---

name: kmp-repository-creator description: | Generate Repository classes for Kotlin Multiplatform (KMP) projects using Ktor for networking and multiplatform-settings for local storage. Use when: (1) Creating data layer repositories, (2) Implementing network data fetching with Ktor, (3) Adding local caching with multiplatform-settings, (4) Writing repository unit tests, (5) Following TDD workflow for repository implementation. Triggers: repository, data layer, ktor, multiplatform-settings, network fetch, local storage, cache, data source, TDD repository, repository pattern, kmp repository. context: fork agent: general-purpose user-invocable: true allowed-tools:

• Read
• Write
• Edit
• Glob
• Grep
• Bash

---

KMP Repository Creator

Generate Repository classes following TDD methodology with Ktor networking and multiplatform-settings caching.

Coding Standards

MANDATORY: Follow kotlin-compose-standards:

• Trailing comma EXCEPT for Modifier arguments
• New line at end of file

TDD Workflow

Always follow Red-Green-Refactor cycle:

1. RED - Write failing test first
2. GREEN - Implement minimal code to pass
3. REFACTOR - Improve code while keeping tests green

See references/tdd-workflow.md for detailed TDD patterns.

Repository Structure

Copyfeature/
  data/
    repository/
      UserRepository.kt          # Interface
      UserRepositoryImpl.kt      # Implementation
    datasource/
      UserRemoteDataSource.kt    # Ktor network calls
      UserLocalDataSource.kt     # multiplatform-settings
    model/
      UserDto.kt                 # Network DTOs
  domain/
    model/
      User.kt                    # Domain models

Quick Start

1. Define Repository Interface

Copyinterface UserRepository {
    suspend fun getUser(id: String): User
    suspend fun getUsers(): List<User>
    fun observeUser(id: String): Flow<User?>
}

2. Write Tests First (TDD)

Copyclass UserRepositoryImplTest {
    private lateinit var repository: UserRepository
    private val remoteDataSource = FakeUserRemoteDataSource()
    private val localDataSource = FakeUserLocalDataSource()

    @BeforeTest
    fun setup() {
        repository = UserRepositoryImpl(remoteDataSource, localDataSource)
    }

    @Test
    fun `getUser returns cached user when available`() = runTest {
        // Given
        val cachedUser = User(id = "1", name = "John")
        localDataSource.saveUser(cachedUser)

        // When
        val result = repository.getUser("1")

        // Then
        assertEquals(cachedUser, result)
    }

    @Test
    fun `getUser fetches from remote when cache empty`() = runTest {
        // Given
        val remoteUser = UserDto(id = "1", name = "John")
        remoteDataSource.setResponse(remoteUser)

        // When
        val result = repository.getUser("1")

        // Then
        assertEquals("John", result.name)
    }

    @Test
    fun `getUser throws exception when network fails`() = runTest {
        // Given
        remoteDataSource.setError(IOException("Network error"))

        // When/Then
        assertFailsWith<IOException> {
            repository.getUser("1")
        }
    }
}

3. Implement Repository

Copyclass UserRepositoryImpl(
    private val remoteDataSource: UserRemoteDataSource,
    private val localDataSource: UserLocalDataSource,
) : UserRepository {

    override suspend fun getUser(id: String): User {
        // Try cache first
        localDataSource.getUser(id)?.let { return it }

        // Fetch from remote
        val dto = remoteDataSource.getUser(id)
        val user = dto.toDomain()
        localDataSource.saveUser(user)
        return user
    }

    override suspend fun getUsers(): List<User> {
        return remoteDataSource.getUsers().map { it.toDomain() }
    }

    override fun observeUser(id: String): Flow<User?> {
        return localDataSource.observeUser(id)
    }
}

Workflow

1. Analyze requirements - What data? Caching strategy?
2. Define interface - Repository contract with suspend functions
3. Write tests - Cover cache hit, cache miss, error cases
4. Implement minimal - Just enough to pass tests
5. Refactor - Extract data sources, improve error handling
6. Verify - Run tests, ensure all pass

References

Koin Integration

Repository injected via Koin DSL:

Copyval dataModule = module {
    // Data sources
    single { UserRemoteDataSource(get()) }
    single { UserLocalDataSource(get()) }

    // Repository
    single<UserRepository> { UserRepositoryImpl(get(), get()) }
}

ViewModel receives repository:

Copyclass UserViewModel(
    private val userRepository: UserRepository,
) : ViewModel() {

    private val _uiState = MutableStateFlow<UserUiState>(UserUiState.Loading)
    val uiState: StateFlow<UserUiState> = _uiState.asStateFlow()

    fun loadUser(id: String) {
        viewModelScope.launch {
            _uiState.value = UserUiState.Loading
            try {
                val user = userRepository.getUser(id)
                _uiState.value = UserUiState.Success(user)
            } catch (e: Exception) {
                _uiState.value = UserUiState.Error(e.message)
            }
        }
    }
}

Best Practices

• Repository throws exceptions for errors (ViewModel handles with try-catch)
• Use Flow<T> for observable data
• Keep domain models separate from DTOs
• Cache strategy: cache-first with network fallback
• Test all paths: success, cache hit, network error