DEV Community

Niranjan Pal
Niranjan Pal

Posted on

AstroGrass AI — Dark-Sky Stargazing & Meteor Lookout

#devchallenge #hf26challenge

Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass Submission 🌿

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass

What I Built

AstroGrass AI is an offline-first dark-sky stargazing and meteor lookout companion designed around a strict anti-screen philosophy: getting people to lie down on open grass under the autumn night sky without blinding themselves with phone screens.

Modern astronomy applications paradoxically trap stargazers into staring down at glowing screens. Screen blue light bleaches rhodopsin in human retinal rod cells, destroying dark adaptation for up to 30 minutes.

AstroGrass AI flips this dynamic using open-weight models via Ollama:

  1. You select your celestial targets (October Orionid meteors, Cassiopeia/Pegasus autumn constellations, Saturn/Jupiter, or deep-sky objects like the Pleiades and Andromeda Galaxy M31) and your Bortle darkness level.
  2. The AI generates a 3-step naked-eye visual finding guide in 15 seconds.
  3. You activate Astro Red Mode, which converts the entire interface into deep scotopic red (#ff4444 on #070000), preserving rhodopsin.
  4. You start the Hands-Free Dark Adaptation Timer: a 15-minute countdown that rings gentle harmonic Web Audio singing-bowl bells at 5-minute milestones so you can keep your phone face down on the grass.
  5. A rapid pocket meteor tally button and offline Grass-Watch Journal record your real-world sightings without looking at a screen.

Demo

The web app includes an integrated astronomical knowledge engine fallback if Ollama is not running locally, allowing instant use anywhere in the world.

Code

You can explore the full open-source codebase on GitHub:

GitHub logo ahanghosh77 / astrograss-ai

AstroGrass AI 🔭🌌

Dark-Sky Stargazing & Meteor Lookout — Powered by Open-Weight Models via Ollama
Built for Hacktoberfest 2026 Open-Source AI Challenge Week 1: Touch Grass

Hacktoberfest 2026 Inference Engine Models Zero Cloud Fees

🍃 Overview

Stargazing and experiencing the autumn cosmos requires literally touching grass in open meadows and dark-sky parks away from artificial lights.

However, modern stargazing apps flood users with intense screen blue-light that instantly destroys retinal rod rhodopsin (night-vision adaptation), forcing users to stare at phone screens rather than the actual stars above.

AstroGrass AI is designed with an anti-screen philosophy:

  1. Select your celestial targets (October Orionid meteors, Cassiopeia/Pegasus asterisms, Saturn/Jupiter, or deep-sky objects like the Pleiades and Andromeda Galaxy M31).
  2. Synthesize an observation plan in 15 seconds using local open-weight models (llama3.2:3b, gemma2:2b) running via Ollama.
  3. Toggle Astro Red Mode to eliminate blue-light and preserve retinal night adaptation.
  4. Start the Hands-Free Dark Adaptation Timer: Lie on the grass…
View on GitHub

The repository contains:

  • index.html: Complete semantic dark-sky observatory workstation featuring observation dispatcher, red night-vision mode, dark adaptation timer, SVG sky chart, and meteor counter.
  • style.css: Cosmic midnight indigo theme with custom photopic/scotopic deep-red night-vision filter and responsive glassmorphic cards.
  • app.js: Local Ollama integration (llama3.2:3b / gemma2:2b), fallback astronomical matrix, Web Audio API harmonic bell synthesizer, and persistent localStorage journal tracker.

How I Built It

AstroGrass AI is built with an offline-first, private edge architecture:

  • Local Open-Source Inference: Powered by open-weight models (Meta Llama 3.2 3B and Google Gemma 2 2B) running locally via Ollama (http://localhost:11434/api/generate) with structured JSON schema responses.
  • Deep Red Scotopic Spectrum Mode: An accessible CSS night-vision filter matching astronomical red flashlight standards to prevent night blindness.
  • Harmonic Audio Synthesizer: Implemented using the browser's native Web Audio API (OscillatorNode + GainNode), synthesizing 528Hz harmonic singing-bowl chimes to mark dark-adaptation milestones hands-free.
  • Dynamic Sky Chart SVG: Visualizes constellation vectors, stellar magnitudes, and meteor radiants.
  • Zero Cloud Dependence: 100% offline capable when stargazing in remote wilderness pastures with no cell signal.

Why Does Open Innovation Matter?

Open innovation and open-weight models via Ollama are vital for outdoor stargazing:

  1. Remote Wilderness Independence: The best stargazing locations (Bortle Class 1-3 dark-sky preserves, mountain pastures, national parks) inherently have zero cellular service. Local inference through Ollama ensures stargazers can query and generate observation plans completely offline.
  2. Zero Commercial Distractions: Closed-source commercial astronomy apps are bloated with in-app purchases, invasive ad tracking, and bright pop-ups that ruin night vision. Open-source AI keeps the experience pure, serene, and distraction-free.
  3. Hardware Freedom & Zero API Cost: Running lightweight open models like Llama 3.2 3B costs $0.00 and runs smoothly on standard laptops and edge devices.

My Agent Session

This project was built pair-programming with Google Antigravity. The agent structured the observation dispatcher, designed the deep-red night-vision theme and celestial SVG chart, engineered the prompt schema for Ollama, implemented the Web Audio API singing-bowl chime synthesizer, and prepared this submission post.

Prize Categories

  • Best Use of Ollama ($200 USD)

Top comments (0)