feature: Add audio transcriptions entrypoint

osugikoji · osugikoji · commit 08c4e9c988a1 · 2023-04-04T01:24:01.000-03:00
diff --git a/ychat/src/androidMain/kotlin/co/yml/ychat/domain/model/FileBytes.kt b/ychat/src/androidMain/kotlin/co/yml/ychat/domain/model/FileBytes.kt
@@ -0,0 +1,7 @@
+package co.yml.ychat.domain.model
+
+actual typealias FileBytes = ByteArray
+
+actual fun FileBytes.toByteArray(): ByteArray {
+    return this
+}
diff --git a/ychat/src/commonMain/kotlin/co/yml/ychat/YChat.kt b/ychat/src/commonMain/kotlin/co/yml/ychat/YChat.kt
@@ -1,5 +1,6 @@
 package co.yml.ychat
 
+import co.yml.ychat.entrypoint.features.AudioTranscriptions
 import co.yml.ychat.entrypoint.features.ChatCompletions
 import co.yml.ychat.entrypoint.features.Completion
 import co.yml.ychat.entrypoint.features.Edits
@@ -137,6 +138,8 @@ interface YChat {
      */
     fun edits(): Edits
 
+    fun audioTranscriptions(): AudioTranscriptions
+
     /**
      * Callback is an interface used for handling the results of an operation.
      * It provides two methods, `onSuccess` and `onError`, for handling the success
diff --git a/ychat/src/commonMain/kotlin/co/yml/ychat/data/api/ChatGptApi.kt b/ychat/src/commonMain/kotlin/co/yml/ychat/data/api/ChatGptApi.kt
@@ -1,5 +1,7 @@
 package co.yml.ychat.data.api
 
+import co.yml.ychat.data.dto.AudioParamsDto
+import co.yml.ychat.data.dto.AudioResultDto
 import co.yml.ychat.data.dto.ChatCompletionParamsDto
 import co.yml.ychat.data.dto.ChatCompletionsDto
 import co.yml.ychat.data.dto.CompletionDto
@@ -25,4 +27,6 @@ internal interface ChatGptApi {
     suspend fun models(): ApiResult<ModelListDto>
 
     suspend fun model(id: String): ApiResult<ModelDto>
+
+    suspend fun audioTranscriptions(audioParamsDto: AudioParamsDto): ApiResult<AudioResultDto>
 }
diff --git a/ychat/src/commonMain/kotlin/co/yml/ychat/data/api/impl/ChatGptApiImpl.kt b/ychat/src/commonMain/kotlin/co/yml/ychat/data/api/impl/ChatGptApiImpl.kt
@@ -1,6 +1,8 @@
 package co.yml.ychat.data.api.impl
 
 import co.yml.ychat.data.api.ChatGptApi
+import co.yml.ychat.data.dto.AudioParamsDto
+import co.yml.ychat.data.dto.AudioResultDto
 import co.yml.ychat.data.dto.ChatCompletionParamsDto
 import co.yml.ychat.data.dto.ChatCompletionsDto
 import co.yml.ychat.data.dto.CompletionDto
@@ -13,6 +15,7 @@ import co.yml.ychat.data.dto.ModelDto
 import co.yml.ychat.data.dto.ModelListDto
 import co.yml.ychat.data.infrastructure.ApiExecutor
 import co.yml.ychat.data.infrastructure.ApiResult
+import co.yml.ychat.domain.model.toByteArray
 import io.ktor.http.HttpMethod
 
