Code - Data - People | Software Engineering | Data Science | ML/AI | ToKnow.ai | Oark Library | mckabue.com

Kabue Charles  • Peerlist

A local-first workflow engine for cloud storage transfers with metadata-aware path organization. The easiest way to transfer files between cloud providers.----For the past few months I've been managing over 1TB per month of media archives using custom Python scripts. It worked, but the operational debt kept growing-regex-heavy metadata pathing (eg: container/{file_created_date}/{file_created_datetime}:{filename} ), async I/O quirks across OSes, and fragile long-running transfers became harder to reason about than the actual data movement.Over the past Christmas break, I challenged myself to replace that setup with a state-aware, local-first workflow engine. The result is Blober: <a target="_blank" rel="noopener noreferrer" class="break-all link-break text-gray-500 hover:underline" href="https://blober.io/">https://blober.io/</a><h3>Architecture highlights</h3><ol><li>Local-first store Blober isn't a CLI wrapper. It persists configurations, workflows, tasks, and state in a local SQLite DB so transfers can be resumed, inspected, and replayed deterministically.</li><li>Immutable jobs Each run snapshots configuration into a Job object. Once started, the job doesn't change, even if the user edits the workflow later.</li><li>Background executor A background loop (Electron-based) manages long-running transfers and prevents system sleep during multi-hour jobs. This was critical for my use case and a advantage over rclone.</li><li>Metadata-aware paths Files are inspected at runtime (stat() or provider APIs), allowing variables like {file_created_date} or {file_size_mb} to be injected into destination paths without pre-scanning. This is the biggest advantage over rclone.</li></ol>----I built the desktop app in ~14 days using heavy AI assistance. What worked surprisingly well was forcing a planning gate before any code generation: having the model produce a phased implementation plan first reduced over-engineering and large hallucinated files.I'm looking to extend this to support all cloud providers and use cases, so any feedback is welcome.I would also love brutal feedback, especially on:<ul><li>Simplicity of the application and the workflow model</li><li>Whether this actually reduces "organizational debt" for large archives</li><li>Where this breaks down compared to rclone or similar tools</li></ul>Blog post on the build process: <a target="_blank" rel="noopener noreferrer" class="break-all link-break text-gray-500 hover:underline" href="https://toknow.ai/posts/vibe-coding-challenge/">https://toknow.ai/posts/vibe-coding-challenge/</a>Blober.io: <a target="_blank" rel="noopener noreferrer" class="break-all link-break text-gray-500 hover:underline" href="https://blober.io/">https://blober.io/</a>

Blober.io

A digital education hub serving over 130,000 users with 400,000+ learning resources. - Engineered the platform from scratch - web, mobile, and backend - Designed robust security against repeated cyberattacks - Achieved peak growth of 1,500 daily signups

Oark Library

Simplex Router https://github.com/McKabue/simplex-router Common C# (C Sharp) Utility Methods https://github.com/McKabue/Common-C-Sharp-Utility-Methods

Open source projects

A platform for publishing data-driven research and analysis in AI, security, finance, and technology. Visit Site