Better Parakeet error handling

2025-08-03 12:44:13 +05:45 · 2025-08-03 12:44:13 +05:45 · 3eebbc4e3b
commit 3eebbc4e3b
parent 29722d0a31
3 changed files with 93 additions and 110 deletions
--- a/VoiceInk.xcodeproj/project.pbxproj
+++ b/VoiceInk.xcodeproj/project.pbxproj
@ -7,7 +7,7 @@
 	objects = {

 /* Begin PBXBuildFile section */
-		E12E7E972E3F109C006276F2 /* FluidAudio in Frameworks */ = {isa = PBXBuildFile; productRef = E12E7E962E3F109C006276F2 /* FluidAudio */; };
+		E10F06092E3F390600F7FBDC /* FluidAudio in Frameworks */ = {isa = PBXBuildFile; productRef = E10F06082E3F390600F7FBDC /* FluidAudio */; };
 		E1A261122CC143AC00B233D1 /* KeyboardShortcuts in Frameworks */ = {isa = PBXBuildFile; productRef = E1A261112CC143AC00B233D1 /* KeyboardShortcuts */; };
 		E1ADD45A2CC5352A00303ECB /* LaunchAtLogin in Frameworks */ = {isa = PBXBuildFile; productRef = E1ADD4592CC5352A00303ECB /* LaunchAtLogin */; };
 		E1ADD45F2CC544F100303ECB /* Sparkle in Frameworks */ = {isa = PBXBuildFile; productRef = E1ADD45E2CC544F100303ECB /* Sparkle */; };
@ -82,7 +82,7 @@
 			files = (
 				E1ADD45A2CC5352A00303ECB /* LaunchAtLogin in Frameworks */,
 				E1D7EF992E35E16C00640029 /* MediaRemoteAdapter in Frameworks */,
-				E12E7E972E3F109C006276F2 /* FluidAudio in Frameworks */,
+				E10F06092E3F390600F7FBDC /* FluidAudio in Frameworks */,
 				E1B2DCAB2E3DE70A008DFD68 /* whisper.xcframework in Frameworks */,
 				E1ADD45F2CC544F100303ECB /* Sparkle in Frameworks */,
 				E1A261122CC143AC00B233D1 /* KeyboardShortcuts in Frameworks */,
@ -162,7 +162,7 @@
 				E1ADD45E2CC544F100303ECB /* Sparkle */,
 				E1F5FA792DA6CBF900B1FD8A /* Zip */,
 				E1D7EF982E35E16C00640029 /* MediaRemoteAdapter */,
-				E12E7E962E3F109C006276F2 /* FluidAudio */,
+				E10F06082E3F390600F7FBDC /* FluidAudio */,
 			);
 			productName = VoiceInk;
 			productReference = E11473B02CBE0F0A00318EE4 /* VoiceInk.app */;
@ -252,7 +252,7 @@
 				E1ADD45D2CC544F100303ECB /* XCRemoteSwiftPackageReference "Sparkle" */,
 				E1F5FA782DA6CBF900B1FD8A /* XCRemoteSwiftPackageReference "Zip" */,
 				E1D7EF972E35E16C00640029 /* XCRemoteSwiftPackageReference "mediaremote-adapter" */,
-				E10A90B12E3F0E0400E5D6AF /* XCLocalSwiftPackageReference "../FluidAudio" */,
+				E10FFA112E3F37D100F7FBDC /* XCRemoteSwiftPackageReference "FluidAudio" */,
 			);
 			preferredProjectObjectVersion = 77;
 			productRefGroup = E11473B12CBE0F0A00318EE4 /* Products */;
@ -621,14 +621,15 @@
 		};
 /* End XCConfigurationList section */

-/* Begin XCLocalSwiftPackageReference section */
-		E10A90B12E3F0E0400E5D6AF /* XCLocalSwiftPackageReference "../FluidAudio" */ = {
-			isa = XCLocalSwiftPackageReference;
-			relativePath = ../FluidAudio;
-		};
-/* End XCLocalSwiftPackageReference section */
-
 /* Begin XCRemoteSwiftPackageReference section */
+		E10FFA112E3F37D100F7FBDC /* XCRemoteSwiftPackageReference "FluidAudio" */ = {
+			isa = XCRemoteSwiftPackageReference;
+			repositoryURL = "https://github.com/FluidInference/FluidAudio";
+			requirement = {
+				branch = main;
+				kind = branch;
+			};
+		};
 		E1A261102CC143AC00B233D1 /* XCRemoteSwiftPackageReference "KeyboardShortcuts" */ = {
 			isa = XCRemoteSwiftPackageReference;
 			repositoryURL = "https://github.com/sindresorhus/KeyboardShortcuts";
@ -663,7 +664,7 @@
 		};
 		E1F5FA782DA6CBF900B1FD8A /* XCRemoteSwiftPackageReference "Zip" */ = {
 			isa = XCRemoteSwiftPackageReference;
-			repositoryURL = "https://github.com/marmelroy/Zip?tab=readme-ov-file";
+			repositoryURL = "https://github.com/marmelroy/Zip";
 			requirement = {
 				kind = upToNextMajorVersion;
 				minimumVersion = 2.1.2;
@ -672,8 +673,9 @@
 /* End XCRemoteSwiftPackageReference section */

