This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass
What I Built
The Local Micro-Climate Garden Planner is a full-stack, local-first web application designed to encourage people to spend less time staring at digital screens and more time outside interacting with nature ("Touch Grass").
🌿 The Problem
Urban micro-climates vary drastically within the exact same city. A 4th-floor south-facing balcony experiences intense sun and dry winds, while a ground-floor terrace remains shaded. Generic online advice leads to withered seedlings and frustrated beginner gardeners.
🎯 The Solution
Our application acts as a personal, intelligent gardening mentor:
- Micro-Climate Profiling: Captures garden location, growing space (balcony, terrace, windowsill, plot), daily sunlight, soil mix, water routine, and temperature.
- Transparent 6-Factor Rule Engine: Scores plant suitability from 0 to 100% using deterministic biological rules—ensuring plant safety without black-box AI hallucinations.
- Live & Offline Climate Dashboard: Integrates real-time weather from Open-Meteo API with a zero-network manual fallback.
-
Local Open-Weight AI Assistant: Powered by Ollama (
qwen2.5/llama3.2) with 100% local privacy (no paid cloud API keys required) and deterministic fallbacks. - 7-Day Outdoor Action Plan: Climate-aware daily tasks (soil finger test, pot drainage check, seed sowing, leaf inspection) specifically designed to get users outdoors.
- My Garden Progress Tracker: SQLite-backed tracker to monitor growth stages (seeded, sprouted, growing, harvestable) and record observations.
Demo
- GitHub Repository: https://github.com/Hyma27/Garden-Planner
-
Local Application URL:
http://localhost:5173 -
FastAPI Backend & Swagger Docs:
http://localhost:8000/docs
📸 Application Highlights
- Home Hero: Nature-inspired modern SaaS dashboard inviting users to plan their garden.
- Micro-Climate Setup: Accessible forms with tooltips explaining direct vs partial sunlight and soil aeration.
- Climate Dashboard: Recharts crop suitability visualization and real-time Open-Meteo weather warnings for extreme heat or frost.
- Plant Recommendations: Suitability score badges with full point breakdown rationale ("Why this suits your garden").
- AI Gardening Assistant: Context-grounded chat interface that highlights a "Suggested Outdoor Action" with every response.
Code
The project is built with a modular, maintainable full-stack architecture:
- Frontend: React 19, Vite, Tailwind CSS v4, Lucide React Icons, Recharts
- Backend: Python 3.12, FastAPI, Pydantic v2, Uvicorn, HTTPX
- Database: SQLite (Zero-config local persistence)
-
AI Engine: Ollama running open-weight models (
qwen2.5:latestorllama3.2:latest) - Live Weather: Open-Meteo REST API (No API key needed)
- Test Suite: Pytest (11/11 tests passing for recommendation scoring, SQLite CRUD, and AI fallback)
🐙 Repository Link
👉 GitHub Repository: Hyma27/Garden-Planner
Open Innovation & Local-First AI
We believe open innovation makes technology more private, transparent, and accessible:
- 100% Local Inference: Runs locally via Ollama. Personal garden details and chat messages never leave your machine.
- Zero Paid API Keys: Requires no cloud AI subscriptions or API keys.
- Deterministic Rule Safety: Recommendations are calculated via verified biological constraints rather than LLM text generation.
- Offline Fallback Engine: If Ollama is offline or hardware is limited, the app seamlessly switches to a rule-based fallback response engine.
My Agent Session
This full-stack application was pair-programmed and built using Google Antigravity AI Agent. The agent performed:
- End-to-end project scaffolding (React + FastAPI + SQLite).
- Designing the rule-based plant recommendation engine.
- Open-Meteo API integration with fallback handling.
- Ollama local AI client integration.
- Unit and integration testing with Pytest (11 passing tests).
Prize Categories
- Primary Category: Hacktoberfest Open-Source AI Challenge — Week 1: Touch Grass
Grow smarter. Step outside. Touch grass. 🌱
Top comments (0)