Features

Everything you need.
Nothing you don't.

Genesis OS ships with a complete set of applications, an animated AI assistant, persistent memory, voice control, and agentic tooling — all running locally on your hardware.

Get Started Explore Features ↓
AI Orb

Your AI is always
one click away

The AI Orb lives in the bottom-right corner of your screen at all times — always accessible, never obtrusive. Four animation states communicate exactly what Genesis is doing without a single word.

✦
Idle

Slow breathing pulse glow. Genesis is calm and ready.

🎤
Listening

Concentric ripple waves. Voice input active.

🧠
Thinking

Neural-net particle orbit. Processing your request.

🔊
Speaking

Waveform visualiser. TTS response playing.

  • Click to open the floating AI chat panel
  • Hold to activate push-to-talk voice input
  • Implemented with CSS animations + Framer Motion — no WebGL
  • Always rendered on top of every window
AI Orb — States
🪐

Screenshot: AI Orb animation states — assets/images/ai-orb-screenshot.png

Desktop Shell

A complete OS experience
in your browser

The Genesis OS shell renders entirely in React — no native dependencies, no Electron required for Docker deployments. Every component is designed with glassmorphism depth and fluid 60fps motion.

🖥️

Desktop

Wallpaper, desktop icon grid, right-click context menu. Spatial file launching and drag-and-drop support.

⬛

Taskbar

Dock at the bottom. Running app indicators, system clock, notification tray, and the AI Orb. Always at the bottom layer.

🪟

Window Manager

Draggable, resizable floating windows. Glassmorphism borders. Snap to edges. Z-order management. Multiple windows side by side.

🔍

App Launcher

Cmd+Space (or click the orb) for full-screen spotlight-style search. Instant fuzzy search across apps, files, and AI commands.

🔔

Notification Panel

Slides in from the right. Receives AI messages, system events, scheduled proactive updates, and tool execution confirmations.

🎨

Design System

Single configurable accent colour (default: electric violet). Glassmorphism panels. Dark-first theme. Fully themeable via CSS tokens.

Genesis OS — Window Manager
🪟

Screenshot: Window Manager with multiple apps — assets/images/window-manager-screenshot.png

Built-in Applications

Everything you need,
shipped on day one

📁
File Manager
React + /api/fs

Tree sidebar navigation, icon grid view, drag-and-drop file operations, preview pane. Ask the AI to find, summarise, or act on any file in natural language.

📂 Browse ✂️ Cut/Copy/Paste 👁️ Preview
⬛
Terminal
xterm.js + WebSocket PTY

Full PTY terminal over WebSocket. Ask the AI to explain error output, suggest commands, or execute entire shell scripts with your approval. Connects to daemon /api/shell.

🔧 Full PTY 🤖 AI assist ✅ Approval gate
📄
PDF Viewer
@react-pdf-viewer + pdfjs-dist

Native PDF rendering with page navigation, zoom, and text selection. Open any PDF from the file manager or by asking the AI to "open invoice.pdf".

📖 Render 🔍 Search 🤖 Summarise
📊
Office Viewer
docx-preview + SheetJS

Opens .docx, .xlsx, and .pptx files without Office installed. Read-only preview with faithful rendering. Ask the AI to extract data, compare revisions, or summarise content.

.docx .xlsx .pptx
📝
Text Editor
Monaco Editor (VS Code engine)

The same engine powering VS Code. Syntax highlighting for 60+ languages, IntelliSense, multi-cursor editing. AI can edit code inline on request.

60+ langs AI inline edit Multi-cursor
🌐
AI Browser
React iframe + /api/browse

Browse local and remote pages through a proxy with a URL bar. Click "Ask AI about this page" to get an instant summary, extract data, or ask questions about any web content.

🔗 Proxy ✨ AI Summary ❓ Ask about page
⚙️
Settings
React form + /api/config

Model picker (change Ollama model live), wallpaper selection, accent colour customiser, voice toggle, approval mode, and user profile configuration.

🦙 Model picker 🎨 Theming 🔧 Config
📋
System Logs
Real-time SSE + /api/logs

Live stream of daemon logs, LLM calls, tool executions, and memory retrievals. Colour-coded log levels. Filter by service. Essential for debugging and transparency.

