Overview
GrassBuddy is an open‑source AI‑driven plant‑care companion that encourages you to step outside, tend to your garden, and learn about plant health. Built with local inference using the open‑weight Llama 2 model and a lightweight computer‑vision pipeline, it runs entirely on your own device – no cloud API keys required.
Why Open‑Source AI?
- Privacy & Offline First – All model inference happens locally, keeping your garden data private.
- Customizability – Fork the repo, swap the model, or add new plant species without vendor lock‑in.
- Community Learning – Share models and datasets with the wider open‑source community.
Core Features
- Plant Image Diagnosis – Snap a photo of a leaf; the app classifies diseases using a fine‑tuned vision model.
- Watering & Care Reminders – Based on species‑specific watering schedules, the app sends desktop notifications reminding you to water.
- Garden Map – Visualize all your plants on a simple map, encouraging you to explore your backyard.
Tech Stack
- Language: JavaScript (Node.js) with Electron for a cross‑platform desktop UI.
- AI Models: Llama 2 (7B) for natural‑language advice; a small ResNet‑based vision model for disease detection.
-
Local Inference:
ggmlbindings for efficient CPU inference. -
Packaging:
electron‑builderto produce binaries for Windows, macOS, and Linux.
How It Fits the “Touch Grass” Theme
GrassBuddy’s primary goal is to get you outside – the app only activates when you take a photo of a real plant. By providing instant AI feedback, it makes plant care fun and educational, turning screen time into outdoor activity.
Getting Started
# Clone the repository (URL omitted intentionally)
git clone <REPO_URL>
cd grassbuddy
npm install
npm start
Note: The repository URL is intentionally left blank as per your request. You can host the code on any platform you prefer.
Next Steps & Community Involvement
- Submit pull requests to improve the plant‑disease model.
- Add support for additional plant species.
- Contribute translations for the UI.
This project was developed for the **Hacktoberfest Open‑Source AI Challenge – Week 1* (Theme: “Touch Grass”).*
Top comments (1)
tr.ee/dev-to