 /* Begin XCSwiftPackageProductDependency section */
-		E12E7E962E3F109C006276F2 /* FluidAudio */ = {
+		E10F06082E3F390600F7FBDC /* FluidAudio */ = {
 			isa = XCSwiftPackageProductDependency;
+			package = E10FFA112E3F37D100F7FBDC /* XCRemoteSwiftPackageReference "FluidAudio" */;
 			productName = FluidAudio;
 		};
 		E1A261112CC143AC00B233D1 /* KeyboardShortcuts */ = {
--- a/VoiceInk.xcodeproj/project.xcworkspace/xcshareddata/swiftpm/Package.resolved
+++ b/VoiceInk.xcodeproj/project.xcworkspace/xcshareddata/swiftpm/Package.resolved
@ -1,6 +1,15 @@
 {
-  "originHash" : "ae3f634e8c4b39a1a80bcd04f018e2208c0491e42ee824cd94a92d7b88893420",
+  "originHash" : "0b9379abd19d2f53581c233273d09235e935a8d2b1180cf253dd69baa2784b39",
  "pins" : [
+    {
+      "identity" : "fluidaudio",
+      "kind" : "remoteSourceControl",
+      "location" : "https://github.com/FluidInference/FluidAudio",
+      "state" : {
+        "branch" : "main",
+        "revision" : "2a3d6a948cb332b3fd8ae479a9942e33ade2cc9e"
+      }
+    },
    {
      "identity" : "keyboardshortcuts",
      "kind" : "remoteSourceControl",
@ -33,14 +42,14 @@
      "kind" : "remoteSourceControl",
      "location" : "https://github.com/sparkle-project/Sparkle",
      "state" : {
-        "revision" : "0ca3004e98712ea2b39dd881d28448630cce1c99",
-        "version" : "2.7.0"
+        "revision" : "df074165274afaa39539c05d57b0832620775b11",
+        "version" : "2.7.1"
      }
    },
    {
-      "identity" : "zip?tab=readme-ov-file",
+      "identity" : "zip",
      "kind" : "remoteSourceControl",
-      "location" : "https://github.com/marmelroy/Zip?tab=readme-ov-file",
+      "location" : "https://github.com/marmelroy/Zip",
      "state" : {
        "revision" : "67fa55813b9e7b3b9acee9c0ae501def28746d76",
        "version" : "2.1.2"
--- a/VoiceInk/Services/ParakeetTranscriptionService.swift
+++ b/VoiceInk/Services/ParakeetTranscriptionService.swift
@ -46,27 +46,22 @@ class ParakeetTranscriptionService: TranscriptionService {
                models = try await AsrModels.downloadAndLoad()
            }
            
-            // Check vocabulary file before initialization
-            let vocabPath = getVocabularyPath()
-            let vocabExists = FileManager.default.fileExists(atPath: vocabPath.path)
-            logger.notice("🦜 Vocabulary file exists at \(vocabPath.lastPathComponent): \(vocabExists)")
-            
-            if vocabExists {
-                do {
-                    let vocabData = try Data(contentsOf: vocabPath)
-                    let vocabDict = try JSONSerialization.jsonObject(with: vocabData) as? [String: String] ?? [:]
-                    logger.notice("🦜 Vocabulary loaded with \(vocabDict.count) entries")
-                } catch {
-                    logger.notice("🦜 Failed to parse vocabulary file: \(error.localizedDescription)")
-                }
-            }
-            
            try await asrManager?.initialize(models: models)
            isModelLoaded = true
            logger.notice("🦜 Parakeet model loaded successfully")
            
+        } catch let error as ASRError {
+            logger.notice("🦜 Parakeet-specific error loading model: \(error.localizedDescription)")
+            isModelLoaded = false
+            asrManager = nil
+            throw error
+        } catch let error as AsrModelsError {
+            logger.notice("🦜 Parakeet model management error loading model: \(error.localizedDescription)")
+            isModelLoaded = false
+            asrManager = nil
+            throw error
        } catch {
-            logger.notice("🦜 Failed to load Parakeet model: \(error.localizedDescription)")
+            logger.notice("🦜 Unexpected error loading Parakeet model: \(error.localizedDescription)")
            isModelLoaded = false
            asrManager = nil
            throw error
@ -74,87 +69,64 @@ class ParakeetTranscriptionService: TranscriptionService {
    }

