Compare commits

...
3 Commits
7 changed files with 360 additions and 57 deletions
+4 -4
View File
@@ -276,7 +276,7 @@
CODE_SIGN_IDENTITY = "Apple Development"; CODE_SIGN_IDENTITY = "Apple Development";
CODE_SIGN_STYLE = Automatic; CODE_SIGN_STYLE = Automatic;
COMBINE_HIDPI_IMAGES = YES; COMBINE_HIDPI_IMAGES = YES;
CURRENT_PROJECT_VERSION = 38; CURRENT_PROJECT_VERSION = 41;
DEVELOPMENT_ASSET_PATHS = "\"meetingnotes/Preview Content\""; DEVELOPMENT_ASSET_PATHS = "\"meetingnotes/Preview Content\"";
DEVELOPMENT_TEAM = G9LVHZAJNX; DEVELOPMENT_TEAM = G9LVHZAJNX;
ENABLE_HARDENED_RUNTIME = YES; ENABLE_HARDENED_RUNTIME = YES;
@@ -290,7 +290,7 @@
"@executable_path/../Frameworks", "@executable_path/../Frameworks",
); );
MACOSX_DEPLOYMENT_TARGET = 15.0; MACOSX_DEPLOYMENT_TARGET = 15.0;
MARKETING_VERSION = 1.1.26; MARKETING_VERSION = 1.1.29;
ONLY_ACTIVE_ARCH = NO; ONLY_ACTIVE_ARCH = NO;
OTHER_SWIFT_FLAGS = "$(inherited) -D ENABLE_TCC_SPI"; OTHER_SWIFT_FLAGS = "$(inherited) -D ENABLE_TCC_SPI";
PRODUCT_BUNDLE_IDENTIFIER = net.jamesbone.meetingnotes; PRODUCT_BUNDLE_IDENTIFIER = net.jamesbone.meetingnotes;
@@ -312,7 +312,7 @@
CODE_SIGN_IDENTITY = "Apple Development"; CODE_SIGN_IDENTITY = "Apple Development";
CODE_SIGN_STYLE = Automatic; CODE_SIGN_STYLE = Automatic;
COMBINE_HIDPI_IMAGES = YES; COMBINE_HIDPI_IMAGES = YES;
CURRENT_PROJECT_VERSION = 38; CURRENT_PROJECT_VERSION = 41;
DEVELOPMENT_ASSET_PATHS = "\"meetingnotes/Preview Content\""; DEVELOPMENT_ASSET_PATHS = "\"meetingnotes/Preview Content\"";
DEVELOPMENT_TEAM = G9LVHZAJNX; DEVELOPMENT_TEAM = G9LVHZAJNX;
ENABLE_HARDENED_RUNTIME = YES; ENABLE_HARDENED_RUNTIME = YES;
@@ -326,7 +326,7 @@
"@executable_path/../Frameworks", "@executable_path/../Frameworks",
); );
MACOSX_DEPLOYMENT_TARGET = 15.0; MACOSX_DEPLOYMENT_TARGET = 15.0;
MARKETING_VERSION = 1.1.26; MARKETING_VERSION = 1.1.29;
ONLY_ACTIVE_ARCH = YES; ONLY_ACTIVE_ARCH = YES;
OTHER_SWIFT_FLAGS = "$(inherited) -D ENABLE_TCC_SPI"; OTHER_SWIFT_FLAGS = "$(inherited) -D ENABLE_TCC_SPI";
PRODUCT_BUNDLE_IDENTIFIER = net.jamesbone.meetingnotes; PRODUCT_BUNDLE_IDENTIFIER = net.jamesbone.meetingnotes;
+221 -29
View File
@@ -20,6 +20,26 @@ private enum RecoveryTranscriptionError: LocalizedError {
} }
} }
private struct SystemCaptureDiagnostics {
var tapAdvertisedFormat = "unavailable"
var aggregateInputFormat = "unavailable"
var selectedInputFormat = "unavailable"
var targetFormat = "unavailable"
var selectedInputSampleRate: Double?
var targetSampleRate: Double?
var firstBufferLayout: String?
var callbackCount: UInt64 = 0
var inputFrameCount: UInt64 = 0
var outputFrameCount: UInt64 = 0
var discardedCallbackCount: UInt64 = 0
var firstSampleTime: Double?
var lastSampleTime: Double?
var firstHostTime: UInt64?
var lastHostTime: UInt64?
var firstCallbackAt: Date?
var lastCallbackAt: Date?