 internal class ChatGptApiImpl(private val apiExecutor: ApiExecutor) : ChatGptApi {
@@ -62,4 +65,14 @@ internal class ChatGptApiImpl(private val apiExecutor: ApiExecutor) : ChatGptApi
             .setHttpMethod(HttpMethod.Get)
             .execute()
     }
+
+    override suspend fun audioTranscriptions(audioParamsDto: AudioParamsDto): ApiResult<AudioResultDto> {
+        val byteArray = audioParamsDto.byteArray.toByteArray()
+        val apiBuilder = apiExecutor
+            .setEndpoint("v1/audio/transcriptions")
+            .setHttpMethod(HttpMethod.Post)
+            .addFormPart("file", audioParamsDto.filename, byteArray)
+        audioParamsDto.getMap().forEach { apiBuilder.addFormPart(it.key, it.value) }
+        return apiBuilder.execute()
+    }
 }
diff --git a/ychat/src/commonMain/kotlin/co/yml/ychat/data/dto/AudioParamsDto.kt b/ychat/src/commonMain/kotlin/co/yml/ychat/data/dto/AudioParamsDto.kt
@@ -0,0 +1,24 @@
+package co.yml.ychat.data.dto
+
+import co.yml.ychat.domain.model.FileBytes
+
+internal data class AudioParamsDto(
+    val filename: String,
+    val byteArray: FileBytes,
+    val model: String,
+    val prompt: String,
+    val responseFormat: String,
+    val temperature: Double,
+    val language: String,
+) {
+
+    fun getMap(): Map<String, Any> {
+        return mapOf(
+            "model" to model,
+            "prompt" to prompt,
+            "response_format" to responseFormat,
+            "temperature" to temperature,
+            "language" to language,
+        )
+    }
+}
diff --git a/ychat/src/commonMain/kotlin/co/yml/ychat/data/dto/AudioResultDto.kt b/ychat/src/commonMain/kotlin/co/yml/ychat/data/dto/AudioResultDto.kt
@@ -0,0 +1,10 @@
+package co.yml.ychat.data.dto
+
+import kotlinx.serialization.SerialName
+import kotlinx.serialization.Serializable
+
+@Serializable
+internal data class AudioResultDto(
+    @SerialName("text")
+    val text: String,
+)
diff --git a/ychat/src/commonMain/kotlin/co/yml/ychat/data/infrastructure/ApiExecutor.kt b/ychat/src/commonMain/kotlin/co/yml/ychat/data/infrastructure/ApiExecutor.kt
@@ -1,16 +1,18 @@
 package co.yml.ychat.data.infrastructure
 
-import co.yml.ychat.data.exception.ChatGptException
 import io.ktor.client.HttpClient
-import io.ktor.client.call.body
 import io.ktor.client.plugins.ResponseException
+import io.ktor.client.request.forms.FormPart
+import io.ktor.client.request.forms.formData
+import io.ktor.client.request.forms.submitFormWithBinaryData
 import io.ktor.client.request.request
 import io.ktor.client.request.setBody
 import io.ktor.client.statement.HttpResponse
 import io.ktor.client.utils.EmptyContent
+import io.ktor.http.ContentType
+import io.ktor.http.Headers
+import io.ktor.http.HttpHeaders
 import io.ktor.http.HttpMethod
-import io.ktor.http.isSuccess
-import io.ktor.util.toMap
 import io.ktor.utils.io.errors.IOException
 import kotlin.collections.set
 
@@ -24,6 +26,8 @@ internal class ApiExecutor(private val httpClient: HttpClient) {
 
     private var query: HashMap<String, String> = HashMap()
 
+    private val formParts = mutableListOf<FormPart<*>>()
+
     fun setEndpoint(endpoint: String): ApiExecutor {
         this.endpoint = endpoint
         return this
@@ -49,13 +53,23 @@ internal class ApiExecutor(private val httpClient: HttpClient) {
         return this
     }
 
+    fun <T : Any> addFormPart(key: String, value: T): ApiExecutor {
+        formParts += FormPart(key, value)
+        return this
+    }
+
+    fun addFormPart(key: String, fileName: String, value: ByteArray): ApiExecutor {
+        val headers = Headers.build {
+            append(HttpHeaders.ContentType, ContentType.Application.OctetStream.contentType)
+            append(HttpHeaders.ContentDisposition, "filename=$fileName")
+        }
+        formParts += FormPart(key, value, headers = headers)
+        return this
+    }
+
     suspend inline fun <reified T> execute(): ApiResult<T> {
         return try {
-            val response = httpClient.request(endpoint) {
-                url { query.forEach { parameters.append(it.key, it.value) } }
-                method = httpMethod
-                setBody(this@ApiExecutor.body)
-            }
+            val response = if (formParts.isEmpty()) executeRequest() else executeRequestAsForm()
             return response.toApiResult()
         } catch (responseException: ResponseException) {
             responseException.toApiResult()
@@ -64,28 +78,18 @@ internal class ApiExecutor(private val httpClient: HttpClient) {
         }
     }
 
-    private suspend inline fun <reified T> HttpResponse.toApiResult(): ApiResult<T> {
-        val headers = this.headers.toMap()
-        val statusCode = this.status.value
-        return if (!this.status.isSuccess()) {
-            val exception = ChatGptException(null, statusCode)
-            ApiResult(null, headers, statusCode, exception)
-        } else {
-            ApiResult(this.body<T>(), headers, statusCode, null)
+    private suspend fun executeRequest(): HttpResponse {
+        return httpClient.request(endpoint) {
+            url { query.forEach { parameters.append(it.key, it.value) } }
+            method = httpMethod
+            setBody(this@ApiExecutor.body)
         }
     }
 
