This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass.
What I Built
🌱 JalRakshak Garden AI is a local-first, AI-powered gardening assistant designed to connect digital intelligence with real-world plant care.
The idea is simple: instead of spending all our time looking at screens, let's use technology to help us spend more time outdoors, care for plants, understand our gardens, and make better watering decisions.
JalRakshak combines a garden management web application with an IoT hardware prototype. My goal is to create a system that can monitor environmental conditions, collect soil moisture readings, maintain a plant-care journal, and eventually automate watering based on real sensor data.
The project is being built for home gardeners, beginners, and anyone interested in combining open-source AI, IoT, and sustainable gardening.
Hardware I'm using
My current prototype includes:
LOLIN NodeMCU V3 ESP8266: Wi-Fi-enabled microcontroller for collecting sensor data.
DHT11: Temperature and humidity monitoring.
Soil moisture sensor module with LM393 comparator: Provides a moisture threshold output and, depending on the module, an analog reading.
A second fork-shaped soil moisture probe: Intended to provide an additional soil measurement once its interface and output are verified.
Mini DC water pump: Intended for controlled plant watering.
Single-channel relay module: Intended to switch the pump using an appropriately rated external power supply.
The DHT11 has already produced real readings of approximately 30–31 °C and 64–66% relative humidity during testing.
The next hardware milestone is to verify both soil sensor outputs, calibrate their readings, and integrate reliable telemetry into the application. The pump will remain disabled until its driver circuit, power supply, and watering safeguards have been tested.
Demo
Project repository: JalRakshak Garden AI on GitHub
The repository contains the project implementation I'm developing.
The hardware prototype is currently under active development. I have verified DHT11 readings over serial, but the ESP8266-to-backend connection still needs to be validated on a shared network. I don't yet have a verified public deployment or a complete end-to-end hardware demo to share.
My next demo will show live sensor readings reaching the application and a controlled watering test using real soil moisture data.
Code
Explore the source code and follow the project's progress:
GitHub: https://github.com/gunmasterg9/JalRakshak-Garden-AI
The intended data flow is:
DHT11 ───────────────┐
│
Soil Sensor 1 ───────┤
├──> NodeMCU ESP8266
Soil Sensor 2 ───────┤ │
│ │ Wi-Fi
└──────────┘
│
▼
FastAPI Backend
│
▼
SQLite Database
│
▼
React Dashboard
ESP8266 GPIO
│
▼
Relay / Driver
│
▼
Mini Water Pump
This diagram represents the intended architecture. The complete sensor-to-dashboard and automatic-watering paths are still being integrated and tested.
How I Built It
I wanted JalRakshak to combine practical hardware with an AI system that can run locally instead of depending entirely on paid cloud APIs.
Software stack
- React: Responsive web interface for garden management.
- FastAPI: Python backend and API endpoints.
- SQLite: Local data storage.
- Ollama: Local inference runtime for experimenting with open-weight language models.
- ESP8266 Arduino framework: Firmware for reading sensors and communicating over Wi-Fi.
- DHT11 and soil moisture sensors: Environmental and soil data collection.
I've been experimenting with locally available models, including Llama 3.2, Qwen3 8B, and Gemma 4 12B. The intention is to use a suitable local model for gardening assistance, plant-care explanations, and recommendations without sending private garden notes to an external AI service.
Development milestones
1. Building the application
I started with the garden application's frontend and backend, with plant management, journal functionality, recommendations, and IoT API integration as core parts of the project.
2. Connecting the ESP8266
I configured the NodeMCU to connect to Wi-Fi and read temperature and humidity from the DHT11. The sensor readings are working over serial.
3. Integrating real telemetry
The backend provides an IoT telemetry endpoint. I am working on getting real ESP8266 readings accepted by the backend and displayed by the application. A network connection problem currently prevents me from claiming successful end-to-end telemetry.
4. Adding two soil moisture sensors
I want to compare readings from two soil sensors instead of relying on a single measurement. Before using those values for decisions, I need to identify their output interfaces, check voltage compatibility, and calibrate their readings in dry and wet soil.
5. Preparing safe automatic watering
The mini pump and relay are part of the hardware prototype. Before enabling automatic watering, I plan to add safeguards such as configurable moisture thresholds, a maximum pump runtime, cooldown periods, and a manual emergency stop.
The pump will use a suitable external power supply and a correctly rated switching circuit. It will never be powered directly from an ESP8266 GPIO pin.
Why Does Open Innovation Matter?
Gardening technology should be accessible, understandable, and adaptable.
Open-source software and open-weight AI make it possible to experiment with local inference, inspect implementation details, and adapt the system to different plants, gardens, and hardware configurations.
For a project like JalRakshak, this matters in several ways:
- Privacy: Garden journals and personal notes can remain on the local machine when local inference is used.
- Affordability: Hobbyists can experiment with inexpensive microcontrollers and sensors.
- Transparency: Developers can inspect the code and understand how the system processes sensor readings.
- Community collaboration: Others can contribute support for new sensors, plant-care workflows, models, and hardware.
- Learning by building: The project connects AI development with electronics, programming, and hands-on gardening.
A closed AI API might provide useful recommendations, but an open and locally runnable architecture gives developers greater control over where data is processed and how the system evolves.
My goal is not simply to build another chatbot. I want to build something that encourages people to step away from their screens, look at their plants, and use technology to support real-world activity.
My Agent Session
I haven't included a DevRelay agent-session recording because I don't yet have a verified session link to share.
If I capture a relevant session during development, I'll add it here using the challenge's required embedding format.
What's Next?
My immediate priorities are:
- Verify and calibrate both soil moisture sensors.
- Establish reliable Wi-Fi communication between the ESP8266 and FastAPI backend.
- Store real sensor readings and distinguish them from simulated data.
- Display live temperature, humidity, and soil moisture on the dashboard.
- Validate the relay and pump with a suitable external supply and safe switching circuit.
- Implement bounded watering rules and an emergency stop.
- Record a short demonstration showing the complete system working with a real plant.
JalRakshak Garden AI is a work in progress, but each step brings the project closer to connecting open-source AI with practical, outdoor plant care.
Let's build technology that helps us touch grass — literally. 🌱
Top comments (0)