Hidros

Water monitoring and anomaly alerting for hydroponic greenhouses.

What it does

Companies run greenhouses with hydroponic stations, each equipped with a Hermes IoT device and water sensors (flow in, flow out, pH). The system watches the water and warns people before a problem ruins a crop:

1. Collects sensor readings — Hermes devices publish measurements through the Midfield platform, which delivers them to this API via webhook (/midfield). Each reading batch becomes a "collect" tied to a station.
2. Detects anomalies — during each station's configured operating hours, water flow readings are sent to an ML prediction service that flags abnormal flow, typically a clog (insufficient flow) or a leak. Stations track consecutive warnings so one noisy reading doesn't trigger an alarm.
3. Alerts operators via WhatsApp — when an anomaly is confirmed, alert messages (critical/warning templates) go to the phone numbers registered on the greenhouse. Operators reply through WhatsApp buttons ("CONFIRMAR" / "NÃO É ENTUPIMENTO"); replies come back via the Meta webhook (/messages/webhook) and are recorded as human feedback on each measurement — which also feeds retraining of the prediction model.
4. Serves the dashboard — REST endpoints expose companies, greenhouses, stations, and paginated collect history for a frontend.

The product targets Brazilian operations: alert numbers follow the +55 format and timestamps use America/Sao_Paulo.

Technical overview

Spring Boot (Kotlin) API: collects sensor data, detects anomalies via a prediction service, and sends WhatsApp alerts.

Stack

• Kotlin / Spring Boot 3.5, Java 21
• PostgreSQL + jOOQ (generated Kotlin classes) + Liquibase migrations
• OpenFeign integrations: WhatsApp (Meta Graph API), prediction service, Midfield webhooks

Requirements

• JDK 21
• Docker (for local Postgres)
• GitHub credentials with access to the folhastech/midfield package registry

Setup

1. Copy .env-example to .env and fill in GITHUB_USER / GITHUB_TOKEN (needed to resolve the com.folhastech.midfield:models dependency).

2. Start the database:

   docker compose up -d

3. Run the app:

   ./gradlew bootRun

The API is served at http://localhost:8082/api/.

Configuration

Everything is env-var driven with local defaults (see src/main/resources/application.yml):

Variable	Default	Purpose
DATABASE_URL	jdbc:postgresql://localhost:5432/hidros_db	Postgres connection
DATABASE_USERNAME / DATABASE_PASSWORD	hidros / hidros	DB credentials
LIQUIBASE_ENABLED	false	Run migrations on startup
WHATSAPP_URL / WHATSAPP_TOKEN / WHATSAPP_PHONE_NUMBER_ID	—	WhatsApp alert integration
PREDICT_URL	http://localhost:8000	Anomaly prediction service

Database

• Migrations live in src/main/resources/db/changelog/ (Liquibase).

• After schema changes, regenerate jOOQ classes:

  ./gradlew jooqCodegen

  Generated code goes to src/main/kotlin/.../resources/jooq/generated.

Tests

./gradlew test              # unit tests
./gradlew integrationTest   # integration tests (Testcontainers, needs Docker)
./gradlew check             # both

Docker

docker build --build-arg GITHUB_USER=... --build-arg GITHUB_TOKEN=... -t hidros .

The container exposes port 8082. ./gradlew bootBuildImage builds and publishes an image using IMAGE_NAME, DOCKER_USERNAME, DOCKER_PASSWORD.

Project layout

src/main/kotlin/com/folhastech/hidros/
├── application/    # controllers, DTOs, config, scheduled jobs
├── domain/         # entities, services, gateways (Feign clients)
└── resources/      # jOOQ generated code, repositories