-    private fun <T> ResponseException.toApiResult(): ApiResult<T> {
-        return ApiResult(
-            statusCode = this.response.status.value,
-            exception = ChatGptException(this.cause, this.response.status.value)
-        )
-    }
-
-    private fun <T> Throwable.toApiResult(): ApiResult<T> {
-        return ApiResult(
-            statusCode = null,
-            exception = ChatGptException(this.cause)
+    private suspend fun executeRequestAsForm(): HttpResponse {
+        return httpClient.submitFormWithBinaryData(
+            url = endpoint,
+            formData = formData { formParts.forEach { append(it) } }
         )
     }
 }
diff --git a/ychat/src/commonMain/kotlin/co/yml/ychat/data/infrastructure/ApiResultExtensions.kt b/ychat/src/commonMain/kotlin/co/yml/ychat/data/infrastructure/ApiResultExtensions.kt
@@ -0,0 +1,33 @@
+package co.yml.ychat.data.infrastructure
+
+import co.yml.ychat.data.exception.ChatGptException
+import io.ktor.client.call.body
+import io.ktor.client.plugins.ResponseException
+import io.ktor.client.statement.HttpResponse
+import io.ktor.http.isSuccess
+import io.ktor.util.toMap
+
+internal suspend inline fun <reified T> HttpResponse.toApiResult(): ApiResult<T> {
+    val headers = this.headers.toMap()
+    val statusCode = this.status.value
+    return if (!this.status.isSuccess()) {
+        val exception = ChatGptException(null, statusCode)
+        ApiResult(null, headers, statusCode, exception)
+    } else {
+        ApiResult(this.body<T>(), headers, statusCode, null)
+    }
+}
+
+internal fun <T> ResponseException.toApiResult(): ApiResult<T> {
+    return ApiResult(
+        statusCode = this.response.status.value,
+        exception = ChatGptException(this.cause, this.response.status.value)
+    )
+}
+
+internal fun <T> Throwable.toApiResult(): ApiResult<T> {
+    return ApiResult(
+        statusCode = null,
+        exception = ChatGptException(this.cause)
+    )
+}
diff --git a/ychat/src/commonMain/kotlin/co/yml/ychat/di/module/LibraryModule.kt b/ychat/src/commonMain/kotlin/co/yml/ychat/di/module/LibraryModule.kt
@@ -5,18 +5,21 @@ import co.yml.ychat.data.api.impl.ChatGptApiImpl
 import co.yml.ychat.data.infrastructure.ApiExecutor
 import co.yml.ychat.data.storage.ChatLogStorage
 import co.yml.ychat.di.provider.NetworkProvider
+import co.yml.ychat.domain.usecases.AudioUseCase
 import co.yml.ychat.domain.usecases.ChatCompletionsUseCase
 import co.yml.ychat.domain.usecases.CompletionUseCase
 import co.yml.ychat.domain.usecases.EditsUseCase
 import co.yml.ychat.domain.usecases.ImageGenerationsUseCase
 import co.yml.ychat.domain.usecases.ListModelsUseCase
 import co.yml.ychat.domain.usecases.RetrieveModelUseCase
+import co.yml.ychat.entrypoint.features.AudioTranscriptions
 import co.yml.ychat.entrypoint.features.ChatCompletions
 import co.yml.ychat.entrypoint.features.Completion
 import co.yml.ychat.entrypoint.features.Edits
 import co.yml.ychat.entrypoint.features.ImageGenerations
 import co.yml.ychat.entrypoint.features.ListModels
 import co.yml.ychat.entrypoint.features.RetrieveModel
+import co.yml.ychat.entrypoint.impl.AudioTranscriptionsImpl
 import co.yml.ychat.entrypoint.impl.ChatCompletionsImpl
 import co.yml.ychat.entrypoint.impl.CompletionImpl
 import co.yml.ychat.entrypoint.impl.EditsImpl
@@ -39,6 +42,7 @@ internal class LibraryModule(private val apiKey: String) {
         factory<ChatCompletions> { ChatCompletionsImpl(Dispatchers.Default, get()) }
         factory<ImageGenerations> { ImageGenerationsImpl(Dispatchers.Default, get()) }
         factory<Edits> { EditsImpl(Dispatchers.Default, get()) }
+        factory<AudioTranscriptions> { AudioTranscriptionsImpl(Dispatchers.Default, get()) }
     }
 
