openeye-sh

Ollama for vision AI models — pull, run, and serve computer-vision models from your terminal.

Install

pipx install openeye-sh

With extras:

pipx install "openeye-sh[yolo]"        # YOLO object detection
pipx install "openeye-sh[depth]"       # Monocular depth estimation
pipx install "openeye-sh[camera]"      # OpenCV camera support
pipx install "openeye-sh[all]"         # YOLO + depth + grounding + camera

Quick Start

# List available models
openeye list

# Pull a model
openeye pull yolov8

# Run inference on an image
openeye run yolov8 image.jpg

# Start the API server
openeye serve yolov8

# Live camera stream with detection
openeye watch --models yolov8

Features

• Model management — pull, list, and run vision models (YOLO, depth, grounding, SAM, RF-DETR)
• REST & WebSocket API — FastAPI server with /predict, streaming WebSocket endpoints
• Live perception — real-time camera inference with VLM reasoning
• Fleet control plane — manage distributed vision agents
• Agentic loop — continuous perception + reasoning + planning pipeline
• Safety guardian — configurable safety zone monitoring for robotics

Links

• Homepage
• Repository