Release 1.1.8

sgusakovsky · Mar 5, 2023 · 3b9889b · 3b9889b
1 parent 63118c2
commit 3b9889b
Show file tree

Hide file tree

Showing 8 changed files with 176 additions and 6 deletions.
diff --git a/OpenAIService.podspec b/OpenAIService.podspec
@@ -1,6 +1,6 @@
 Pod::Spec.new do |spec|
   spec.name         = 'OpenAIService'
-  spec.version      = '1.1.7'
+  spec.version      = '1.1.8'
   spec.homepage     = 'https://github.com/sgusakovsky/OpenAIService'
   spec.license      = {
     :type => 'MIT',

diff --git a/README.md b/README.md
@@ -147,6 +147,23 @@ service?.sendImageVariation(with: body, completionHandler: { result in
 ```
 The API will return an `OpenAIImageResponse` object containing the corresponding image url items.
 
+Create a call to the audio transcription API.
+
+```swift
+let urlPath = Bundle.main.url(forResource: "audio", withExtension: "mp3")!
+let data = try? Data(contentsOf: urlPath)!
+let body = OpenAIAudioTranscriptionBody(file: data, fileFormat: .mp3)
+service?.sendAudioTranscription(with: body, completionHandler: { result in
+    switch result {
+    case .success(let response):
+        print(response.text)
+    case .failure(let error):
+        print(error.localizedDescription)
+    }
+})
+```
+The API will return an `OpenAIAudioResponse` object containing the corresponding image url items.
+
 For a full list of the supported models see [OpenAICompletionModelType.swift](https://github.com/sgusakovsky/OpenAIService/blob/main/Sources/OpenAIService/Models/Completion/OpenAICompletionModelType.swift), [OpenAIChatModelType.swift](https://github.com/sgusakovsky/OpenAIService/blob/main/Sources/OpenAIService/Models/Chat/OpenAIChatModelType.swift), [OpenAIEditsModelType.swift](https://github.com/sgusakovsky/OpenAIService/blob/main/Sources/OpenAIService/Models/Edits/OpenAIEditsModelType.swift). For more information on the models see the [OpenAI API Documentation](https://platform.openai.com/docs/models).
 
 OpenAIService also supports Swift concurrency so you can use Swift’s async/await syntax to fetch completions.

diff --git a/Sources/OpenAIService/Models/Audio/AudioFileFormat.swift b/Sources/OpenAIService/Models/Audio/AudioFileFormat.swift
@@ -0,0 +1,25 @@
+//
+//  AudioFileFormat.swift
+//  OpenAIDemo
+//
+//  Created by Gusakovsky, Sergey on 5.03.23.
+//
+
+import Foundation
+
+public enum AudioFileFormat: String {
+    case mp3, mp4, mpeg, mpga, m4a, wav, webm
+
+    public var mimeType: String {
+        switch self {
+        case .mp3, .mpeg, .mpga:
+            return "audio/mpeg"
+        case .mp4, .m4a:
+            return "audio/mp4"
+        case .wav:
+            return "audio/x-wav"
+        case .webm:
+            return "audio/webm"
+        }
+    }
+}
diff --git a/Sources/OpenAIService/Models/Audio/OpenAIAudioResponse.swift b/Sources/OpenAIService/Models/Audio/OpenAIAudioResponse.swift
@@ -0,0 +1,12 @@
+//
+//  OpenAIAudioResponse.swift
+//  OpenAIDemo
+//
+//  Created by Gusakovsky, Sergey on 5.03.23.
+//
+
+import Foundation
+
+public struct OpenAIAudioResponse: Codable {
+    public let text: String
+}
diff --git a/Sources/OpenAIService/Models/Audio/Transcriptions/OpenAIAudioModelType.swift b/Sources/OpenAIService/Models/Audio/Transcriptions/OpenAIAudioModelType.swift
@@ -0,0 +1,14 @@
+//
+//  OpenAIAudioModelType.swift
+//  OpenAIDemo
+//
+//  Created by Gusakovsky, Sergey on 5.03.23.
+//
+
+import Foundation
+
+public enum OpenAIAudioModelType: String, Encodable {
+
+    /// > Model Name: whisper
+    case whisper = "whisper-1"
+}
diff --git a/Sources/OpenAIService/Models/Audio/Transcriptions/OpenAIAudioTranscriptionBody.swift b/Sources/OpenAIService/Models/Audio/Transcriptions/OpenAIAudioTranscriptionBody.swift
@@ -0,0 +1,50 @@
+//
+//  OpenAIAudioTranscriptionBody.swift
+//  OpenAIDemo
+//
+//  Created by Gusakovsky, Sergey on 5.03.23.
+//
+
+import Foundation
+
+/// https://platform.openai.com/docs/api-reference/audio/create
+public struct OpenAIAudioTranscriptionBody {
+    public let file: FormData
+    public let model: OpenAIAudioModelType
+    public let prompt: String?
+    public let temperature: Double
+    public let language: String?
+
+    public init(
+        file: Data,
+        fileFormat: AudioFileFormat,
+        model: OpenAIAudioModelType = .whisper,
+        prompt: String? = nil,
+        temperature: Double = 0,
+        language: String? = nil
+    ) {
+        self.file = FormData(data: file, mimeType: fileFormat.mimeType, fileName: "aiduo.\(fileFormat.rawValue)")
+        self.model = model
+        self.prompt = prompt
+        self.temperature = temperature
+        self.language = language
+    }
+
+    public var body: [String: Any] {
+        var result: [String: Any] = [
+            "file": self.file,
+            "model": self.model.rawValue,
+            "temperature": self.temperature
+        ]
+
+        if let prompt = self.prompt {
+            result["prompt"] = prompt
+        }
+
+        if let language = self.language {
