Chapter 02 · Multi-platform product engineering
V3 shipped · Phases 0–5
A household finance platform shipped across web, Expo, and Android with unified Inbox, Goals, Reports, Health, Phase 5 insights, and honest backend maturity framing.
V3 UI flow
Assets below come from the current Android V3 emulator evidence set. Each flow pairs product behavior with the backend boundary it proves.
Flow 01
The redesigned home surface combines spend command cards, widget density, and Phase 5 insight charts including Sankey cashflow and predictive burndown.
InsightAggregatorService exposes cashflow, burndown, leakage, heatmap, correlation, and trade-off endpoints.
Flow 02
Gmail, WhatsApp, AI, and receipt drafts share one review queue with confidence sorting and draft inspector routing.
IntakeDraft source filtering and confirm/reject actions stay backend-owned before ledger mutation.
Flow 03
Savings goals, portfolio holdings, and challenges sit behind a segmented V3 control with gap analysis and fastest-runway logic.
Goal APIs protect family membership while progress and runway calculations stay centralized.
Flow 04
Monthly summaries, year-over-year comparisons, and export paths use the same V3 token system across web and native shells.
Report endpoints remain the first candidates for read-model caching as household traffic grows.
Flow 05
Power curve and muscle heatmap visualizations translate synced activity summaries into household fitness context.
Health summaries reuse backend activity contracts with client-side chart adapters under the frozen-backend constraint.
Flow 06
Income and expense lists, filters, and recurring controls share V3 chrome with state-preserving edit paths.
Finance mutations invalidate the shared financeInvalidationKeys set, including reports-summary.
Flow 07
Category budgets, coach insight banners, and progress bars keep spend visibility aligned across platforms.
Category writes carry nameId for Indonesian localization parity with web.
Flow 08
Household switching, invites, Gmail, Strava, Hevy, WhatsApp, and Google Sheets sync live under V3 settings surfaces.
OAuth tokens and provider state stay server-side behind protected backend services.
Multi-platform journey
V3 parity work wired inbox, goals, reports, health, and dashboard insights into each native shell while keeping finance mutations backend-owned.
Journey 01
Android and Expo home screens surface spend, savings, widgets, and insight charts from the same mobile API contracts.
Dashboard DTOs stay family-scoped without exposing raw persistence models.
Journey 02
Draft review, duplicate hints, and inspector routing compress intake triage into one mobile-native queue.
Draft fetch enforces familyId boundaries; errors surface real API codes instead of generic copy.
Journey 03
Savings, portfolio holdings, and challenges share one tabbed hub with parity logic ported from Android.
Goal CRUD endpoints enforce membership before any progress writeback.
Journey 04
Pushed routes under Settings/More carry Reports and Health where the five-slot tab bar has no room.
Read APIs stay stable while V3 navigation graphs add deep links without backend churn.
Journey 05
Confirm→applying→rollback flow with a two-key crash-safe commit marker lets households opt into V3 safely.
Version preference persists server-side; clients never trust local route state for finance scope.
Journey 06
Web, Expo, and Jetpack Compose V3 screens share inbox, goals, reports, health, and Phase 5 insight coverage.
Backend remains frozen while frontends consolidate IA around the same chapter map.
Tech stack
Backend maturity
Redis is not currently implemented in family-finance-backend.
Add Redis later for dashboard and reporting read models, using family-scoped keys, short TTLs, and write invalidation.
Kafka is not currently implemented in family-finance-backend.
Use an outbox-first async model before Kafka; reserve Kafka for replayable multi-consumer event streams.
The backend has Gmail Pub/Sub callbacks, worker-style backfill, goroutines, and an in-process monthly scheduler.
Move monthly summaries, push notifications, and provider retries to durable outbox-backed workers.
Request IDs, Cloud Armor planning, and tests exist, but metrics/tracing are not yet a full observability platform.
Add structured logs, metrics, traces, queue-depth alerts, provider latency tracking, and DB pool saturation alerts.
Backend proof points