fix: preserve recordings before transcription

.github/workflows/release.yml

@@ -20,6 +20,9 @@ jobs:
     runs-on: macos-15
     env:
       VERSION: ${{ inputs.version }}
+      APPLE_ID: ${{ secrets.APPLE_ID }}
+      APPLE_TEAM_ID: ${{ secrets.APPLE_TEAM_ID }}
+      APPLE_APP_PASSWORD: ${{ secrets.APPLE_APP_PASSWORD }}
       SPARKLE_PRIVATE_KEY: ${{ secrets.SPARKLE_PRIVATE_KEY }}
     steps:
       - uses: actions/checkout@v7
@@ -31,7 +34,7 @@ jobs:
           APPLE_CERTIFICATE_P12: ${{ secrets.APPLE_CERTIFICATE_P12 }}
           APPLE_CERTIFICATE_PASSWORD: ${{ secrets.APPLE_CERTIFICATE_PASSWORD }}
         run: |
-          for variable in APPLE_CERTIFICATE_P12 APPLE_CERTIFICATE_PASSWORD SPARKLE_PRIVATE_KEY; do
+          for variable in APPLE_CERTIFICATE_P12 APPLE_CERTIFICATE_PASSWORD APPLE_ID APPLE_TEAM_ID APPLE_APP_PASSWORD SPARKLE_PRIVATE_KEY; do
             if [[ -z "${!variable:-}" ]]; then
               echo "Missing GitHub Actions secret: $variable" >&2
               exit 1
@@ -53,7 +56,7 @@ jobs:
           fi
           echo "SIGNING_IDENTITY=$signing_identity" >> "$GITHUB_ENV"
 
-      - name: Build and sign release
+      - name: Build, sign, and notarize release
         timeout-minutes: 30
         run: scripts/package_release.sh
 
@@ -70,13 +73,16 @@ jobs:
           GH_TOKEN: ${{ github.token }}
         run: |
           tag="v$VERSION"
-          if git rev-parse "$tag" >/dev/null 2>&1; then
+          if gh release view "$tag" >/dev/null 2>&1; then
-            echo "Tag already exists: $tag" >&2
+            gh release upload "$tag" \
-            exit 1
+              "$RUNNER_TEMP/meetingnotes-release/release/Meetingnotes-$VERSION.zip" \
-          fi
+              "$RUNNER_TEMP/meetingnotes-release/release/appcast.xml" \
+              --clobber
+          else
             gh release create "$tag" \
               "$RUNNER_TEMP/meetingnotes-release/release/Meetingnotes-$VERSION.zip" \
               "$RUNNER_TEMP/meetingnotes-release/release/appcast.xml" \
               --target "$GITHUB_SHA" \
               --title "Meetingnotes $VERSION" \
               --generate-notes
+          fi

Meetingnotes.xcodeproj/project.pbxproj

@@ -276,7 +276,7 @@
 				CODE_SIGN_IDENTITY = "Apple Development";
 				CODE_SIGN_STYLE = Automatic;
 				COMBINE_HIDPI_IMAGES = YES;
-				CURRENT_PROJECT_VERSION = 31;
+				CURRENT_PROJECT_VERSION = 35;
 				DEVELOPMENT_ASSET_PATHS = "\"meetingnotes/Preview Content\"";
 				DEVELOPMENT_TEAM = G9LVHZAJNX;
 				ENABLE_HARDENED_RUNTIME = YES;
@@ -290,7 +290,7 @@
 					"@executable_path/../Frameworks",
 				);
 				MACOSX_DEPLOYMENT_TARGET = 15.0;
-				MARKETING_VERSION = 1.1.19;
+				MARKETING_VERSION = 1.1.23;
 				ONLY_ACTIVE_ARCH = NO;
 				OTHER_SWIFT_FLAGS = "$(inherited) -D ENABLE_TCC_SPI";
 				PRODUCT_BUNDLE_IDENTIFIER = net.jamesbone.meetingnotes;
@@ -312,7 +312,7 @@
 				CODE_SIGN_IDENTITY = "Apple Development";
 				CODE_SIGN_STYLE = Automatic;
 				COMBINE_HIDPI_IMAGES = YES;
-				CURRENT_PROJECT_VERSION = 31;
+				CURRENT_PROJECT_VERSION = 35;
 				DEVELOPMENT_ASSET_PATHS = "\"meetingnotes/Preview Content\"";
 				DEVELOPMENT_TEAM = G9LVHZAJNX;
 				ENABLE_HARDENED_RUNTIME = YES;
@@ -326,7 +326,7 @@
 					"@executable_path/../Frameworks",
 				);
 				MACOSX_DEPLOYMENT_TARGET = 15.0;
-				MARKETING_VERSION = 1.1.19;
+				MARKETING_VERSION = 1.1.23;
 				ONLY_ACTIVE_ARCH = YES;
 				OTHER_SWIFT_FLAGS = "$(inherited) -D ENABLE_TCC_SPI";
 				PRODUCT_BUNDLE_IDENTIFIER = net.jamesbone.meetingnotes;

