┌─────────────────────────────────────────────────────────────┐
│                 Webcam / Video Device Feed                  │
└──────────────────────────────┬──────────────────────────────┘
                               │ 640x480 @ 60 FPS (In-Memory)
                               ▼
┌─────────────────────────────────────────────────────────────┐
│                 Landmark Extraction Engine                  │
│             (21 3D Spatial Joints + Face Mesh)              │
└──────────────────────────────┬──────────────────────────────┘
                               │
                ┌──────────────┴──────────────┐
                ▼                             ▼
┌───────────────────────────────┐ ┌───────────────────────────┐
│     Vector Geometry Engine    │ │  1€ Adaptive Jitter Filter│
│  (Angles, Distance, Normals)  │ │ (Velocity-based smoothing)│
└───────────────┬───────────────┘ └───────────┬───────────────┘
                │                             │
                └──────────────┬──────────────┘
                               ▼
┌─────────────────────────────────────────────────────────────┐
│                 Gesture State Machine Engine                │
│    (Open Palm, Pinch, Point, Victory, Fist, Swipe, Dwell)   │
└───────────────┬─────────────────────────────┬───────────────┘
                │                             │
                ▼                             ▼
┌───────────────────────────────┐ ┌───────────────────────────┐
│   Interactive Web Studio      │ │ MCP Server for AI Agents  │
│   • Presentation Controller   │ │ • Claude Desktop          │
│   • Air Drawing Canvas        │ │ • Cursor / Cline / Zed    │
│   • Visual Filters/Telemetry  │ │ • Autonomous Tool Loops   │
└───────────────────────────────┘ └───────────────────────────┘

# Clone the repository
git clone https://github.com/google/vision-gesture-control.git
cd vision-gesture-control

# Install with pip
pip install -e .

python3 -m http.server 8000 --directory public

{
  "mcpServers": {
    "vision-gesture-control": {
      "command": "python",
      "args": ["-m", "vision_gesture_control.mcp"]
    }
  }
}

from vision_gesture_control.engine import GestureEngine

# Initialize the real-time gesture engine
engine = GestureEngine(camera_index=0, smoothing_factor=0.6)

# Register a custom callback for swipe events
@engine.on("SWIPE_RIGHT")
def handle_swipe_right(event):
    print(f"👉 Swiped Right with confidence {event.confidence:.2f}!")

# Start the non-blocking background tracking loop
engine.start()

vision-gesture-control/
├── .github/
│   └── workflows/
│       ├── ci.yml                 # 15-job cross-platform CI matrix
│       └── release.yml            # Wheel packaging, checksums & GitHub release
├── docs/
│   ├── GESTURE_RECOGNITION_GUIDE.md # Landmark math, vector geometry & smoothing
│   ├── MCP_GUIDE.md               # MCP server & client integration specification
│   └── CAMERA_SECURITY_PRIVACY.md # Local-first in-memory privacy & GDPR compliance
├── examples/
│   ├── presentation-controller/   # Touchless slide deck web application
│   ├── air-drawing-app/           # 3D spatial air painting canvas
│   ├── mcp-clients/               # Client config JSONs for Claude, Cursor, Cline, Zed
│   └── README.md                  # Examples directory guide
├── public/
│   └── index.html                 # Google Vision Studio (Material 3 Light UI)
├── src/
│   └── vision_gesture_control/    # Core Python package & MCP server
├── tests/
│   └── test_examples.py           # Comprehensive unit & integration test suite
└── README.md                      # This master documentation

PYTHONPATH=src pytest tests/test_examples.py -v