agentleFS
PL-Coding-GmbHPL-Coding-GmbH/Skills/skills/kmp-data-layer/SKILL.md

kmp-data-layer

Skill18 starsChanged 4 months agoMITView on GitHub

Data layer patterns for Kotlin Multiplatform (Android + iOS) - data sources, repositories, DTOs, mappers, Room entities, Ktor HttpClient, safe call helpers, token storage, and offline-first. Use this skill whenever writing or reviewing a data source or repository, creating DTOs or Room entities, writing mappers, setting up the Ktor HttpClient, handling network errors, or implementing token refresh. Trigger on phrases like "create a repository", "create a data source", "add a DAO", "Ktor client", "write a mapper", "DTO", "Room entity", "network call", "token storage", or "offline-first".

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 1 is distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    All data layer code lives in commonMain unless it needs a platform API.

    line 9
  2. Distinctive:

    Always separate: DTOs (data layer) ↔ Domain Models (domain layer).

    line 50
  3. Very specific:

    Use names like RoomNoteDataSource, KtorNoteDataSource, OfflineFirstNoteRepository.

    line 96
  4. Very specific:

    Use safeCall / responseToResult helpers and typed extension functions (HttpClient.get, HttpClient.post, HttpClient.delete) to keep data …

    line 144
  5. Very specific:

    Create exactly one DataStore instance per file (Koin single).

    line 188
  6. Very specific:

    Prefer @Database(autoMigrations = [AutoMigration(from = 1, to = 2)]).

    line 237
  7. Very specific:

    Follow Room as single source of truth: fetch from network → persist to Room → expose DB Flow to the ViewModel.

    line 243

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

KMP Data Layer

All data layer code lives in commonMain unless it needs a platform API. The only platform-specific pieces are the Ktor engine, the Room database builder, and the DataStore file path — each is a few lines in androidMain / iosMain.

Error Handling

This skill uses Result<T, E>, DataError, and the extension helpers defined in the kmp-error-handling skill. Refer to that skill for the full Result wrapper, DataError sealed interface, and map/onSuccess/onFailure/asEmptyResult extensions.


Data Source vs Repository

  • Data source — accesses a single data source (local DB, remote API, file system). Most classes in the data layer are data sources.
  • Repository — combines multiple data sources (e.g., a remote API + a local DB for offline-first). Only use the term "repository" when the class genuinely coordinates multiple sources.
// Single source → data source
interface NoteLocalDataSource {
    suspend fun getNotes(): Result<List<Note>, DataError.Local>
    suspend fun insertNote(note: Note): EmptyResult<DataError.Local>
}

interface NoteRemoteDataSource {
    suspend fun fetchNotes(): Result<List<Note>, DataError.Network>
}

// Multiple sources → repository
interface NoteRepository {
    suspend fun getNotes(): Result<List<Note>, DataError>
    suspend fun sync(): EmptyResult<DataError>
}

Domain Layer Contracts

  • Pure Kotlin in commonMain — no Android/framework imports and no java.* imports (use kotlin.time / kotlinx-datetime instead of java.time, kotlin.uuid.Uuid instead of java.util.UUID).
  • Contains: domain models, data source/repository interfaces, error types.
  • Every data source or repository used by a ViewModel must have an interface in domain — enforces that presentation never depends on data, and enables testing.

DTOs and Domain Models

  • Always separate: DTOs (data layer) ↔ Domain Models (domain layer).
  • Domain models never go directly into Room entities or Ktor request/response bodies.
  • Mappers are simple extension functions living in the data layer alongside the DTO:
fun NoteDto.toNote(): Note = Note(id = id, title = title, ...)
fun Note.toNoteDto(): NoteDto = NoteDto(id = id, title = title, ...)
fun NoteEntity.toNote(): Note = ...
fun Note.toNoteEntity(): NoteEntity = ...

Implementations

Name implementations for what makes them unique — never suffix with Impl.

Data source (single source)

class RoomNoteDataSource(private val dao: NoteDao) : NoteLocalDataSource {
    override suspend fun getNotes(): Result<List<Note>, DataError.Local> {
        return try {
            Result.Success(dao.getAllNotes().map { it.toNote() })
        } catch (e: Exception) {
            Result.Error(DataError.Local.UNKNOWN)
        }
    }
}

Repository (multiple sources)

class OfflineFirstNoteRepository(
    private val localDataSource: NoteLocalDataSource,
    private val remoteDataSource: NoteRemoteDataSource
) : NoteRepository {
    override suspend fun getNotes(): Result<List<Note>, DataError> {
        return remoteDataSource.fetchNotes()
            .onSuccess { notes -> localDataSource.insertAll(notes) }
            .onFailure { localDataSource.getNotes() }
    }
}

Use names like RoomNoteDataSource, KtorNoteDataSource, OfflineFirstNoteRepository. The name should tell you what the class wraps or how it behaves.


Ktor — HttpClient Factory (core:data)

Configure the client once. Accept the engine externally so tests can swap in a mock engine:

object HttpClientFactory {
    fun create(engine: HttpClientEngine): HttpClient = HttpClient(engine) {
        install(ContentNegotiation) { json() }
        install(Auth) {
            bearer {
                loadTokens { /* load from DataStore */ }
                refreshTokens { /* call refresh endpoint, save new tokens */ }
            }
        }
        install(Logging) { logger = Logger.DEFAULT; level = LogLevel.ALL }
        defaultRequest { contentType(ContentType.Application.Json) }
    }
}

Inject HttpClient via Koin. The engine is the only platform-specific part — it is provided by the platform Koin module (see kmp-di-koin):

