Skip to main content

Luna: Night-sky guide running Gemma on-device to get users outside

Discover Luna, a night-sky guide powered by Gemma 4 E2B via WebGPU in the browser, designed to verify every answer against the real sky and get you off your screen.

AI-written
Inewgen
12 Oct 2026Source: Dev.to3 min read (0 views)
Share
Luna: Night-sky guide running Gemma on-device to get users outside

Stock photo for illustration only, not from the actual event

Font size
  • Luna is a night-sky guide designed to pull users away from their screens and into the real outdoors.
  • It runs the Gemma 4 E2B model directly inside the browser using WebGPU and MediaPipe LLM Inference.
  • A dedicated verifier cross-checks every answer's name, direction, altitude, and time against astronomical data.
  • Created as a submission for Week 1 of the Hacktoberfest 2026 Open-Source AI Challenge.

While most stargazing applications keep users glued to their mobile displays, a developer has taken the complete opposite approach with Luna. Built for the Hacktoberfest 2026 Open-Source AI Challenge Week 1 theme 'Touch Grass', this application is engineered with a singular mission: to get you outside, looking at the real night sky for longer periods.

Luna functions as a night-sky guide that you set up by looking up, ask by pointing your phone, and finish by putting the device away. It is intentionally useless indoors; when accessed on a desktop computer, it immediately informs the user that it lives on mobile and provides a QR code to transfer the session. Whether on a city rooftop or a mountain pass, the app operates seamlessly.

smartphone mobile app night sky stargazing

Stock photo for illustration only, not from the actual event

Under the hood, Luna is powered by Gemma 4 E2B (released under the Apache 2.0 license as a litert-community web build), running entirely in-browser via WebGPU through MediaPipe LLM Inference. By leveraging the phone's compass, tilt sensors, GPS, and the sky above, the application processes astronomical data locally without relying on closed cloud APIs.

Utilizing open-weight models like Gemma directly on mobile devices represents a major milestone for local spatial AI. It eliminates connectivity barriers in remote outdoor environments such as dark-sky reserves or mountain trails, allowing users to explore astronomy freely while keeping personal interactions private and secure on-device.

The workflow begins when a user asks a question, which resolves a subject before the sky engine computes physical facts including altitude, direction, rise and set times. Gemma then writes an answer restricted strictly to those facts. Following this, a verifier checks every name, direction within plus or minus 45 degrees, height, and time within plus or minus 10 minutes. Passing results are marked as 'Checked against the sky', while failures display the raw engine reasoning openly.

Never miss the latest news?

Subscribe to get news summaries by email - not often enough to be annoying.

โฆษณา

"At Midnight, I took Luna onto a street in Bangalore. I held the phone up to a bright star low in the south-east and Luna named it as Sirius. It matched. I asked 'Where is jupiter rigth now?' and it auto-guided me and answered and help me spot jupiter in the sky. I Always wanted that! Never knew that star i saw so many times was JUPITER! Then I put the phone away and just looked up for a while embracing the beauty."

Luna Developer

The application features a dark, red-safe UI suited for night viewing, while its underlying libraries including Astronomy Engine and the Hipparcos catalogue via d3-celestial are fully open-sourced on GitHub for public collaboration.

Source: Dev.to

Comments

Leave a Comment
0/2000

Found something wrong in this article? Report an issue with this article