meetingnotes/Managers/AudioManager.swift

@@ -86,9 +86,22 @@ final class AudioManager: NSObject, ObservableObject {
             isProcessing = false
         }
 
+        repairHalfDurationSystemWAVIfNeeded(in: files)
+        let completedFiles = files.compactMap { $0 }
+        let audioFolder = preserveAudioFiles(completedFiles, meetingID: completedMeetingID)
+        let transcriptionFiles = preservedAudioFiles(files, in: audioFolder)
+        lastRecoveryAudioFolderName = audioFolder?.lastPathComponent
+        if let mismatch = captureDurationMismatch(in: transcriptionFiles) {
+            let recoveryMessage = audioFolder == nil
+                ? " The audio remains in the app's temporary folder."
+                : " Audio was kept so it can be recovered."
+            errorMessage = "System audio timing was invalid (\(mismatch)). Transcription was stopped to avoid an out-of-order result." + recoveryMessage
+            return transcriptChunks.filter(\.isFinal)
+        }
+
         let model = UserDefaultsManager.shared.transcriptionModel
-        async let micResult = transcribe(files[0], model: model, diarization: false)
+        async let micResult = transcribe(transcriptionFiles[0], model: model, diarization: false)
-        async let systemResult = transcribe(files[1], model: model, diarization: true)
+        async let systemResult = transcribe(transcriptionFiles[1], model: model, diarization: true)
         let (micTranscription, systemTranscription) = await (micResult, systemResult)
         let results = [micTranscription, systemTranscription]
 
@@ -98,9 +111,6 @@ final class AudioManager: NSObject, ObservableObject {
             existingChunks: transcriptChunks.filter(\.isFinal)
         )
         transcriptChunks = updated
-        let completedFiles = files.compactMap { $0 }
-        let audioFolder = preserveAudioFiles(completedFiles, meetingID: completedMeetingID)
-        lastRecoveryAudioFolderName = audioFolder?.lastPathComponent
         if !failures.isEmpty {
             let retentionDays = UserDefaultsManager.shared.audioRetentionDays
             let retentionUnit = retentionDays == 1 ? "day" : "days"
@@ -209,15 +219,18 @@ final class AudioManager: NSObject, ObservableObject {
     
     private func prepareAudioFiles() throws {
         let settings: [String: Any] = [
-            AVFormatIDKey: kAudioFormatMPEG4AAC,
+            AVFormatIDKey: kAudioFormatLinearPCM,
             AVSampleRateKey: 16_000,
             AVNumberOfChannelsKey: 1,
-            AVEncoderBitRateKey: 48_000
+            AVLinearPCMBitDepthKey: 16,
+            AVLinearPCMIsFloatKey: false,
+            AVLinearPCMIsBigEndianKey: false,
+            AVLinearPCMIsNonInterleaved: false
         ]
         let base = FileManager.default.temporaryDirectory
         let id = sessionID.uuidString
-        let micURL = base.appendingPathComponent("meetingnotes-\(id)-mic.m4a")
+        let micURL = base.appendingPathComponent("meetingnotes-\(id)-mic.wav")
-        let systemURL = base.appendingPathComponent("meetingnotes-\(id)-system.m4a")
+        let systemURL = base.appendingPathComponent("meetingnotes-\(id)-system.wav")
         let newMicAudioFile = try AVAudioFile(
             forWriting: micURL,
             settings: settings,
@@ -353,13 +366,25 @@ final class AudioManager: NSObject, ObservableObject {
     