+            result["language"] = language
+        }
+
+        return result
+    }
+}
diff --git a/Sources/OpenAIService/Networking/OpenAIEndpoint.swift b/Sources/OpenAIService/Networking/OpenAIEndpoint.swift
@@ -14,6 +14,7 @@ enum OpenAIEndpoint {
     case imagesGenerations
     case imageEdits
     case imageVariation
+    case audioTranscriptions
 
     var path: String {
         switch self {
@@ -29,19 +30,21 @@ enum OpenAIEndpoint {
             return "/v1/images/edits"
         case .imageVariation:
             return "/v1/images/variations"
+        case .audioTranscriptions:
+            return "/v1/audio/transcriptions"
         }
     }
 
     var method: HTTPMethod {
         switch self {
-        case .completions, .edits, .chatCompletions, .imagesGenerations, .imageEdits, .imageVariation:
+        case .completions, .edits, .chatCompletions, .imagesGenerations, .imageEdits, .imageVariation, .audioTranscriptions:
             return .post
         }
     }
 
     func baseURL() -> String {
         switch self {
-        case .completions, .edits, .chatCompletions, .imagesGenerations, .imageEdits, .imageVariation:
+        case .completions, .edits, .chatCompletions, .imagesGenerations, .imageEdits, .imageVariation, .audioTranscriptions:
             return "https://api.openai.com"
         }
     }

diff --git a/Sources/OpenAIService/OpenAIService.swift b/Sources/OpenAIService/OpenAIService.swift
@@ -163,6 +163,33 @@ public final class OpenAIService {
         )
     }
 
+    /// Send a audio transcription request to the OpenAI API
+    /// - Parameters:
+    ///   - body: Body of audio transcription  request
+    ///   - completionHandler: Returns an OpenAIAudioResponse Data Model
+    public func sendAudioTranscription(
+        with body: OpenAIAudioTranscriptionBody,
+        networkQueue: DispatchQueue = .global(qos: .background),
+        responseQueue: DispatchQueue = .main,
+        completionHandler: @escaping (Result<OpenAIAudioResponse, OpenAIAPIError>) -> Void
+    ) {
+        let endpoint = OpenAIEndpoint.audioTranscriptions
+        guard let request = apiClient.prepareMultipartFormDataRequest(endpoint, body: body.body, config: config) else {
+            completionHandler(.failure(.genericError(error: RequestError())))
+            return
+        }
+
+        apiClient.makeRequest(
+            request: request,
+            networkQueue: networkQueue,
+            responseQueue: responseQueue,
+            completionHandler: completionHandler
+        )
+    }
+
+}
+
+extension OpenAIService {
     /// Send a Completion to the OpenAI API
     /// - Parameters:
     ///   - body: Body of chat completion request
@@ -232,7 +259,7 @@ public final class OpenAIService {
     /// Send a Image generation request to the OpenAI API
     /// - Parameters:
     ///   - body: Body of image generation request
-    /// - Returns: Returns an OpenAIGenerationImageResponse Data Model
+    /// - Returns: Returns an OpenAIImageResponse Data Model
     @available(swift 5.5)
     @available(macOS 10.15, iOS 13, watchOS 6, tvOS 13, *)
     public func sendImageGeneration(
@@ -254,7 +281,7 @@ public final class OpenAIService {
     /// Send a Image edits request to the OpenAI API
     /// - Parameters:
     ///   - body: Body of image edits request
-    /// - Returns: Returns an OpenAIGenerationImageResponse Data Model
+    /// - Returns: Returns an OpenAIImageResponse Data Model
     @available(swift 5.5)
     @available(macOS 10.15, iOS 13, watchOS 6, tvOS 13, *)
     public func sendImageEdits(
@@ -276,7 +303,7 @@ public final class OpenAIService {
     /// Send a Image vatiation request to the OpenAI API
     /// - Parameters:
     ///   - body: Body of image variation request
-    /// - Returns: Returns an OpenAIGenerationImageResponse Data Model
+    /// - Returns: Returns an OpenAIImageResponse Data Model
     @available(swift 5.5)
     @available(macOS 10.15, iOS 13, watchOS 6, tvOS 13, *)
     public func sendImageVariation(
@@ -294,4 +321,26 @@ public final class OpenAIService {
             }
         }
     }
+
+    /// Send a Image vatiation request to the OpenAI API
+    /// - Parameters:
+    ///   - body: Body of audio transcription  request
+    /// - Returns: Returns an OpenAIAudioResponse Data Model
+    @available(swift 5.5)
+    @available(macOS 10.15, iOS 13, watchOS 6, tvOS 13, *)
+    public func sendAudioTranscription(
+        with body: OpenAIAudioTranscriptionBody,
+        networkQueue: DispatchQueue = .global(qos: .background),
+        responseQueue: DispatchQueue = .main
+    ) async throws -> OpenAIAudioResponse {
+        return try await withCheckedThrowingContinuation { continuation in
+            sendAudioTranscription(
+                with: body,
+                networkQueue: networkQueue,
+                responseQueue: responseQueue
+            ) { result in
+                continuation.resume(with: result)
+            }
+        }
+    }
 }