}
/// Captures microphone and system audio locally, then sends completed files to Coder. /// Captures microphone and system audio locally, then sends completed files to Coder.
@MainActor @MainActor
final class AudioManager: NSObject, ObservableObject { final class AudioManager: NSObject, ObservableObject {
@@ -50,6 +70,7 @@ final class AudioManager: NSObject, ObservableObject {
private var micAudioURL: URL? private var micAudioURL: URL?
private var systemAudioURL: URL? private var systemAudioURL: URL?
private var recordingStartedAt = Date() private var recordingStartedAt = Date()
private var systemDiagnostics = SystemCaptureDiagnostics()
private override init() { private override init() {
super.init() super.init()
@@ -67,6 +88,7 @@ final class AudioManager: NSObject, ObservableObject {
sessionID = UUID() sessionID = UUID()
self.meetingID = meetingID self.meetingID = meetingID
recordingStartedAt = Date() recordingStartedAt = Date()
systemDiagnostics = SystemCaptureDiagnostics()
do { do {
try prepareAudioFiles() try prepareAudioFiles()
startMicrophoneTap() startMicrophoneTap()
@@ -79,6 +101,7 @@ final class AudioManager: NSObject, ObservableObject {
func stopRecordingAndTranscribe() async -> [TranscriptChunk] { func stopRecordingAndTranscribe() async -> [TranscriptChunk] {
let completedMeetingID = meetingID let completedMeetingID = meetingID
let completedSessionID = sessionID
let captureStartedAt = recordingStartedAt let captureStartedAt = recordingStartedAt
let files = stopCaptureAndCloseFiles() let files = stopCaptureAndCloseFiles()
isProcessing = true isProcessing = true
@@ -86,11 +109,20 @@ final class AudioManager: NSObject, ObservableObject {
isProcessing = false isProcessing = false
} }
repairHalfDurationSystemWAVIfNeeded(in: files) let preRepairFileSummaries = files.map(audioFileSummary)
let repairApplied = repairHalfDurationSystemWAVIfNeeded(in: files)
let completedFiles = files.compactMap { $0 } let completedFiles = files.compactMap { $0 }
let audioFolder = preserveAudioFiles(completedFiles, meetingID: completedMeetingID) let audioFolder = preserveAudioFiles(completedFiles, meetingID: completedMeetingID)
let transcriptionFiles = preservedAudioFiles(files, in: audioFolder) let transcriptionFiles = preservedAudioFiles(files, in: audioFolder)
lastRecoveryAudioFolderName = audioFolder?.lastPathComponent lastRecoveryAudioFolderName = audioFolder?.lastPathComponent
writeCaptureDiagnostics(
sessionID: completedSessionID,
captureStartedAt: captureStartedAt,
preRepairFileSummaries: preRepairFileSummaries,
repairedFiles: transcriptionFiles,
repairApplied: repairApplied,
audioFolder: audioFolder
)
if let mismatch = captureDurationMismatch(in: transcriptionFiles) { if let mismatch = captureDurationMismatch(in: transcriptionFiles) {
let recoveryMessage = audioFolder == nil let recoveryMessage = audioFolder == nil
? " The audio remains in the app's temporary folder." ? " The audio remains in the app's temporary folder."
@@ -405,9 +437,13 @@ final class AudioManager: NSObject, ObservableObject {
advertisedInputFormat.sampleRate > 0 else { advertisedInputFormat.sampleRate > 0 else {
throw NSError(domain: "AudioManager", code: -1, userInfo: [NSLocalizedDescriptionKey: "Unsupported system audio format"]) throw NSError(domain: "AudioManager", code: -1, userInfo: [NSLocalizedDescriptionKey: "Unsupported system audio format"])
} }
systemDiagnostics.tapAdvertisedFormat = streamDescriptionSummary(tap.tapAdvertisedStreamDescription)
systemDiagnostics.aggregateInputFormat = streamDescriptionSummary(tap.aggregateInputStreamDescription)
systemDiagnostics.targetFormat = audioFormatSummary(targetFormat)
systemDiagnostics.targetSampleRate = targetFormat.sampleRate
var inputFormat: AVAudioFormat? var inputFormat: AVAudioFormat?
