finds.dev← search

// the find

ringhyacinth/Star-Office-UI

★ 7,400 · HTML · NOASSERTION · updated Mar 2026

A pixel office for your OpenClaw: turn invisible work states into a cozy little space with characters, daily notes, and guest agents. Code under MIT; art assets for non-commercial learning only.

A pixel-art office dashboard that visualizes AI agent work states in real time — idle, writing, researching, error, and so on — as animated characters moving around a tiny office scene. Built on Flask + Phaser.js, it integrates tightly with OpenClaw but works standalone via a simple HTTP API. Aimed at people running local AI agents who want a glanceable status display rather than log files.

The multi-agent guest system is genuinely well thought out — join keys with per-key concurrency limits, a push script visitors can drop in, and stale-state auto-reset to idle so the board doesn't lie when an agent crashes. Security got real attention for a hobby project: weak-password blocking, session cookie hardening, and a smoke test script that actually verifies the deployment. The split-license approach (MIT code, non-commercial art assets) is honest and correctly documented. Phaser.js for the game layer is the right call — canvas rendering with spritesheet animation is exactly what Phaser is for.

State is stored in a flat JSON file on disk, which means any concurrent writes from multiple agents will corrupt it — there's no locking mentioned and Flask's dev server is single-threaded by default, so this is a latent race condition in the exact multi-agent scenario the project advertises. The desktop pet switched from Electron to Tauri mid-development (both `electron-shell/` and `desktop-pet/src-tauri/` exist in the tree) and the README still says Electron in one place and the directory structure tells a different story — confusing for anyone trying to build it. The backend is a single `app.py` with no tests, and the only validation is the smoke test hitting live endpoints; the Gemini image generation is a network call with no timeout or retry documented. Not production-ready as a multi-user service, but that's probably fine given what it is.

View on GitHub →

// want more like this?

We dig through GitHub every week and send a few repos picked for what you actually care about — each with an honest take like this one.

Get finds in your inbox → Search again →