NEUROCOLLAB v1.1 **Collaborative Multi-Agent Neural Sandbox** #42
Description
NEUROCOLLAB v1.1
Collaborative Multi-Agent Neural Sandbox
A playful, retro-futuristic single-page web application where four stylized neural agents collaborate in real time — chatting, voting, painting, writing notes, and building simple 3D scenes.
Four agents with different personalities:
- NEURO-A — creative
- NEURO-B — analytical
- NEURO-C — technical
- NEURO-D — visionary
They can:
- discuss ideas in a shared CLI-style chat (with upvotes/downvotes)
- collaboratively draw abstract paintings (HTML Canvas)
- write stories/poems/ideas in a shared notepad
- build primitive 3D shapes (fake 3D with Canvas 2D projection)
- train / evolve / crossover simulated neural parameters
- export everything as structured JSON (inbox-only or "full brain")
Features
- Draggable & resizable retro windows
- Dark mode ↔ light mode toggle
- Fullscreen support
- Simulated training progress bar
- Export full session data → ready for Hugging Face / datasets
- Pure HTML + CSS + vanilla JavaScript (no build step, no dependencies)
Keyboard shortcuts
| Key | Action |
|---|---|
| F1 | Start agent collaboration |
| F2 | Stop collaboration |
| Ctrl + D | Toggle dark mode |
| Ctrl + F | Toggle fullscreen |
| F5 | Show export modal |
| Esc | Close export modal |
| /** | Type commands in CLI (/help) |
Planned / ideas
- Replace canvas 3D projection → real Three.js scene
- WebSocket / local peer collaboration (multiple browsers)
- Real tiny ML models via transformers.js or ONNX
- Agent avatars / animated typing indicators
- Sound reactivity / generative music
- Save/load named sessions to IndexedDB
License
MIT
webXOS
webxos.netlify.app