📡 Live stream 🎨 Colour coded 🔍 Filter
🔨
App Builder
AI-generated React components

Describe an app in plain English. Genesis generates a fully functional React component, sandboxes it, and registers it in the app registry — without reloading.

✨ AI generate 🪄 Live sandbox 📦 Auto-register
1
Message received — user says "open that contract I was editing yesterday"
2
ChromaDB retrieval — semantic search finds 5 relevant past interactions including file edit history
3
Context injection — memories prepended to LLM prompt. 128K context window, compression if full
4
Grounded response — AI opens the correct file and says "I found contract_v2.docx from Tuesday"
5
Stored to memory — this interaction appended to SQLite + ChromaDB for future retrieval
Memory System

AI that
actually remembers

Every single conversation, file access, and context is permanently stored in ChromaDB (vector) and SQLite (structured). When you ask Genesis anything, it retrieves the most relevant memories before responding — so it always knows what you mean.

  • sentence-transformers embeddings for semantic memory search
  • chokidar file watcher auto-indexes your entire workspace continuously
  • Context compression (packages/memory/compress.py) when 128K window fills
  • Append-only — memories are never silently deleted without your consent
  • Identity profile: your name, preferences, projects, and habits
Memory Architecture →
Voice Control

Talk to your OS.
Privately.

Full hands-free operation with a wake-word, local speech-to-text, and natural-sounding text-to-speech. Everything runs on-device — no audio ever leaves your machine.

  • "Hey Genesis" — openWakeWord always-listening detection
  • faster-whisper STT — tiny/base/small/medium model choices
  • Piper TTS — natural voices, no cloud, zero latency
  • Web Audio API spectrum visualisation in the browser
  • Screenshot → multimodal vision (Gemma 4 supports images natively)
  • Disable voice entirely with one env flag
Voice Architecture →
Voice Control Interface
🎤

Screenshot: Voice interface — assets/images/voice-screenshot.png

Agentic Tool Use

More than a chatbot —
an agent that acts

Genesis uses Gemma 4's native function calling to execute real actions on your behalf. Every destructive operation waits for your explicit approval before proceeding.

Built-in Tools

read_file Read any file in the workspace Safe
write_file Write or patch a file Approval
run_shell Execute a shell command Approval
search_web Search via local SearxNG instance Safe
send_email Send email via configured SMTP Approval
browse_url Fetch and summarise a URL Safe

Plugin System

Any tool. Same interface.

Every tool — built-in or custom — implements the same interface. Drop a file in packages/tools/plugins/ and it's instantly available to the AI.

my-tool.js 
module.exports = {
  name: 'my_tool',
  description: 'Does something useful',
  requiresApproval: false,
  schema: {
    type: 'object',
    properties: {
      input: { type: 'string' }
    }
  },
  async execute({ input }, context) {
    // context = { bus, projectRoot }
    return `Result: ${input}`;
  }
};
Agentic Loop — Tool Execution
🔧

Screenshot: Agent loop with approval UI — assets/images/agent-screenshot.png

Model Agnostic

Switch models.
Zero code changes.

Genesis never hardcodes a model name. One environment variable swap and you're running a completely different LLM — all tool schemas, memory retrieval, and function calling adapt automatically.

🔷
gemma4:e4b

Default. 4B effective params (MoE). 9.6 GB. 128K context. Multimodal. Fast CPU.

Default
⚡
gemma4:e2b

2B effective params. 7.2 GB. Fastest CPU option. Low RAM machines.

Lightweight
🏔️
qwen3:8b

Excellent reasoning. Great for code. Strong multilingual support.

Code-focused
🦙
Any Ollama model

Mistral, Phi-4, LLaMA, Deepseek — any model in the Ollama library works instantly.

Your choice
docker-compose.yml — switch models 
# Change ONE variable to switch models
environment:
  GENESIS_MODEL: gemma4:e2b   # ← change this
  OLLAMA_BASE_URL: http://ollama:11434
  GENESIS_PORT: 3000
  GENESIS_VOICE_ENABLED: true

Ready to try every feature?

Launch Genesis OS with Docker and explore all features in under 5 minutes.

Launch with Docker See Use Cases