documented engineering and honest limits. Clucky is me; CluckyMedia is my media server project. CluckyWorks is the home for the wider collection, with credit to the open-source projects behind it.