Harden Cohere artifact loading and language selection

Info.plist

@@ -13,9 +13,9 @@
 	<key>CFBundleIdentifier</key>
 	<string>$(PRODUCT_BUNDLE_IDENTIFIER)</string>
 	<key>CFBundleVersion</key>
-	<string>9</string>
+		<string>10</string>
 	<key>CFBundleShortVersionString</key>
-	<string>1.5.11-beta.2</string>
+		<string>1.5.11-beta.3</string>
 	<key>LSMinimumSystemVersion</key>
 	<string>$(MACOSX_DEPLOYMENT_TARGET)</string>
 	<key>LSApplicationCategoryType</key>

Sources/Fluid/Persistence/SettingsStore.swift

@@ -2359,7 +2359,7 @@ final class SettingsStore: ObservableObject {
             case .parakeetTDTv2: return "Parakeet TDT v2 (English Only)"
             case .parakeetRealtime: return "Parakeet Flash (Beta)"
             case .qwen3Asr: return "Qwen3 ASR (Beta)"
-            case .cohereTranscribeSixBit: return "Cohere Transcribe 6-bit"
+            case .cohereTranscribeSixBit: return "Cohere Transcribe"
             case .appleSpeech: return "Apple ASR Legacy"
             case .appleSpeechAnalyzer: return "Apple Speech - macOS 26+"
             case .whisperTiny: return "Whisper Tiny"
@@ -2378,7 +2378,7 @@ final class SettingsStore: ObservableObject {
             case .parakeetTDTv2: return "English Only (Higher Accuracy)"
             case .parakeetRealtime: return "English Only (Live Streaming)"
             case .qwen3Asr: return "30 Languages"
-            case .cohereTranscribeSixBit: return "14 Languages"
+            case .cohereTranscribeSixBit: return "14 Languages (Select Manually)"
             case .appleSpeech: return "System Languages"
             case .appleSpeechAnalyzer: return "EN, ES, FR, DE, IT, JA, KO, PT, ZH"
             case .whisperTiny, .whisperBase, .whisperSmall, .whisperMedium, .whisperLargeTurbo, .whisperLarge:
@@ -2532,7 +2532,7 @@ final class SettingsStore: ObservableObject {
             case .qwen3Asr:
                 return "Qwen3 multilingual ASR via FluidAudio. Higher quality, heavier memory footprint."
             case .cohereTranscribeSixBit:
-                return "High-accuracy multilingual transcription. Supports English, French, German, Italian, Spanish, Portuguese, Greek, Dutch, Polish, Mandarin, Japanese, Korean, Vietnamese, and Arabic."
+                return "High-accuracy multilingual transcription. Select the language manually before dictation for best results."
             case .appleSpeech:
                 return "Built-in macOS speech recognition. No download required."
             case .appleSpeechAnalyzer:
@@ -2999,6 +2999,7 @@ private extension SettingsStore {
 
         // Unified Speech Model (replaces above two)
         static let selectedSpeechModel = "SelectedSpeechModel"
+        static let selectedCohereLanguage = "SelectedCohereLanguage"
         static let externalCoreMLArtifactsDirectories = "ExternalCoreMLArtifactsDirectories"
 
         // Overlay Position
@@ -3130,7 +3131,7 @@ extension SettingsStore.SpeechModel {
         case .parakeetRealtime:
             return "EN"
         case .cohereTranscribeSixBit:
-            return "14 Languages"
+            return "AR, DE, EL, EN, ES, FR, IT, JA, KO, NL, PL, PT, VI, ZH"
         case .appleSpeechAnalyzer:
             return "EN, ES, FR, DE, IT, JA, KO, PT, ZH"
         default:
@@ -3144,13 +3145,55 @@ extension SettingsStore.SpeechModel {
             return """
             Bulgarian, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, Greek, Hungarian, Italian, Latvian, Lithuanian, Maltese, Polish, Portuguese, Romanian, Slovak, Slovenian, Spanish, Swedish, Russian, and Ukrainian
             """
+        case .cohereTranscribeSixBit:
+            return "Arabic, German, Greek, English, Spanish, French, Italian, Japanese, Korean, Dutch, Polish, Portuguese, Vietnamese, and Mandarin Chinese"
         default:
             return nil
         }
     }
 }
 
 extension SettingsStore {
+    enum CohereLanguage: String, CaseIterable, Identifiable, Codable {
+        case arabic = "ar"
+        case german = "de"
+        case greek = "el"
+        case english = "en"
+        case spanish = "es"
+        case french = "fr"
+        case italian = "it"
+        case japanese = "ja"
+        case korean = "ko"
+        case dutch = "nl"
+        case polish = "pl"
+        case portuguese = "pt"
+        case vietnamese = "vi"
+        case mandarinChinese = "zh"
+
+        var id: String { self.rawValue }
+
+        var displayName: String {
+            switch self {
+            case .arabic: return "Arabic"
+            case .german: return "German"
+            case .greek: return "Greek"
+            case .english: return "English"
+            case .spanish: return "Spanish"
+            case .french: return "French"
+            case .italian: return "Italian"
+            case .japanese: return "Japanese"
+            case .korean: return "Korean"
+            case .dutch: return "Dutch"
+            case .polish: return "Polish"
+            case .portuguese: return "Portuguese"
+            case .vietnamese: return "Vietnamese"
+            case .mandarinChinese: return "Mandarin Chinese"
+            }
+        }
+
+        var tokenString: String { "<|\(self.rawValue)|>" }
+    }
+
     // MARK: - Unified Speech Model Selection
 
