ENG-1846

Package.swift

@@ -92,6 +92,7 @@ let package = Package(
     ],
     dependencies: [
         .package(url: "https://github.com/vmanot/CorePersistence.git", branch: "main"),
+        .package(url: "https://github.com/vmanot/Media", branch: "main"),
         .package(url: "https://github.com/vmanot/Merge.git", branch: "master"),
         .package(url: "https://github.com/vmanot/NetworkKit.git", branch: "master"),
         .package(url: "https://github.com/vmanot/Swallow.git", branch: "master"),
@@ -115,7 +116,7 @@ let package = Package(
                 "Merge",
                 "NetworkKit",
                 "Swallow",
-                "SwiftUIX",
+                "SwiftUIX"
             ],
             path: "Sources/LargeLanguageModels",
             resources: [
@@ -191,7 +192,8 @@ let package = Package(
                 "LargeLanguageModels",
                 "Merge",
                 "NetworkKit",
-                "Swallow"
+                "Swallow",
+                "Media"
             ],
             path: "Sources/_Gemini",
             swiftSettings: [
@@ -389,6 +391,7 @@ let package = Package(
                 "Ollama",
                 "OpenAI",
                 "Swallow",
+                "NeetsAI",
             ],
             path: "Sources/AI",
             swiftSettings: [

Sources/AI/AnySpeechSynthesisRequestHandling.swift

@@ -0,0 +1,42 @@
+//
+//  AnySpeechSynthesisRequestHandling.swift
+//  AI
+//
+//  Created by Jared Davidson on 1/14/25.
+//
+
+import ElevenLabs
+import LargeLanguageModels
+import NeetsAI
+
+public struct AnySpeechSynthesisRequestHandling: Hashable {
+    private let _hashValue: Int
+
+    public let base: any CoreMI._ServiceClientProtocol & SpeechSynthesisRequestHandling
+
+    public var displayName: String {
+        switch base {
+            case is ElevenLabs.Client:
+                return "ElevenLabs"
+            case is NeetsAI.Client:
+                return "NeetsAI"
+            default:
+                fatalError()
+        }
+    }
+
+    public init(
+        _ base: any CoreMI._ServiceClientProtocol & SpeechSynthesisRequestHandling
+    ) {
+        self.base = base
+        self._hashValue = ObjectIdentifier(base as AnyObject).hashValue
+    }
+
+    public static func == (lhs: AnySpeechSynthesisRequestHandling, rhs: AnySpeechSynthesisRequestHandling) -> Bool {
+        lhs._hashValue == rhs._hashValue
+    }
+
+    public func hash(into hasher: inout Hasher) {
+        hasher.combine(_hashValue)
+    }
+}

Sources/ElevenLabs/Intramodular/API/ElevenLabs.APISpecification.swift

@@ -128,9 +128,6 @@ extension ElevenLabs.APISpecification {
             context: DecodeOutputContext
         ) throws -> Output {
             do {
-                if Input.self == RequestBodies.EditVoiceInput.self {
-                    print("TEsts")
-                }
                 try response.validate()
             } catch {
                 let apiError: Error

Sources/ElevenLabs/Intramodular/ElevenLabs.Client.swift

@@ -10,10 +10,15 @@ import SwiftAPI
 import Merge
 import FoundationX
 import Swallow
+import LargeLanguageModels
 
 extension ElevenLabs {
     @RuntimeDiscoverable
     public final class Client: SwiftAPI.Client, ObservableObject {
+        public static var persistentTypeRepresentation: some IdentityRepresentation {
+            CoreMI._ServiceVendorIdentifier._ElevenLabs
+        }
+
         public typealias API = ElevenLabs.APISpecification
         public typealias Session = HTTPSession
 
