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: https://blober.io/

Architecture highlights

1. 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.

2. 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.

3. 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.

4. 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.

----

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:

• Simplicity of the application and the workflow model

• Whether this actually reduces "organizational debt" for large archives

• Where this breaks down compared to rclone or similar tools

Blog post on the build process: https://toknow.ai/posts/vibe-coding-challenge/

Blober.io: https://blober.io/