var converter: AVAudioConverter? var converter: AVAudioConverter?
try tap.run(on: tapQueue) { [weak self] _, inputData, _, _, _ in try tap.run(on: tapQueue) { [weak self] inNow, inputData, inInputTime, _, _ in
guard let self else { return } guard let self else { return }
if inputFormat == nil { if inputFormat == nil {
inputFormat = self.inputFormat( inputFormat = self.inputFormat(
@@ -415,6 +451,8 @@ final class AudioManager: NSObject, ObservableObject {
advertisedFormat: advertisedInputFormat advertisedFormat: advertisedInputFormat
) )
if let inputFormat { if let inputFormat {
self.systemDiagnostics.selectedInputFormat = self.audioFormatSummary(inputFormat)
self.systemDiagnostics.selectedInputSampleRate = inputFormat.sampleRate
converter = AVAudioConverter(from: inputFormat, to: targetFormat) converter = AVAudioConverter(from: inputFormat, to: targetFormat)
} }
} }
@@ -425,7 +463,9 @@ final class AudioManager: NSObject, ObservableObject {
{ self.copyAudioBuffer(from: inputData, format: inputFormat) }, { self.copyAudioBuffer(from: inputData, format: inputFormat) },
converter: converter, converter: converter,
targetFormat: targetFormat, targetFormat: targetFormat,
source: .system source: .system,
callbackTimestamp: inInputTime.pointee,
ioTimestamp: inNow.pointee
) )
} invalidationHandler: { [weak self] _ in } invalidationHandler: { [weak self] _ in
guard let self, self.isRecording else { return } guard let self, self.isRecording else { return }
@@ -443,14 +483,15 @@ final class AudioManager: NSObject, ObservableObject {
let channelCount = buffers.reduce(UInt32(0)) { $0 + $1.mNumberChannels } let channelCount = buffers.reduce(UInt32(0)) { $0 + $1.mNumberChannels }
guard channelCount > 0 else { return nil } guard channelCount > 0 else { return nil }
// HAL tap metadata can advertise interleaved stereo while the callback // HAL I/O proc samples use the canonical Float32 representation. The
// supplies one mono buffer per channel (or the reverse). Constructing a // tap's stream description can advertise a different common format;
// PCM buffer with that mismatched layout halves its frame count and // using that to interpret the callback bytes can halve the frame count
// produces 2x-speed system audio. The callback's AudioBufferList is the // (for example, treating four-byte Float32 samples as eight-byte
// authoritative layout for the memory we are copying. // Float64 samples). The callback's AudioBufferList is authoritative for
// its channel layout, while its sample rate comes from the input stream.
let isInterleaved = buffers.count == 1 && channelCount > 1 let isInterleaved = buffers.count == 1 && channelCount > 1
return AVAudioFormat( return AVAudioFormat(
commonFormat: advertisedFormat.commonFormat, commonFormat: .pcmFormatFloat32,
sampleRate: advertisedFormat.sampleRate, sampleRate: advertisedFormat.sampleRate,
channels: AVAudioChannelCount(channelCount), channels: AVAudioChannelCount(channelCount),
interleaved: isInterleaved interleaved: isInterleaved
@@ -461,20 +502,32 @@ final class AudioManager: NSObject, ObservableObject {
from inputData: UnsafePointer<AudioBufferList>, from inputData: UnsafePointer<AudioBufferList>,
format: AVAudioFormat format: AVAudioFormat
) -> AVAudioPCMBuffer? { ) -> AVAudioPCMBuffer? {
guard let borrowedBuffer = AVAudioPCMBuffer(
pcmFormat: format,
bufferListNoCopy: inputData,
deallocator: nil
), borrowedBuffer.frameLength > 0,
let ownedBuffer = AVAudioPCMBuffer(
pcmFormat: format,
frameCapacity: borrowedBuffer.frameLength
) else { return nil }
ownedBuffer.frameLength = borrowedBuffer.frameLength
