Submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass
Built in collaboration with @dkb73
PlantMyPlanet is an open-source, hyper-local garden planning application that starts with the ground under your feet instead of relying on a generic plant catalog.
When a user drops a pin on the map, the application fetches real-time soil composition and climate parameters for those exact coordinates, matches them against $2,568$ crops in the FAO EcoCrop dataset, and generates a stage-by-stage outdoor care checklist accompanied by an AI chat advisor tied directly to that specific planting instance.
Working alongside @dkb73, my primary responsibilities centered around the FastAPI backend architecture, weather API integration, frontend user interface, and the LLM chat advisor system:
FastAPI Backend Architecture: Designed and built the asynchronous FastAPI endpoints that coordinate requests, manage planting sessions, and handle data flow between the user interface, database, and external services.
Weather API Integration: Built the integration with Open-Meteo and NASA POWER to pull live climate metrics, elevation, solar radiation, and $16$-day local weather forecasts dynamically based on user-selected coordinates.
Frontend UI & Interactive Experience: Created the responsive user interface (HTML/JS) allowing users to drop a location pin, view suitabilities across crops, select planting scales, and track their garden progress stage-by-stage.
Contextual LLM Chat Advisor: Developed the interactive chat system powered by open-weight models (Google Gemma via OpenRouter). The prompt dynamically injects the specific planting context—crop type, days in the ground, current growth stage, pending tasks, and recent garden field notes—to deliver grounded, actionable advice.
Backend: Python, FastAPI, Asyncio
Frontend: HTML5, CSS3, JavaScript
Weather & Climate APIs: Open-Meteo API, NASA POWER API
AI Chat & Models: Google Gemma (google/gemma-4-31b-it:free, google/gemma-4-26b-a4b-it:free), NVIDIA Nemotron, Liquid LFM via OpenRouter
Database: PostgreSQL 16 (FAO EcoCrop catalog)
GitHub Repository: github.com/dkb73/plantTheplanet
Collaborator Submission Post: Check out Dhwanan's Post for full project architecture breakdown and additional screenshots.
docker compose up -d
python db/ingest_ecocrop.py
uvicorn app.main:app --host 0.0.0.0 --port 8000 --reload
By combining transparent, public weather and environmental data with swappable open-weight models like Google Gemma, PlantMyPlanet provides grounded advice without hiding its reasoning inside a closed model. If an API key is omitted, the app gracefully falls back to a deterministic agronomic timeline—ensuring that real-world gardeners always have actionable guidance.