@@ -33,6 +38,25 @@ extension ElevenLabs {
     }
 }
 
+extension ElevenLabs.Client: CoreMI._ServiceClientProtocol {
+    public convenience init(
+        account: (any CoreMI._ServiceAccountProtocol)?
+    ) async throws {
+        let account: any CoreMI._ServiceAccountProtocol = try account.unwrap()
+        let serviceVendorIdentifier: CoreMI._ServiceVendorIdentifier = try account.serviceVendorIdentifier.unwrap()
+
+        guard serviceVendorIdentifier == CoreMI._ServiceVendorIdentifier._ElevenLabs else {
+            throw CoreMI._ServiceClientError.incompatibleVendor(serviceVendorIdentifier)
+        }
+
+        guard let credential = try account.credential as? CoreMI._ServiceCredentialTypes.APIKeyCredential else {
+            throw CoreMI._ServiceClientError.invalidCredential(try account.credential)
+        }
+
+        self.init(apiKey: credential.apiKey)
+    }
+}
+
 extension ElevenLabs.Client {
     public func availableVoices() async throws -> [ElevenLabs.Voice] {
         try await run(\.listVoices).voices
@@ -50,7 +74,6 @@ extension ElevenLabs.Client {
             voiceSettings: voiceSettings,
             model: model
         )
-
         return try await run(\.textToSpeech, with: .init(voiceId: voiceID, requestBody: requestBody))
     }
 
@@ -107,3 +130,56 @@ extension ElevenLabs.Client {
         try await run(\.deleteVoice, with: voice.rawValue)
     }
 }
+
+// MARK: - Conformances
+
+extension ElevenLabs.Client: SpeechSynthesisRequestHandling {
+    public func availableVoices() async throws -> [AbstractVoice] {
+        return try await self.availableVoices().map({try $0.__conversion()})
+    }
+
+    public func speech(for text: String, voiceID: String, voiceSettings: AbstractVoiceSettings, model: String) async throws -> Data {
+        try await self.speech(
+            for: text,
+            voiceID: voiceID,
+            voiceSettings: .init(settings: voiceSettings),
+            model: .init(rawValue: model) ?? .MultilingualV1
+        )
+    }
+
+    public func speechToSpeech(inputAudioURL: URL, voiceID: String, voiceSettings: AbstractVoiceSettings, model: String) async throws -> Data {
+        try await self.speechToSpeech(
+            inputAudioURL: inputAudioURL,
+            voiceID: voiceID,
+            voiceSettings: .init(settings: voiceSettings),
+            model: .init(rawValue: model) ?? .MultilingualV1
+        )
+    }
+
+    public func upload(voiceWithName name: String, description: String, fileURL: URL) async throws -> AbstractVoice.ID {
+        let voice: ElevenLabs.Voice.ID = try await self.upload(
+            voiceWithName: name,
+            description: description,
+            fileURL: fileURL
+        )
+
+        return .init(rawValue: voice.rawValue)
+    }
+
+    public func edit(voice: AbstractVoice.ID, name: String, description: String, fileURL: URL?) async throws -> Bool {
+        try await self.edit(
+            voice: ElevenLabs.Voice.ID(rawValue: voice.rawValue),
+            name: name,
+            description: description,
+            fileURL: fileURL
+        )
+    }
+
+    public func delete(voice: AbstractVoice.ID) async throws {
+        try await self.delete(
+            voice: ElevenLabs.Voice.ID(
+                rawValue: voice.rawValue
+            )
+        )
+    }
+}

Sources/ElevenLabs/Intramodular/ElevenLabs.Voice.swift

@@ -4,6 +4,7 @@
 
 import Foundation
 import Swift
+import LargeLanguageModels
 
 extension ElevenLabs {
     public struct Voice: Hashable, Identifiable, Sendable {
@@ -42,3 +43,24 @@ extension ElevenLabs.Voice: Codable {
         case isOwner
     }
 }
+
+extension ElevenLabs.Voice: AbstractVoiceConvertible {
+    public func __conversion() throws -> AbstractVoice {
+        return AbstractVoice(
+            voiceID: self.voiceID,
+            name: self.name,
+            description: self.description
+        )
+    }
+}
+
+extension ElevenLabs.Voice: AbstractVoiceInitiable {
+    public init(voice: AbstractVoice) throws {
+        self.init(
+            voiceID: voice.voiceID,
+            name: voice.name,
+            description: voice.description,
+            isOwner: nil
+        )
+    }
+}