    func transcribe(audioURL: URL, model: any TranscriptionModel) async throws -> String {
-        do {
-
-            if !isModelLoaded {
-                try await loadModel()
-            }
-            
-            guard let asrManager = asrManager else {
-                logger.notice("🦜 ASR manager is nil after model loading")
-                throw NSError(domain: "ParakeetTranscriptionService", code: -1, userInfo: [NSLocalizedDescriptionKey: "Failed to initialize ASR manager."])
-            }
-
-            logger.notice("🦜 Starting Parakeet transcription")
-            let audioSamples = try readAudioSamples(from: audioURL)
-            logger.notice("🦜 Audio samples loaded: \(audioSamples.count) samples")
-            
-            let result = try await asrManager.transcribe(audioSamples)
-            logger.notice("🦜 Parakeet transcription completed")
-            
-            // Check for empty results (vocabulary issue indicator)
-            if result.text.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty {
-                logger.notice("🦜 Warning: Empty transcription result for \(audioSamples.count) samples - possible vocabulary issue")
-            }
-            
-            if UserDefaults.standard.object(forKey: "IsTextFormattingEnabled") as? Bool ?? true {
-                return WhisperTextFormatter.format(result.text)
-            }
-            return result.text
-        } catch {
-            logger.notice("🦜 Parakeet transcription failed: \(error.localizedDescription)")
-            let errorMessage = error.localizedDescription
-            await MainActor.run {
-                NotificationManager.shared.showNotification(
-                    title: "Transcription Failed: \(errorMessage)",
-                    type: .error
-                )
-            }
-            return ""
+        if asrManager == nil || !isModelLoaded {
+            try await loadModel()
        }
+
+        guard let asrManager = asrManager else {
+            logger.notice("🦜 Parakeet manager is still nil after attempting to load the model.")
+            throw ASRError.notInitialized
+        }
+        
+        let audioSamples = try readAudioSamples(from: audioURL)
+        
+        // Validate audio data before transcription
+        guard audioSamples.count >= 16000 else {
+            logger.notice("🦜 Audio too short for transcription: \(audioSamples.count) samples")
+            throw ASRError.invalidAudioData
+        }
+        
+        let result = try await asrManager.transcribe(audioSamples)
+        
+        Task {
+            asrManager.cleanup()
+            isModelLoaded = false
+            logger.notice("🦜 Parakeet ASR models cleaned up from memory")
+        }
+        
+        // Check for empty results (vocabulary issue indicator)
+        if result.text.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty {
+            logger.notice("🦜 Warning: Empty transcription result for \(audioSamples.count) samples - possible vocabulary issue")
+        }
+        
+        if UserDefaults.standard.object(forKey: "IsTextFormattingEnabled") as? Bool ?? true {
+            return WhisperTextFormatter.format(result.text)
+        }
+        return result.text
    }

    private func readAudioSamples(from url: URL) throws -> [Float] {
-        logger.notice("🦜 Reading audio file: \(url.lastPathComponent)")
-        let data = try Data(contentsOf: url)
-        logger.notice("🦜 Audio file size: \(data.count) bytes")
-        
-        // A basic check, assuming a more robust check happens elsewhere.
-        guard data.count > 44 else { 
-            logger.notice("🦜 Warning: Audio file too small (\(data.count) bytes), expected > 44 bytes")
-            return [] 
-        }
-
-        let floats = stride(from: 44, to: data.count, by: 2).map {
-            return data[$0..<$0 + 2].withUnsafeBytes {
-                let short = Int16(littleEndian: $0.load(as: Int16.self))
-                return max(-1.0, min(Float(short) / 32767.0, 1.0))
+        do {
+            let data = try Data(contentsOf: url)
+            
+            // Check minimum file size for valid WAV header
+            guard data.count > 44 else { 
+                logger.notice("🦜 Audio file too small (\(data.count) bytes), expected > 44 bytes")
+                throw ASRError.invalidAudioData
            }
-        }
-        
-        logger.notice("🦜 Processed audio: \(floats.count) samples from \(data.count) bytes")
-        
-        // Check if we have enough samples for transcription (minimum 16,000 samples = 1 second at 16kHz)
-        if floats.count < 16000 {
-            logger.notice("🦜 Warning: Audio too short (\(floats.count) samples), minimum 16,000 required")
-        }
-        
-        return floats
-    }
-    
-    // Helper function to get vocabulary path based on model directory
-    private func getVocabularyPath() -> URL {
-        if let customDirectory = customModelsDirectory {
-            return customDirectory.appendingPathComponent("parakeet_vocab.json")
-        } else {
-            let applicationSupportURL = FileManager.default.urls(
-                for: .applicationSupportDirectory, in: .userDomainMask
-            ).first!
-            return applicationSupportURL
-                .appendingPathComponent("FluidAudio", isDirectory: true)
-                .appendingPathComponent("Models", isDirectory: true)
-                .appendingPathComponent("parakeet-tdt-0.6b-v2-coreml", isDirectory: true)
-                .appendingPathComponent("parakeet_vocab.json")
+
+            let floats = stride(from: 44, to: data.count, by: 2).map {
+                return data[$0..<$0 + 2].withUnsafeBytes {
+                    let short = Int16(littleEndian: $0.load(as: Int16.self))
+                    return max(-1.0, min(Float(short) / 32767.0, 1.0))
+                }
+            }
+            
+            return floats
+        } catch {
+            logger.notice("🦜 Failed to read audio file: \(error.localizedDescription)")
+            throw ASRError.invalidAudioData
        }
    }
+
 }