     private func startTapIO(_ tap: ProcessTap) throws {
         guard var description = tap.tapStreamDescription,
-              let inputFormat = AVAudioFormat(streamDescription: &description),
+              let advertisedInputFormat = AVAudioFormat(streamDescription: &description),
               let targetFormat = systemAudioFile?.processingFormat,
-              let converter = AVAudioConverter(from: inputFormat, to: targetFormat) else {
+              advertisedInputFormat.sampleRate > 0 else {
             throw NSError(domain: "AudioManager", code: -1, userInfo: [NSLocalizedDescriptionKey: "Unsupported system audio format"])
         }
+        var inputFormat: AVAudioFormat?
+        var converter: AVAudioConverter?
         try tap.run(on: tapQueue) { [weak self] _, inputData, _, _, _ in
             guard let self else { return }
+            if inputFormat == nil {
+                inputFormat = self.inputFormat(
+                    for: inputData,
+                    advertisedFormat: advertisedInputFormat
+                )
+                if let inputFormat {
+                    converter = AVAudioConverter(from: inputFormat, to: targetFormat)
+                }
+            }
+            guard let inputFormat, let converter else { return }
             // The tap queue is serial. Reusing the converter preserves its
             // resampler state instead of discarding audio at every callback.
             self.processAudioBuffer(
@@ -374,6 +399,30 @@ final class AudioManager: NSObject, ObservableObject {
         }
     }
 
+    private func inputFormat(
+        for inputData: UnsafePointer<AudioBufferList>,
+        advertisedFormat: AVAudioFormat
+    ) -> AVAudioFormat? {
+        let buffers = UnsafeMutableAudioBufferListPointer(
+            UnsafeMutablePointer(mutating: inputData)
+        )
+        let channelCount = buffers.reduce(UInt32(0)) { $0 + $1.mNumberChannels }
+        guard channelCount > 0 else { return nil }
+
+        // HAL tap metadata can advertise interleaved stereo while the callback
+        // supplies one mono buffer per channel (or the reverse). Constructing a
+        // PCM buffer with that mismatched layout halves its frame count and
+        // produces 2x-speed system audio. The callback's AudioBufferList is the
+        // authoritative layout for the memory we are copying.
+        let isInterleaved = buffers.count == 1 && channelCount > 1
+        return AVAudioFormat(
+            commonFormat: advertisedFormat.commonFormat,
+            sampleRate: advertisedFormat.sampleRate,
+            channels: AVAudioChannelCount(channelCount),
+            interleaved: isInterleaved
+        )
+    }
+
     private func copyAudioBuffer(
         from inputData: UnsafePointer<AudioBufferList>,
         format: AVAudioFormat
@@ -521,6 +570,99 @@ final class AudioManager: NSObject, ObservableObject {
         LocalStorageManager.shared.preserveAudioFiles(urls, for: meetingID)
     }
 
