Isle — Cross-platform Habit Tracker

Context

The environment

Isle is an active beta built by a two-person team to provide one dependable habit model across browser, native desktop, and Android experiences. The product supports positive, negative, and composite habits while keeping schedule, completion, streak, and account behavior aligned between clients.

Problem

What needed to change

Deliver one cross-platform system without allowing recurrence rules, local dates, offline actions, authentication, or release behavior to drift between three client runtimes.

Constraints

Daily, weekly, and monthly RRULE recurrence must remain exact across local dates and time zones.
Web, Tauri, and Android share product behavior but require different token-storage and platform integration boundaries.
Offline completions must queue and synchronize without silently changing habit history.
A beta release process must test and package TypeScript, Java, Rust, Docker, desktop, and Android surfaces.
Product planning and implementation decisions are collaborative across the two-person team.

Contribution

My responsibility

Recurrence, completion, and timezone-aware streak implementation
Shared-package extraction and cross-client integration
Android delivery and native-platform integration
OAuth, session, and deployment reliability fixes
Docker health checks and delivery hardening
CI gates, semantic-release, and native artifact automation

Approach

How I structured the work

Keep recurrence and streak rules explicit

Model positive, negative, and composite habits around RRULE schedules, local-date completion, and timezone-aware streak calculations so every client receives the same domain answer.

Share behavior without hiding platform boundaries

Extract common TypeScript contracts and client behavior into a pnpm monorepo package while retaining platform-specific storage, navigation, and lifecycle integrations for web, Tauri, and Android.

Bound authentication by runtime

Use Google OAuth 2.0 PKCE with short-lived JWT access and refresh flows, including Tauri Stronghold storage for desktop refresh tokens and rate limiting at the service boundary.

Make delivery part of the product

Combine backend unit and integration checks with web, desktop, Rust, Docker, and Android CI gates, then use semantic-release for Docker publication and native release artifacts.

Evidence

A safe view of the system

Spring Boot habit & auth services

Web client

Tauri desktop

Android mobile

PostgreSQL & delivery

Shared behavior with platform-specific authentication and storage boundaries. This diagram is intentionally abstract and contains no operational identifiers.

Isle desktop dashboard showing schedule filters, completion progress, and habit cards with personal text permanently masked. — Desktop habit dashboard; personal greeting and habit text are irreversibly redacted.Open full-size

Isle analytics interface showing aggregate habit, streak, completion, and weekly consistency summaries without activity names. — Aggregate analytics with the named-activity area excluded from the publication crop.Open full-size

Profile layout with email, account name, and recent-activity values irreversibly redacted.Open full-size

Live application ↗The public beta opens at Isle's authentication boundary.Public repository ↗Source, implementation history, and release workflow for the beta product.

Outcome

What the work established

Established one domain model for recurrence, completion, and streak behavior across three client platforms.
Reduced duplicated client logic through a shared package while preserving platform-specific security boundaries.
Connected test, build, container, desktop, and Android checks to an automated release workflow.
Built offline, authentication, timezone, and delivery concerns into the beta foundation rather than treating them as later add-ons.