Skip to content

Project Summary

This document provides an overview of the GreenThumb project.

Vision

GreenThumb is a controlled-environment plant production system that combines a Raspberry Pi controller, environmental sensors, automation, and a cloud backend to run greenhouses reliably and collect consistent cultivation data.

Current Status

Origin

GreenThumb started as a 12-month PIBITI undergraduate research project at Insper (2025–2026). The research phase closed with the final report in August 2026; development of the system continues.

What exists today

  • Edge node on a Raspberry Pi 5 with Docker Compose (PostgreSQL, device API, controller, local dashboard, Watchtower)
  • Sensor drivers: AHT10, BMP280, TSL2561, DS18B20, float switch, pH and TDS/EC probes, USB camera
  • Relay-switched actuators: grow light, exhaust fan, water and air pumps, peristaltic dosing pumps
  • Controller with threshold, schedule, interval and after-actuator rules, safety bounds and a heartbeat watchdog (safety mode)
  • Periodic photos and a live camera stream
  • Offline-first sync to a self-hosted cloud (PostgreSQL 17 + TimescaleDB; photos in Cloudflare R2)
  • Cloud API, authentication services and an admin dashboard (not public yet)
  • CI/CD: GitHub Actions → Docker Hub

Next

  • 🔄 Finishing the physical prototype
  • 🔄 Calibrating the dosing pumps and the pH and TDS probes
  • First cultivation run (cherry tomato)

Planned

  • Computer Vision: OpenCV for growth analysis
  • Machine Learning: Growth prediction models

Technology Stack

Component Technology
Controller Raspberry Pi 5
Language Python 3.11+
Web Framework FastAPI
Database PostgreSQL 17
ORM SQLModel
Containers Docker Compose
CI/CD GitHub Actions → Docker Hub
Cloud services FastAPI; Java (Spring Boot, Spring Cloud Gateway)
Dashboards React
Cloud data PostgreSQL 17 + TimescaleDB; Cloudflare R2 (photos)
Networking WireGuard

Sensors

  • AHT10: Temperature and humidity
  • BMP280: Atmospheric pressure and temperature
  • TSL2561: Light intensity
  • DS18B20: Water temperature
  • Float switch: Tank level (full or empty)
  • pH and TDS/EC probes: Nutrient solution pH and conductivity
  • USB Camera: Plant photos for computer vision

Actuators

  • Grow light, exhaust fan, water and air pumps: switched by relay
  • Peristaltic dosing pumps: nutrient and pH dosing with safety bounds

System Architecture

The system uses a microservices architecture with centralized device management:

Raspberry Pi 5
├── PostgreSQL (database)
├── microcontroller-api (API + hardware control)
├── controller (Sense-Think-Act loop client)
├── local-dashboard (React SPA)
└── watchtower (auto-updates)

All services run in Docker containers and share a common network.

A self-hosted cloud stack (API, gateway, authentication services, admin dashboard, PostgreSQL + TimescaleDB) receives the synced data. See Cloud Backend.

Repository Organization

See Repositories. Most repositories are private; this documentation and the organisation profile are public.

Data Collection

The system collects:

  • Sensor data, logged periodically
  • Photos, captured periodically for future computer-vision work

Data is stored on the node first and synced to the cloud when a connection is available.

Long-term Goals

  1. Many nodes: the cloud already registers and lists multiple devices; self-service onboarding is still to come.
  2. Improved Automation: Refine environmental control and monitoring

Contact