Source setArtifactsEngine
commonMainio.ktor:ktor-client-core, ktor-client-content-negotiation, ktor-serialization-kotlinx-json, ktor-client-auth, ktor-client-logging—
androidMainio.ktor:ktor-client-okhttpOkHttp.create()
iosMainio.ktor:ktor-client-darwinDarwin.create()
// androidMain
actual val coreDataPlatformModule = module {
    single<HttpClientEngine> { OkHttp.create() }
}

// iosMain
actual val coreDataPlatformModule = module {
    single<HttpClientEngine> { Darwin.create() }
}

Ktor — Safe Call Helpers (core:data)

Use safeCall / responseToResult helpers and typed extension functions (HttpClient.get, HttpClient.post, HttpClient.delete) to keep data source call sites clean and uniform. See the kmp-error-handling skill for the full implementation of these helpers.

suspend fun getNotes(): Result<List<NoteDto>, DataError.Network> {
    return httpClient.get(route = "/notes")
}

Token Storage

Store tokens in DataStore (in core:data or a dedicated :core:auth / :feature:auth:data module). The Ktor Auth plugin reads/writes tokens and handles 401 refresh automatically.

DataStore is multiplatform: androidx.datastore:datastore-preferences-core goes into commonMain. Only the file path differs per platform:

// commonMain
internal const val DATA_STORE_FILE_NAME = "prefs.preferences_pb"

fun createDataStore(producePath: () -> String): DataStore<Preferences> =
    PreferenceDataStoreFactory.createWithPath(
        produceFile = { producePath().toPath() }   // okio.Path.Companion.toPath
    )

// androidMain
fun createDataStore(context: Context): DataStore<Preferences> = createDataStore {
    context.filesDir.resolve(DATA_STORE_FILE_NAME).absolutePath
}

// iosMain
@OptIn(ExperimentalForeignApi::class)
fun createDataStore(): DataStore<Preferences> = createDataStore {
    val directory = NSFileManager.defaultManager.URLForDirectory(
        directory = NSDocumentDirectory,
        inDomain = NSUserDomainMask,
        appropriateForURL = null,
        create = false,
        error = null
    )
    requireNotNull(directory).path + "/$DATA_STORE_FILE_NAME"
}

Create exactly one DataStore instance per file (Koin single).


Room (KMP setup)

Room is multiplatform. Entities, DAOs, and the @Database class live in commonMain of a database module (:core:database or :feature:<name>:database) that applies the room convention plugin — it adds room-runtime, sqlite-bundled, the Room Gradle plugin, and the per-target KSP configurations (kspAndroid, kspIosArm64, kspIosSimulatorArm64, kspIosX64).

// commonMain
@Database(entities = [NoteEntity::class], version = 1)
@ConstructedBy(AppDatabaseConstructor::class)
abstract class AppDatabase : RoomDatabase() {
    abstract fun noteDao(): NoteDao
}

// Room's KSP processor generates the actual implementations
@Suppress("NO_ACTUAL_FOR_EXPECT")
expect object AppDatabaseConstructor : RoomDatabaseConstructor<AppDatabase> {
    override fun initialize(): AppDatabase
}

fun createDatabase(builder: RoomDatabase.Builder<AppDatabase>): AppDatabase = builder
    .setDriver(BundledSQLiteDriver())
    .setQueryCoroutineContext(Dispatchers.IO)   // import kotlinx.coroutines.IO
    .build()

// androidMain
fun databaseBuilder(context: Context): RoomDatabase.Builder<AppDatabase> {
    val dbFile = context.applicationContext.getDatabasePath("app.db")
    return Room.databaseBuilder<AppDatabase>(context.applicationContext, dbFile.absolutePath)
}

// iosMain
fun databaseBuilder(): RoomDatabase.Builder<AppDatabase> {
    val dbFile = documentDirectory() + "/app.db"   // NSDocumentDirectory, same lookup as DataStore
    return Room.databaseBuilder<AppDatabase>(name = dbFile)
}

KMP-specific rules:

  • Every DAO function must be suspend or return Flow — blocking DAO functions don't compile for iOS.
  • No LiveData, Cursor, SupportSQLiteDatabase, or SupportSQLiteQuery. Use the SQLiteConnection / SQLiteDriver APIs.
  • No allowMainThreadQueries() — it does not exist for the bundled driver.

Room Migrations

Prefer @Database(autoMigrations = [AutoMigration(from = 1, to = 2)]). Use manual Migration objects when the schema change is too complex for auto-migration. In KMP, override migrate(connection: SQLiteConnection) (not the SupportSQLiteDatabase overload) and run statements with connection.execSQL(...).


Offline-First (when applicable)

Follow Room as single source of truth: fetch from network → persist to Room → expose DB Flow to the ViewModel. The ViewModel never observes network responses directly.

This pattern is optional — apply it when the project requires offline support.


Naming Conventions

ThingConventionExample
Data source interface<Entity><Local/Remote>DataSourceNoteLocalDataSource, NoteRemoteDataSource
Data source impldescribe what makes it uniqueRoomNoteDataSource, KtorNoteDataSource
Repository interface<Entity>Repository (multi-source only)NoteRepository
Repository impldescribe what makes it uniqueOfflineFirstNoteRepository
DTO<Model>DtoNoteDto
Room entity<Model>EntityNoteEntity
Mapperextension fun on source typefun NoteDto.toNote()

Checklist: Adding a New Data Source or Repository

  • Define domain model(s) in feature:domain
  • Define data source or repository interface in feature:domain
  • Define feature-specific error type(s) in feature:domain (implement Error) — see kmp-error-handling skill
  • Define DTOs and Room entities in feature:data
  • Write mappers as extension functions in feature:data
  • Implement data source (single source) or repository (multi-source) in feature:data (commonMain), named for what makes it unique
  • Keep anything that needs a platform API behind an interface or expect/actual, with implementations in androidMain and iosMain

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.