EcoDose
Soil readings in, a recommended dose of Rhizobium biofertiliser out, with a sensor kit for taking the readings.
- Kind
- Sensor firmware, API and web app
- Year
- 2025
- Area
- Machines
- Built with
- ESP32
- Flask
- scikit-learn
- React
- Vite
- Gemini
The question
How much biofertiliser should go on this soil?
EcoDose turns five soil numbers into a dose and then explains the soil in words a farmer can use.
Why it exists
The right dose of a Rhizobium biofertiliser depends on the soil. EcoDose takes soil readings and recommends a dose, with plain-language feedback on the soil from Gemini.
How it works
From soil to a dose
- 1
Readings
The sensor kit gives moisture and pH, and the other three values come from a soil test.
- 2
Form
The values are entered in the web app and checked against sensible ranges, such as pH 3 to 12.
- 3
Dose
The chosen model, Random Forest or Linear Regression, predicts a dose in g/m².
- 4
Feedback
Gemini writes a short soil-health note, and a chatbot can answer follow-up questions.
Decisions
The models set the dose, the language model explains
The dose comes from the trained regression models. Gemini’s soil feedback is prompted not to include a dose, so a figure never comes from free text.
Two models to compare
The form offers Random Forest, the default, and Linear Regression, so a user can see how much the choice matters.
Framed as educational
The terms say the output is educational and not professional agricultural advice.
Not built yet
- Sending the sensor readings to the web app automatically (they are entered in the form for now)
Built with
Frameworks and libraries
- Flask
- React
- Vite
AI and models
- scikit-learn
- Gemini
Systems and hardware
- ESP32