+    private func preservedAudioFiles(_ urls: [URL?], in folder: URL?) -> [URL?] {
+        urls.map { sourceURL in
+            guard let sourceURL else { return nil }
+            guard let folder else { return sourceURL }
+            let preservedURL = folder.appendingPathComponent(sourceURL.lastPathComponent)
+            return FileManager.default.fileExists(atPath: preservedURL.path) ? preservedURL : sourceURL
+        }
+    }
+
+    private func captureDurationMismatch(in files: [URL?]) -> String? {
+        guard files.count >= 2,
+              let micDuration = audioDuration(at: files[0]),
+              let systemDuration = audioDuration(at: files[1]),
+              micDuration >= 60 else { return nil }
+        let ratio = systemDuration / micDuration
+        guard (0.45...0.55).contains(ratio) || (1.8...2.2).contains(ratio) else { return nil }
+        return String(format: "mic %.1fs, system %.1fs", micDuration, systemDuration)
+    }
+
+    private func repairHalfDurationSystemWAVIfNeeded(in files: [URL?]) {
+        guard files.count >= 2,
+              let micDuration = audioDuration(at: files[0]),
+              let systemURL = files[1],
+              let systemDuration = audioDuration(at: systemURL),
+              micDuration >= 60,
+              systemURL.pathExtension.caseInsensitiveCompare("wav") == .orderedSame,
+              (0.48...0.52).contains(systemDuration / micDuration) else { return }
+        try? halveWAVSampleRate(at: systemURL)
+    }
+
+    private func halveWAVSampleRate(at url: URL) throws {
+        let handle = try FileHandle(forUpdating: url)
+        defer { try? handle.close() }
+
+        try handle.seek(toOffset: 0)
+        guard let riffHeader = try handle.read(upToCount: 12),
+              riffHeader.count == 12,
+              String(data: riffHeader[0..<4], encoding: .ascii) == "RIFF",
+              String(data: riffHeader[8..<12], encoding: .ascii) == "WAVE" else {
+            throw NSError(domain: "AudioManager", code: -2, userInfo: [NSLocalizedDescriptionKey: "Invalid WAV header"])
+        }
+
+        var offset: UInt64 = 12
+        while true {
+            try handle.seek(toOffset: offset)
+            guard let chunkHeader = try handle.read(upToCount: 8), chunkHeader.count == 8 else { break }
+            let chunkID = String(data: chunkHeader[0..<4], encoding: .ascii)
+            let chunkSize = UInt32(chunkHeader[4])
+                | (UInt32(chunkHeader[5]) << 8)
+                | (UInt32(chunkHeader[6]) << 16)
+                | (UInt32(chunkHeader[7]) << 24)
+            let chunkDataOffset = offset + 8
+
+            if chunkID == "fmt ", chunkSize >= 16 {
+                try handle.seek(toOffset: chunkDataOffset)
+                guard let format = try handle.read(upToCount: 16), format.count == 16 else { break }
+                let audioFormat = UInt16(format[0]) | (UInt16(format[1]) << 8)
+                let blockAlign = UInt16(format[12]) | (UInt16(format[13]) << 8)
+                let sampleRate = UInt32(format[4])
+                    | (UInt32(format[5]) << 8)
+                    | (UInt32(format[6]) << 16)
+                    | (UInt32(format[7]) << 24)
+                guard audioFormat == 1, sampleRate >= 16_000, sampleRate.isMultiple(of: 2) else {
+                    throw NSError(domain: "AudioManager", code: -3, userInfo: [NSLocalizedDescriptionKey: "Unsupported WAV format"])
+                }
+
+                let correctedSampleRate = sampleRate / 2
+                let correctedByteRate = correctedSampleRate * UInt32(blockAlign)
+                try handle.seek(toOffset: chunkDataOffset + 4)
+                try handle.write(contentsOf: littleEndianData(correctedSampleRate))
+                try handle.write(contentsOf: littleEndianData(correctedByteRate))
+                try handle.synchronize()
+                return
+            }
+
+            offset = chunkDataOffset + UInt64(chunkSize) + UInt64(chunkSize % 2)
+        }
+
+        throw NSError(domain: "AudioManager", code: -4, userInfo: [NSLocalizedDescriptionKey: "WAV format chunk was not found"])
+    }
+
+    private func littleEndianData(_ value: UInt32) -> Data {
+        var littleEndianValue = value.littleEndian
+        return withUnsafeBytes(of: &littleEndianValue) { Data($0) }
+    }
+
+    private func audioDuration(at url: URL?) -> TimeInterval? {
+        guard let url,
+              let file = try? AVAudioFile(forReading: url),
+              file.processingFormat.sampleRate > 0 else { return nil }
+        return Double(file.length) / file.processingFormat.sampleRate
+    }
+        
     private func resetAudioLevels() {
         micAudioLevel = 0
         systemAudioLevel = 0

