DEV Community

Sandip Sahoo
Sandip Sahoo

Posted on

๐ŸŒฟ FrostBite: Offline AI Garden Scout

๐ŸŒฟ FrostBite: Offline Open-Weight Garden Scout

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


๐Ÿ“– What I Built

FrostBite is a lightweight, zero-cloud CLI tool that helps gardeners step away from the screen and into the soil.

  • Runs 100% offline on your machine โ€” no internet required.
  • Accepts your USDA hardiness zone or region and looks up historical frost benchmark dates locally.
  • Prompts a locally running open-weight SLM (smollm2:1.7b) to output a concise 5-step outdoor gardening checklist.
  • Renders the checklist in an ASCII index-card format directly in your terminal and saves it as garden_todo.txt.

This way, you can fold the slip, pocket a pencil, close your laptop, and head outdoors. Itโ€™s designed for gardeners, hobbyists, and anyone who wants to spend less time staring at screens and more time tending to nature.


๐ŸŽฅ Demo

๐Ÿ‘‰ Live Demo Repo


๐Ÿ’ป Code

๐Ÿ‘‰ GitHub Repository


๐Ÿ› ๏ธ How I Built It

  • Language: Python 3.10+
  • Inference Engine: Ollama (local loopback at http://localhost:11434)
  • Model: smollm2:1.7b โ€” ultra-compact 1.7B parameter open-weight model designed for fast, local CPU/RAM inference.
  • Dependencies: requests for local REST communication.
  • Offline Frost Data: USDA hardiness zones & frost benchmark lookup table.
  • CLI Workflow: Python scripts (garden_cli.py, frost_data.py) generate the ASCII checklist and save it locally.

๐ŸŒฑ Why Does Open Innovation Matter?

Closed APIs often require subscriptions, internet connectivity, or send your personal data to external servers. With open-weight models and local inference:

  • Privacy: All data stays on your device.
  • Accessibility: Free to run, no API costs.
  • Customization: Fine-tune for your regionโ€™s frost dates and local crops.
  • Resilience: Works offline in the field, even with no signal.

Open innovation made it possible to build a tool that is practical, portable, and empowering for everyday gardeners.


๐Ÿง‘โ€๐Ÿ’ป My Agent Session

I used a local agent workflow to dispatch prompts to smollm2:1.7b via Ollama. The session generates structured, actionable outdoor tasks in under 3 seconds โ€” keeping the screen time minimal and the outdoor time maximal.


๐Ÿ† Prize Categories

  • Best Use of Open-Source AI
  • Touch Grass Challenge
  • Sustainability & Environment

โœจ Closing Note

FrostBite embodies Hacktoberfestโ€™s Touch Grass theme: the screen is the shortest part of the experience. You grab your slip, then head outside to garden.

Built with โค๏ธ by Sandip.

Top comments (0)