     /// The selected speech recognition model.
@@ -3187,6 +3230,21 @@ extension SettingsStore {
         }
     }
 
+    var selectedCohereLanguage: CohereLanguage {
+        get {
+            if let rawValue = self.defaults.string(forKey: Keys.selectedCohereLanguage),
+               let language = CohereLanguage(rawValue: rawValue)
+            {
+                return language
+            }
+            return .english
+        }
+        set {
+            objectWillChange.send()
+            self.defaults.set(newValue.rawValue, forKey: Keys.selectedCohereLanguage)
+        }
+    }
+
     func externalCoreMLArtifactsDirectory(for model: SpeechModel) -> URL? {
         guard let spec = model.externalCoreMLSpec else { return nil }
         let paths = self.defaults.dictionary(forKey: Keys.externalCoreMLArtifactsDirectories) as? [String: String] ?? [:]

Sources/Fluid/Services/ExternalCoreMLModelRegistry.swift

@@ -28,6 +28,10 @@ enum ExternalCoreMLArtifactsValidationError: LocalizedError {
     case manifestUnreadable(URL, Error)
     case unexpectedModelID(expected: String, actual: String)
     case unexpectedSampleRate(expected: Int, actual: Int)
+    case invalidMaxAudioSeconds(Double)
+    case invalidMaxAudioSamples(Int)
+    case inconsistentAudioWindow(samples: Int, seconds: Double, sampleRate: Int)
+    case invalidOverlapSamples(Int, maxAudioSamples: Int)
 
     var errorDescription: String? {
         switch self {
@@ -41,6 +45,14 @@ enum ExternalCoreMLArtifactsValidationError: LocalizedError {
             return "Unexpected model_id '\(actual)'. Expected '\(expected)'."
         case let .unexpectedSampleRate(expected, actual):
             return "Unexpected sample rate \(actual). Expected \(expected)."
+        case let .invalidMaxAudioSeconds(seconds):
+            return "Invalid max_audio_seconds \(seconds)."
+        case let .invalidMaxAudioSamples(samples):
+            return "Invalid max_audio_samples \(samples)."
+        case let .inconsistentAudioWindow(samples, seconds, sampleRate):
+            return "Manifest audio window is inconsistent: \(samples) samples vs \(seconds)s at \(sampleRate) Hz."
+        case let .invalidOverlapSamples(overlapSamples, maxAudioSamples):
+            return "Invalid overlap_samples \(overlapSamples) for max_audio_samples \(maxAudioSamples)."
         }
     }
 }
@@ -61,6 +73,8 @@ struct ExternalCoreMLASRModelSpec {
     let repositoryOwner: String?
     let repositoryName: String?
     let repositoryRevision: String
+    let artifactBundleVersion: String
+    private let maximumAudioWindowSeconds: Double = 60
 
     var requiredEntries: [String] {
         [
@@ -129,6 +143,33 @@ struct ExternalCoreMLASRModelSpec {
                 actual: manifest.sampleRate
             )
         }
+
+        guard manifest.maxAudioSeconds > 0, manifest.maxAudioSeconds <= self.maximumAudioWindowSeconds else {
+            throw ExternalCoreMLArtifactsValidationError.invalidMaxAudioSeconds(manifest.maxAudioSeconds)
+        }
+
+        let maximumAudioSamples = Int((Double(self.expectedSampleRate) * self.maximumAudioWindowSeconds).rounded())
+        guard manifest.maxAudioSamples > 0, manifest.maxAudioSamples <= maximumAudioSamples else {
+            throw ExternalCoreMLArtifactsValidationError.invalidMaxAudioSamples(manifest.maxAudioSamples)
+        }
+
+        let expectedSamples = Int((manifest.maxAudioSeconds * Double(manifest.sampleRate)).rounded())
+        guard abs(expectedSamples - manifest.maxAudioSamples) <= 1 else {
+            throw ExternalCoreMLArtifactsValidationError.inconsistentAudioWindow(
+                samples: manifest.maxAudioSamples,
+                seconds: manifest.maxAudioSeconds,
+                sampleRate: manifest.sampleRate
+            )
+        }
+
+        if let overlapSamples = manifest.overlapSamples {
+            guard overlapSamples >= 0, overlapSamples < manifest.maxAudioSamples else {
+                throw ExternalCoreMLArtifactsValidationError.invalidOverlapSamples(
+                    overlapSamples,
+                    maxAudioSamples: manifest.maxAudioSamples
+                )
+            }
+        }
     }
 }
 
@@ -151,7 +192,8 @@ enum ExternalCoreMLModelRegistry {
                 sourceURL: URL(string: "https://huggingface.co/BarathwajAnandan/cohere-transcribe-03-2026-CoreML-6bit"),
                 repositoryOwner: "BarathwajAnandan",
                 repositoryName: "cohere-transcribe-03-2026-CoreML-6bit",
-                repositoryRevision: "main"
+                repositoryRevision: "main",
+                artifactBundleVersion: "2026-04-02-cohere-refresh-1"
             )
         default:
             return nil