meetingnotes/Managers/LocalStorageManager.swift

@@ -256,50 +256,12 @@ class LocalStorageManager {
     }
 
     func findRecoveryAudioFolder(for meeting: Meeting) -> URL? {
-        if let name = meeting.recoveryAudioFolderName,
+        let canonicalName = meeting.id.uuidString
-           let folder = recoveryAudioFolder(named: name) {
+        guard meeting.recoveryAudioFolderName == nil
-            return folder
+                || meeting.recoveryAudioFolderName?.caseInsensitiveCompare(canonicalName) == .orderedSame else {
-        }
-
-        let claimedFolderNames = Set(
-            loadMeetings()
-                .filter { $0.id != meeting.id }
-                .compactMap(\.recoveryAudioFolderName)
-        )
-        guard let folders = try? FileManager.default.contentsOfDirectory(
-            at: recoveryDirectory,
-            includingPropertiesForKeys: [.isDirectoryKey, .creationDateKey, .contentModificationDateKey],
-            options: [.skipsHiddenFiles]
-        ) else {
             return nil
         }
-
+        return recoveryAudioFolder(named: canonicalName)
-        let candidates = folders.compactMap { folder -> (url: URL, distance: TimeInterval)? in
-            let folderValues = try? folder.resourceValues(
-                forKeys: [.isDirectoryKey, .creationDateKey, .contentModificationDateKey]
-            )
-            let files = recoveryAudioFiles(in: folder)
-            guard folderValues?.isDirectory == true,
-                  !claimedFolderNames.contains(folder.lastPathComponent),
-                  !files.isEmpty else {
-                return nil
-            }
-            let dates = files.compactMap { file -> Date? in
-                let values = try? file.url.resourceValues(forKeys: [.creationDateKey, .contentModificationDateKey])
-                return values?.creationDate ?? values?.contentModificationDate
-            }
-            let referenceDate = dates.min()
-                ?? folderValues?.creationDate
-                ?? folderValues?.contentModificationDate
-            guard let referenceDate else { return nil }
-            return (folder, abs(referenceDate.timeIntervalSince(meeting.date)))
-        }
-
-        // This fallback links recovery files created by older app versions.
-        return candidates
-            .filter { $0.distance <= 12 * 60 * 60 }
-            .min(by: { $0.distance < $1.distance })?
-            .url
     }
 
     func deleteRecoveryAudioFolder(_ folder: URL) {

meetingnotes/Managers/RecordingSessionManager.swift

@@ -124,6 +124,9 @@ class RecordingSessionManager: ObservableObject {
         // Find and update the active meeting
         if let index = meetings.firstIndex(where: { $0.id == meetingId }) {
             meetings[index].transcriptChunks = chunks
+            if let recoveryAudioFolderName = lastRecoveryAudioFolderName {
+                meetings[index].recoveryAudioFolderName = recoveryAudioFolderName
+            }
             
             // Save the updated meeting
             let success = LocalStorageManager.shared.saveMeeting(meetings[index])

meetingnotes/Managers/UserDefaultsManager.swift

@@ -79,13 +79,7 @@ class UserDefaultsManager {
     }
 
     var transcriptionModel: String {
-        get {
+        get { userDefaults.string(forKey: Keys.transcriptionModel) ?? "local-parakeet/parakeet-tdt-0.6b-v3" }
-            let stored = userDefaults.string(forKey: Keys.transcriptionModel)
-            if stored == "local-whisper/whisper-large-v3-turbo" {
-                return "local-parakeet/parakeet-tdt-0.6b-v3"
-            }
-            return stored ?? "local-parakeet/parakeet-tdt-0.6b-v3"
-        }
         set { userDefaults.set(newValue, forKey: Keys.transcriptionModel) }
     }
 

meetingnotes/Services/CoderAPIClient.swift

@@ -275,6 +275,13 @@ final class CoderAPIClient {
         let framesPerChunk = preserveSpeakerIdentity
             ? max(1, input.length)
             : AVAudioFramePosition(format.sampleRate * transcriptionChunkDuration)
+
+        if fileURL.pathExtension.caseInsensitiveCompare("wav") == .orderedSame,
+           input.fileFormat.streamDescription.pointee.mFormatID == kAudioFormatLinearPCM,
+           input.length <= framesPerChunk {
+            return [AudioChunk(url: fileURL, offset: 0, isTemporary: false)]
+        }
+
         var chunks: [AudioChunk] = []
         var frameOffset: AVAudioFramePosition = 0
 

meetingnotes/ViewModels/MeetingViewModel.swift

@@ -254,9 +254,7 @@ class MeetingViewModel: ObservableObject {
 
     private func refreshRecoveryAudioFolder() {
         recoveryAudioFolderURL = LocalStorageManager.shared.findRecoveryAudioFolder(for: meeting)
-        if let recoveryAudioFolderURL {
+        meeting.recoveryAudioFolderName = recoveryAudioFolderURL?.lastPathComponent
-            meeting.recoveryAudioFolderName = recoveryAudioFolderURL.lastPathComponent
-        }
     }
 
     func showAudioInFinder() {

meetingnotes/Views/SettingsView.swift

@@ -59,10 +59,16 @@ struct SettingsView: View {
                             }
                         }
 
