🌱 GrowMate: Your Local AI Gardening Companion with Gemma 3 4B | Hacktoberfest 2026
What if AI could help us spend less time staring at screens and more time taking care of the world around us?
For the Hacktoberfest 2026 — “Touch Grass” challenge, I'm building GrowMate, an AI-powered gardening companion designed to make gardening more approachable through local AI.
🌿 GitHub Repository: NiranjanRSoorej06/growmate
🤖 AI Model: Gemma 3 4B via Ollama
🌿 The Idea
Starting a garden can feel overwhelming, especially for beginners. Which plants should you grow? How often should you water them? What should you do when a plant looks unhealthy?
GrowMate aims to make these questions easier to tackle through an accessible web interface and AI-assisted gardening guidance.
The goal is to use technology to help people learn about plants and turn that knowledge into real-world action.
🤖 Why Gemma 3 4B?
GrowMate uses Gemma 3 4B through Ollama for local AI experimentation.
Running a model locally offers several advantages:
- Privacy: Requests can be processed locally rather than automatically sent to a third-party AI API.
- Control: The model can be integrated into a custom backend and application workflow.
- Accessibility: Developers can experiment with open models without depending entirely on paid proprietary APIs.
- Learning: Local inference provides hands-on experience with AI integration and deployment.
Local inference also has trade-offs, including hardware requirements and response latency.
🛠️ Tech Stack
| Component | Technology |
|---|---|
| Frontend | React, Vite, Tailwind CSS |
| Backend | Python, FastAPI |
| AI model | Gemma 3 4B |
| Model runtime | Ollama |
| Database | SQLite |
| Deployment | Render configuration in progress |
The frontend communicates with the FastAPI backend, which handles application logic and connects to Ollama for AI inference.
🔄 How It Works
- A user submits a gardening question through the frontend.
- The frontend sends the request to FastAPI.
- The backend communicates with the configured Ollama model.
- The generated response is returned to the frontend.
- The user can apply the guidance to their gardening activities.
The broader vision includes gardening guidance, planting plans, plant-care assistance, and a gardening journal. These capabilities will be developed and verified as the project progresses.
💻 Running Locally
Ensure Ollama is running and the model is available:
ollama pull gemma3:4b
ollama list
Start the backend from the backend directory with your Python virtual environment activated:
uvicorn app.main:app --reload --host 127.0.0.1 --port 8000
In a separate terminal, start the frontend:
cd frontend
npm install
npm run dev
Open the local URL printed by Vite, usually http://localhost:5173.
For local AI inference, configure the backend to use:
OLLAMA_URL=http://localhost:11434
Keep real environment files and credentials out of version control.
☁️ Deployment Challenges
I'm preparing GrowMate for deployment on Render.
One important lesson is that deploying an application does not automatically deploy its AI model. Inside a cloud container, localhost:11434 refers to the container itself, not Ollama running on my laptop.
Hosted AI requests require a separately reachable Ollama server. This is one of the infrastructure challenges I'm working through as I prepare the project for production.
🌍 Why “Touch Grass”?
AI is often used to help us do more work on our computers. But what if it could also encourage us to spend more time away from them?
A gardening assistant should ultimately help someone take practical action: water a plant, learn about sunlight requirements, prepare soil, or develop a habit of caring for something living.
The goal is to make AI a bridge between digital knowledge and real-world activity.
🚧 What's Next?
- Improve gardening guidance and the user experience.
- Continue testing local Gemma and Ollama integration.
- Verify backend health checks and database persistence.
- Complete the Render deployment configuration.
- Configure reachable inference infrastructure for hosted AI.
- Test the complete workflow from asking a question to applying the advice.
🤝 Explore the Project
🌱 GitHub: https://github.com/NiranjanRSoorej06/growmate
I'm interested in feedback, suggestions, and contributions that could make GrowMate more useful for beginners and gardening enthusiasts.
🧑💻 Build Process
I also documented my development workflow through a DevRelay build session.
🔗 Explore the GrowMate build session on Dev.to
This provides additional context on the implementation process and the technical decisions behind the project.
💚 Final Thoughts
GrowMate is my experiment in combining open AI models with an everyday activity that gets us away from our screens.
The interesting part isn't just making a model answer questions. It's building an application that helps turn those answers into practical action.
If you were building an AI project to help people spend more time outdoors, what would you create? I'd love to hear your ideas in the comments!
Project: GrowMate
Model: Gemma 3 4B via Ollama
Challenge: Hacktoberfest 2026 — Touch Grass
📸 Project Screenshots
Top comments (0)