For the complete documentation index, see llms.txt. This page is also available as Markdown.

🔧Technology Stack

The languages, frameworks and data stores SPICE is built on, by layer.

SPICE is built entirely on open-source technology. Versions below are major versions taken from the project's build files and will move forward over time; check each repository's build file for exact versions. For how the pieces fit together, see the Architecture Overview and System Architecture (C4).

Mobile app

Concern
Choice
Why

Language

Kotlin 2.x

Standard language for native Android.

Platform

Android, minimum API level 23 (Android 6.0)

Wide device coverage for field devices.

UI

Android Views with view binding, plus Jetpack Compose

Mature screens plus incremental adoption of Compose.

Architecture

MVVM with repositories

Separates UI, data access and sync logic.

Dependency injection

Hilt 2.x

Compile-time DI that integrates with Android lifecycles.

Local database

Room 2.x on an encrypted SQLite (SQLCipher 4.x)

Offline-first storage with encryption at rest.

Networking

Retrofit 2 with OkHttp 4

De facto REST client for Android.

Background work

WorkManager 2.x

Reliable, constraint-aware background synchronisation.

Image loading

Glide 4

Efficient image caching.

Build

Gradle (Kotlin DSL) with the Android Gradle Plugin 9

Standard Android toolchain.

Web front ends

Concern
Admin web portal
Reporting web app

Framework

React 19

React 17

Language

TypeScript 5

TypeScript 4

State management

Redux Toolkit 2

Redux 4

Routing

React Router 7

React Router 5

Styling

Bootstrap 5, SCSS

Bootstrap 5, SCSS

Build tooling

Vite 8 (Node 22 or later)

Create React App (webpack-based)

HTTP client

Axios

Axios

Maps

Leaflet 1 with React-Leaflet

not used

Tests

Jest 30

Jest

React with TypeScript was chosen for typed, component-based single-page applications; Redux centralises client state.

Backend services

Concern
Choice
Why

Language and runtime

Java 25

Long-term support Java with modern language features and a mature ecosystem.

Framework

Spring Boot 3.x (3.5 line)

Convention-driven REST services, security and data access.

Service-to-service calls

Spring Cloud 2025.x with OpenFeign, circuit breakers

Declarative REST clients with resilience.

Persistence

Spring Data JPA with Hibernate

Standard object-relational mapping.

Schema migrations

Flyway (in the user, facility registry and FHIR services)

Versioned, repeatable database changes.

Security

Spring Security with signed tokens (JWT)

Stateless, role-based API protection.

API documentation

OpenAPI via springdoc (Swagger UI)

Interactive API docs generated from code.

Notifications

Thymeleaf email templates; pluggable SMS and email providers

Provider-neutral message delivery.

Scheduling

ShedLock-guarded scheduled jobs

Ensures a job runs on one instance only when scaled out.

Testing and quality

JUnit 5, Mockito, JaCoCo

Unit testing and coverage.

Build

Maven (multi-module)

Standard Java build and dependency management.

Data stores

Store
Choice
Use

Relational databases

PostgreSQL (managed or self-hosted)

Platform data, FHIR store and reporting database. Reference container setup uses PostgreSQL 18.

Cache

Redis

Caching and shared state.

Object storage

Object storage (self-hosted or managed)

Files such as signatures, consent forms and exports.

Messaging

Concern
Choice
Why

Message queue

RabbitMQ, via Spring Cloud Stream

Asynchronous hand-off of offline-sync processing.

FHIR

Concern
Choice
Why

FHIR version

R4

Widely implemented release for interoperability.

FHIR server

HAPI FHIR JPA server 8.x (starter project), PostgreSQL-backed

Mature open-source FHIR server. SPICE adds token validation, optional partitioning and migrations.

FHIR mapper

Spring Boot service using the HAPI FHIR R4 structures and client libraries

Keeps FHIR mapping out of the clinical workflow code.

Clinical rules

Clinical Quality Language (CQL) service using HAPI FHIR client libraries

Standards-based clinical decision support.

See FHIR and Standards.

Reporting, ETL and analytics

Concern
Choice
Why

Reporting ETL

Python 3.11 or later with psycopg2, scheduled batch job

Flattens FHIR and administrative data into reporting tables.

Reporting database

PostgreSQL

Fixed, denormalised reporting schemas.

Report API

Java service (Spring Boot)

Serves the reporting web app.

Analytics stack (optional)

Apache Airflow, Apache Spark, Airbyte, Hive Metastore, Trino, Apache Superset, deployed with Helm charts

Open-source orchestration, processing, query and dashboard components for large-scale analytics.

Build and packaging

Concern
Choice

Containers

Each backend service, the FHIR server, the web apps and the ETL job build to container images (Docker). The Java services run on a Java 25 runtime image; the FHIR server uses a minimal distroless image.

Orchestration

Docker Compose for simple installs; Kubernetes with Helm charts for larger ones.

Reverse proxy

Nginx in the reference compose setup.

Android packaging

Gradle build variants (product flavours) configure the backend endpoints per deployment.

Last updated