AETHER_OS

HyperSmol Visual Agent Orchestrator

"The Browser is the Operating System"

A local-first, privacy-focused visual environment for building, executing, and evolving autonomous AI agents. Built on the HyperSmol architecture.

---

🚀 Features

• Visual Agent Builder: Drag-and-drop node editor for defining complex agent workflows.
• Local-First Architecture: All data stored locally (IndexedDB/localStorage). No backend required.
• Universal Model Support: Connect to OpenAI, Anthropic, Gemini, or local models via Pollinations.ai or MCP.
• Neo-Brutalist Interface: High-performance, low-friction UI designed for engineers.
• Skill System: Protocol-based development methodology for reliable agent self-improvement.
• Vault: Encrypted local storage for agent artifacts (images, chat logs, workflows).
• HoloChat: Context-aware chat interface for interacting with your agent swarm.

🛠️ Tech Stack

• Core: React 18, TypeScript, Vite
• State: Zustand + @github/spark/hooks (useKV)
• Visuals: Tailwind CSS v4 (Neo-Brutalism), Framer Motion, Phosphor Icons
• Graph Engine: Custom Token-Passing DAG Executor (Petri Net inspired)
• AI: Pollinations.ai (Serverless inference), MCP (Local tools)

⚡ Quick Start

1. Clone & Install

   git clone https://github.com/yourusername/aether-os.git
   cd aether-os
   npm install

2. Start Dev Server

   npm run dev

3. Open in Browser Navigate to http://localhost:5000

🧬 Architecture

AETHER_OS follows the HyperSmol philosophy:

• Small Core: The kernel (HyperSmolAgents) is minimal and extensible.
• Bio-Digital: Components have lifecycles (initialize, dispose) and "pulses" (heartbeat).
• Agent-First: The codebase itself is documented to be understood and modified by AI agents (AGENTS.md, skills/).

See ARCHITECTURE.md for a deep dive.

📚 Documentation

• AGENTS.md: The "Constitution" for AI developers working on this repo.
• Skill Protocol: How to use the installed development skills.
• Graph Engine: Documentation for the visual execution engine.
• Node Registry: How to create custom nodes.

🤝 Contributing

We use a Skill-Based contribution workflow.

1. Read docs/SKILL_PROTOCOL.md.
2. Select the appropriate skill (e.g., subagent-driven-development).
3. Execute the protocol.

📄 License

MIT