     private val domainModule = module {
@@ -48,6 +52,7 @@ internal class LibraryModule(private val apiKey: String) {
         factory { ChatCompletionsUseCase(get()) }
         factory { ImageGenerationsUseCase(get()) }
         factory { EditsUseCase(get()) }
+        factory { AudioUseCase(get()) }
     }
 
     private val dataModule = module {
diff --git a/ychat/src/commonMain/kotlin/co/yml/ychat/domain/mapper/AudioMapper.kt b/ychat/src/commonMain/kotlin/co/yml/ychat/domain/mapper/AudioMapper.kt
@@ -0,0 +1,17 @@
+package co.yml.ychat.domain.mapper
+
+import co.yml.ychat.data.dto.AudioParamsDto
+import co.yml.ychat.domain.model.AudioParams
+import co.yml.ychat.domain.model.FileBytes
+
+internal fun AudioParams.toAudioParamsDto(filename: String, fileBytes: FileBytes): AudioParamsDto {
+    return AudioParamsDto(
+        filename = filename,
+        byteArray = fileBytes,
+        model = this.model,
+        prompt = this.prompt,
+        responseFormat = this.responseFormat,
+        temperature = this.temperature,
+        language = this.language,
+    )
+}
diff --git a/ychat/src/commonMain/kotlin/co/yml/ychat/domain/model/AudioParams.kt b/ychat/src/commonMain/kotlin/co/yml/ychat/domain/model/AudioParams.kt
@@ -0,0 +1,9 @@
+package co.yml.ychat.domain.model
+
+internal data class AudioParams(
+    var model: String = "whisper-1",
+    var prompt: String = "",
+    var responseFormat: String = "json",
+    var temperature: Double = 0.0,
+    var language: String = "en"
+)
diff --git a/ychat/src/commonMain/kotlin/co/yml/ychat/domain/model/FileBytes.kt b/ychat/src/commonMain/kotlin/co/yml/ychat/domain/model/FileBytes.kt
@@ -0,0 +1,5 @@
+package co.yml.ychat.domain.model
+
+expect class FileBytes
+
+internal expect fun FileBytes.toByteArray(): ByteArray
diff --git a/ychat/src/commonMain/kotlin/co/yml/ychat/domain/usecases/AudioUseCase.kt b/ychat/src/commonMain/kotlin/co/yml/ychat/domain/usecases/AudioUseCase.kt
@@ -0,0 +1,19 @@
+package co.yml.ychat.domain.usecases
+
+import co.yml.ychat.data.api.ChatGptApi
+import co.yml.ychat.domain.mapper.toAudioParamsDto
+import co.yml.ychat.domain.model.AudioParams
+import co.yml.ychat.domain.model.FileBytes
+
+internal class AudioUseCase(private val chatGptApi: ChatGptApi) {
+
+    suspend fun requestAudioTranscription(
+        filename: String,
+        audioFile: FileBytes,
+        params: AudioParams
+    ): String {
+        val requestDto = params.toAudioParamsDto(filename, audioFile)
+        val response = chatGptApi.audioTranscriptions(requestDto)
+        return response.getBodyOrThrow().text
+    }
+}
diff --git a/ychat/src/commonMain/kotlin/co/yml/ychat/entrypoint/features/AudioTranscriptions.kt b/ychat/src/commonMain/kotlin/co/yml/ychat/entrypoint/features/AudioTranscriptions.kt
diff --git a/ychat/src/commonMain/kotlin/co/yml/ychat/entrypoint/impl/AudioTranscriptionsImpl.kt b/ychat/src/commonMain/kotlin/co/yml/ychat/entrypoint/impl/AudioTranscriptionsImpl.kt
diff --git a/ychat/src/commonMain/kotlin/co/yml/ychat/entrypoint/impl/YChatImpl.kt b/ychat/src/commonMain/kotlin/co/yml/ychat/entrypoint/impl/YChatImpl.kt
diff --git a/ychat/src/iosMain/kotlin/co/yml/ychat/domain/model/FileBytes.kt b/ychat/src/iosMain/kotlin/co/yml/ychat/domain/model/FileBytes.kt
diff --git a/ychat/src/jvmMain/kotlin/co/yml/ychat/domain/model/FileBytes.kt b/ychat/src/jvmMain/kotlin/co/yml/ychat/domain/model/FileBytes.kt