Sources/ElevenLabs/Intramodular/ElevenLabs.VoiceSettings.swift

@@ -3,6 +3,7 @@
 //
 
 import Foundation
+import LargeLanguageModels
 
 extension ElevenLabs {
     public struct VoiceSettings: Codable, Sendable, Hashable {
@@ -98,3 +99,29 @@ extension ElevenLabs.VoiceSettings {
         )
     }
 }
+
+// MARK: - Conformances
+
+extension ElevenLabs.VoiceSettings: AbstractVoiceSettingsConvertible {
+    public func __conversion() throws -> AbstractVoiceSettings {
+        return .init(
+            stability: stability,
+            similarityBoost: similarityBoost,
+            styleExaggeration: styleExaggeration,
+            speakerBoost: speakerBoost,
+            removeBackgroundNoise: removeBackgroundNoise
+        )
+    }
+}
+
+extension ElevenLabs.VoiceSettings: AbstractVoiceSettingsInitiable {
+    public init(settings: AbstractVoiceSettings) throws {
+        self.init(
+            stability: settings.stability,
+            similarityBoost: settings.similarityBoost,
+            styleExaggeration: settings.styleExaggeration,
+            speakerBoost: settings.speakerBoost,
+            removeBackgroundNoise: settings.removeBackgroundNoise
+        )
+    }
+}

Sources/HumeAI/Intramodular/HumeAI.Client+SpeechSynthesisRequestHandling.swift

@@ -0,0 +1,44 @@
+//
+//  HumeAI+ElevenLabsClientProtocol.swift
+//  Voice
+//
+//  Created by Jared Davidson on 11/22/24.
+//
+
+import Foundation
+import SwiftUI
+import AVFoundation
+import LargeLanguageModels
+
+extension HumeAI.Client: SpeechSynthesisRequestHandling {
+    public func availableVoices() async throws -> [AbstractVoice] {
+        return try await getAllAvailableVoices().map(
+            { voice in
+                return AbstractVoice(
+                    voiceID: voice.id,
+                    name: voice.name,
+                    description: nil
+                )
+        })
+    }
+
+    public func speech(for text: String, voiceID: String, voiceSettings: AbstractVoiceSettings, model: String) async throws -> Data {
+        throw HumeAI.APIError.unknown(message: "Text to speech not supported")
+    }
+
+    public func speechToSpeech(inputAudioURL: URL, voiceID: String, voiceSettings: AbstractVoiceSettings, model: String) async throws -> Data {
+        throw HumeAI.APIError.unknown(message: "Speech to speech not supported")
+    }
+
+    public func upload(voiceWithName name: String, description: String, fileURL: URL) async throws -> AbstractVoice.ID {
+        throw HumeAI.APIError.unknown(message: "Voice creation is not supported")
+    }
+
+    public func edit(voice: AbstractVoice.ID, name: String, description: String, fileURL: URL?) async throws -> Bool {
+        throw HumeAI.APIError.unknown(message: "Voice creation is not supported")
+    }
+
+    public func delete(voice: AbstractVoice.ID) async throws {
+        throw HumeAI.APIError.unknown(message: "Voice creation is not supported")
+    }
+}

Sources/LargeLanguageModels/Intramodular/AbstractVoice (WIP)/AbstractVoice.swift

@@ -0,0 +1,41 @@
+//
+//  AudioStore.swift
+//  Voice
+//
+//  Created by Jared Davidson on 10/31/24.
+//
+
+import CorePersistence
+import SwiftUI
+import AVFoundation
+import UniformTypeIdentifiers
+
+public struct AbstractVoice: Codable, Hashable, Identifiable, Sendable {
+    public typealias ID = _TypeAssociatedID<Self, String>
+
+    public let id: ID
+    public let voiceID: String
+    public let name: String
+    public let description: String?
+
+    public init(
+        voiceID: String,
+        name: String,
+        description: String?
+    ) {
+        self.id = .init(rawValue: voiceID)
+        self.voiceID = voiceID
+        self.name = name
+        self.description = description
+    }
+}
+
+// MARK: - Conformances
+
+public protocol AbstractVoiceInitiable {
+    init(voice: AbstractVoice) throws
+}
+
+public protocol AbstractVoiceConvertible {
+    func __conversion() throws -> AbstractVoice
+}