Skip to main content

Moondream Station

Moondream Station: Local Visual Reasoning

How It Works​

🚀 Launches Local Server
All inference runs on your device

🔧 Control via CLI
Caption images, answer questions, and manage settings

🌐 Access via HTTP
Connect to http://localhost:2020/v1 through REST or our Python, Node, or OpenAI client

Installation​

Install from PyPI:

$ pip install moondream-station

Install from source:

$ git clone https://github.com/m87-labs/moondream-station.git
$ cd moondream-station
$ pip install -e .

That's it! Moondream Station will automatically set itself up.

Usage​

Launch Moondream Station​

To fire up Moondream Station, execute this command in your terminal:

$ moondream-station

Model Management​

By default, Moondream Station uses the latest model your machine supports. If you want to view or activate other Moondream models, use the following commands:

  • models - List available models
  • models switch <model> - Switch to a model

Service Control​

We like to think Moondream has 20/20 vision; that's why, by default, we launch Moondream Station on port 2020. If that port is taken, Moondream Station will try a nearby free port. Additionally, you can control the port and status of the inference service with the following commands:

  • start [port] - Start REST server (default: port 2020)
  • stop - Stop server
  • restart - Restart server

Inference​

Video showing Moondream Station running in a terminal

Access via HTTP: Point any of our inference clients at your Moondream Station; for example, with our python client you can do:

import moondream as md
from PIL import Image

# connect to Moondream Station
model = md.vl(endpoint="http://localhost:2020/v1")

# Load an image
image = Image.open("path/to/image.jpg")

# Ask a question
answer = model.query(image, "What's in this image?")["answer"]
print("Answer:", answer)

For more information on our clients visit: Python, Node, Quick Start

Connect via CLI: Use all the capabilities of Moondream directly through your terminal. No need to touch any code!

  • infer <function> [args] - Run single inference
  • inference - Enter interactive inference mode

Settings​

Control the number of workers, queue size, privacy settings, and more through Settings:

  • settings - Show configuration
  • settings set <key> <value> - Set setting value

Moondream Station collects anonymous usage metrics to help us improve the app. The following data is collected:

  • Event data: when you use features like caption, query, detect, or point.
  • Version information: active bootstrap, hypervisor, inference client, and model version.
  • System information: OS version, IP address, and Python version/runtime.

No personal information, images, or prompts/responses are ever collected. To opt out of logging, run: settings set logging false.

Utility​

The utility functions provide insight into what Moondream Station is currently doing. To view statistics for your current session, use the session mode. To view a log of requests processed by Moondream Station, use the history command.

  • session - Show session stats
  • help - Show available commands
  • history - Show command history
  • reset - Reset app data & settings
  • clear - Clear screen
  • exit - Quit application