android-data-architecture

An offline-first Android data layer uses Room as the single source of truth, exposing data via Flow while Retrofit fetches remote updates. Repositories coordinate network and local data sources, keeping persistence and networking logic decoupled from domain contracts.

Standardize error handling in Android repositories using typed Result<T, E> wrappers and DataError classes. Map android-typed-errors from Retrofit and Room operations into domain-specific errors so presentation layers receive consistent, predictable failure states.

Keep domain models pure by defining separate DTOs for network parsing with Moshi and Room entities for persistence. Use mappers to convert DTOs to domain models, preventing any persistence or networking coupling from leaking into domain contracts.

Hilt can provision OkHttp interceptors and DataStore instances for token storage in Android. This setup enables authenticated Retrofit requests and secure token refresh handling while maintaining clean dependency injection boundaries across data sources.

Separate data sources from repositories when offline-first logic needs isolation from domain rules. Data sources handle Room queries and Retrofit calls directly, while repositories orchestrate caching, mapping, and error handling to keep domain interfaces pure Kotlin.

Repositories leak network details when domain layers depend directly on Retrofit responses or Room entities. Defining pure Kotlin domain interfaces and mapping DTOs to domain models prevents data, persistence, and networking abstractions from coupling.