-                        if viewModel.coderModels.first(where: { $0.id == viewModel.settings.transcriptionModel })?.supportsSpeakerDiarization == true {
+                        if let selectedModel = viewModel.coderModels.first(where: { $0.id == viewModel.settings.transcriptionModel }) {
+                            if selectedModel.supportsSpeakerDiarization {
                                 Label("Remote participants are labeled Speaker 1–4; your microphone is labeled Me.", systemImage: "person.2.wave.2")
                                     .font(.caption)
                                     .foregroundColor(.secondary)
+                            } else {
+                                Label("Remote participants are labeled Them; your microphone is labeled Me.", systemImage: "person.2")
+                                    .font(.caption)
+                                    .foregroundColor(.secondary)
+                            }
                         }
                     }
 

scripts/package_release.sh

@@ -14,6 +14,9 @@ APP_PATH="$DERIVED_DATA/Build/Products/Release/$APP_NAME.app"
 required_variables=(
   VERSION
   SIGNING_IDENTITY
+  APPLE_ID
+  APPLE_TEAM_ID
+  APPLE_APP_PASSWORD
   SPARKLE_PRIVATE_KEY
   GITHUB_REPOSITORY
 )
@@ -80,6 +83,33 @@ ARCHIVE_NAME="$APP_NAME-$VERSION.zip"
 ARCHIVE_PATH="$RELEASE_DIR/$ARCHIVE_NAME"
 ditto -c -k --sequesterRsrc --keepParent "$APP_PATH" "$ARCHIVE_PATH"
 
+NOTARY_RESULT="$BUILD_ROOT/notary-result.json"
+xcrun notarytool submit "$ARCHIVE_PATH" \
+  --apple-id "$APPLE_ID" \
+  --team-id "$APPLE_TEAM_ID" \
+  --password "$APPLE_APP_PASSWORD" \
+  --wait \
+  --timeout 20m \
+  --output-format json > "$NOTARY_RESULT"
+
+NOTARY_STATUS=$(plutil -extract status raw -o - "$NOTARY_RESULT")
+if [[ "$NOTARY_STATUS" != "Accepted" ]]; then
+  submission_id=$(plutil -extract id raw -o - "$NOTARY_RESULT")
+  xcrun notarytool log "$submission_id" \
+    --apple-id "$APPLE_ID" \
+    --team-id "$APPLE_TEAM_ID" \
+    --password "$APPLE_APP_PASSWORD" || true
+  echo "Apple notarization failed with status: $NOTARY_STATUS" >&2
+  exit 1
+fi
+
+xcrun stapler staple "$APP_PATH"
+xcrun stapler validate "$APP_PATH"
+spctl --assess --type execute --verbose=2 "$APP_PATH"
+
+rm -f "$ARCHIVE_PATH"
+ditto -c -k --sequesterRsrc --keepParent "$APP_PATH" "$ARCHIVE_PATH"
+
 GENERATE_APPCAST=$(find "$DERIVED_DATA/SourcePackages/artifacts" -type f -name generate_appcast -print -quit)
 if [[ -z "$GENERATE_APPCAST" ]]; then
   echo "Sparkle generate_appcast tool was not found" >&2
@@ -97,8 +127,8 @@ grep -q "$DOWNLOAD_URL$ARCHIVE_NAME" "$RELEASE_DIR/appcast.xml"
 grep -q 'sparkle:edSignature=' "$RELEASE_DIR/appcast.xml"
 
 if [[ -n "${GITHUB_STEP_SUMMARY:-}" ]]; then
-  printf 'Built and Developer ID-signed Meetingnotes %s. The GitHub release is ready to publish.\n' \
+  printf 'Built, Developer ID-signed, notarized, and stapled Meetingnotes %s. The GitHub release is ready to publish.\n' \
     "$VERSION" >> "$GITHUB_STEP_SUMMARY"
 fi
 
-echo "Signed release artifacts are ready in $RELEASE_DIR"
+echo "Signed and notarized release artifacts are ready in $RELEASE_DIR"