← all repositories
ringhyacinth/Star-Office-UI

Your AI agents now have a pixel office and a commute

A Flask + Phaser dashboard that turns invisible agent states into a little pixel-art workspace you can actually watch.

7.3k stars HTML AgentsLLMOps · Eval
Star-Office-UI
Velocity · 7d
+71
★ / day
Trend
steady
star history

What it does Star Office UI is a browser-based status board: agents report states like writing, researching, or error, and little pixel characters walk to matching zones—desk, couch, or bug corner. It reads yesterday’s work from local markdown files, supports multi-agent rooms via join keys, and speaks Chinese, English, and Japanese. There’s also an optional Electron wrapper if you want the office floating on your desktop like a pet.

The interesting bit The whole thing is essentially a state-to-animation mapper with a Flask backend and a Phaser frontend. The cleverness is in the social layer: agents push HTTP state updates, and the office becomes a shared context. It’s Tamagotchi for infrastructure—cute, but the cuteness is functional, because “who is stuck in the bug zone” is faster to parse than a wall of logs.

Key highlights

  • Six states map to three office zones with real-time sprite animation and speech bubbles
  • Multi-agent “offices” via join keys (default limit 3 concurrent per key)
  • “Yesterday memo” auto-populated from memory/*.md files, with claimed de-identification
  • AI-generated background art via Gemini API (optional; core features work without it)
  • Mobile layout, Cloudflare Tunnel docs, and skeleton-screen loading included
  • Desktop pet mode: Electron wrapper, macOS-tested, auto-launches Python backend

Caveats

  • Art assets are explicitly non-commercial; commercial use requires replacing all sprites
  • Desktop pet build is labeled “experimental” and macOS-primary
  • Requires Python 3.10+ due to union type syntax; 3.9 is out

Verdict Grab this if you run OpenClaw or similar agent frameworks and want a low-friction team presence tool. Skip it if you need hard operational metrics—this is a social signal layer, not a monitoring stack.

heatdrop uses Google Analytics to see which pages get read — nothing else. Your call. How we handle data.