let sourceBuffers = UnsafeMutableAudioBufferListPointer( let sourceBuffers = UnsafeMutableAudioBufferListPointer(
UnsafeMutablePointer(mutating: inputData) UnsafeMutablePointer(mutating: inputData)
) )
if systemDiagnostics.firstBufferLayout == nil {
systemDiagnostics.firstBufferLayout = sourceBuffers.enumerated().map { index, buffer in
"buffer\(index):channels=\(buffer.mNumberChannels),bytes=\(buffer.mDataByteSize)"
}.joined(separator: "; ")
}
let frameLengths = sourceBuffers.compactMap { source -> AVAudioFrameCount? in
let bytesPerFrame = Int(source.mNumberChannels) * MemoryLayout<Float32>.size
guard bytesPerFrame > 0,
Int(source.mDataByteSize).isMultiple(of: bytesPerFrame) else { return nil }
return AVAudioFrameCount(Int(source.mDataByteSize) / bytesPerFrame)
}
guard frameLengths.count == sourceBuffers.count,
let frameLength = frameLengths.first,
frameLength > 0,
frameLengths.allSatisfy({ $0 == frameLength }),
let ownedBuffer = AVAudioPCMBuffer(
pcmFormat: format,
frameCapacity: frameLength
) else { return nil }
// Set the frame count explicitly instead of asking AVAudioPCMBuffer to
// infer it from potentially inconsistent tap metadata.
ownedBuffer.frameLength = frameLength
let destinationBuffers = UnsafeMutableAudioBufferListPointer( let destinationBuffers = UnsafeMutableAudioBufferListPointer(
ownedBuffer.mutableAudioBufferList ownedBuffer.mutableAudioBufferList
) )
@@ -497,20 +550,44 @@ final class AudioManager: NSObject, ObservableObject {
_ inputBufferProvider: () -> AVAudioPCMBuffer?, _ inputBufferProvider: () -> AVAudioPCMBuffer?,
converter: AVAudioConverter, converter: AVAudioConverter,
targetFormat: AVAudioFormat, targetFormat: AVAudioFormat,
source: AudioSource source: AudioSource,
callbackTimestamp: AudioTimeStamp? = nil,
ioTimestamp: AudioTimeStamp? = nil
) { ) {
// The system callback copies its borrowed Core Audio memory while this // The system callback copies its borrowed Core Audio memory while this
// lock prevents teardown, then conversion operates on the owned copy. // lock prevents teardown, then conversion operates on the owned copy.
audioFileLock.lock() audioFileLock.lock()
defer { audioFileLock.unlock() } defer { audioFileLock.unlock() }
guard isAcceptingAudio, guard isAcceptingAudio else { return }
let inputBuffer = inputBufferProvider(), if source == .system {
inputBuffer.frameLength > 0 else { return } systemDiagnostics.callbackCount += 1
let callbackAt = Date()
if systemDiagnostics.firstCallbackAt == nil { systemDiagnostics.firstCallbackAt = callbackAt }
systemDiagnostics.lastCallbackAt = callbackAt
if let callbackTimestamp {
if systemDiagnostics.firstSampleTime == nil { systemDiagnostics.firstSampleTime = callbackTimestamp.mSampleTime }
systemDiagnostics.lastSampleTime = callbackTimestamp.mSampleTime
}
if let ioTimestamp {
if systemDiagnostics.firstHostTime == nil { systemDiagnostics.firstHostTime = ioTimestamp.mHostTime }
systemDiagnostics.lastHostTime = ioTimestamp.mHostTime
}
}
guard let inputBuffer = inputBufferProvider(), inputBuffer.frameLength > 0 else {
if source == .system { systemDiagnostics.discardedCallbackCount += 1 }
return
}
if source == .system {
systemDiagnostics.inputFrameCount += UInt64(inputBuffer.frameLength)
}
updateAudioLevel(inputBuffer, source: source) updateAudioLevel(inputBuffer, source: source)
let ratio = targetFormat.sampleRate / inputBuffer.format.sampleRate let ratio = targetFormat.sampleRate / inputBuffer.format.sampleRate
let capacity = max(1, AVAudioFrameCount(ceil(Double(inputBuffer.frameLength) * ratio))) let capacity = max(1, AVAudioFrameCount(ceil(Double(inputBuffer.frameLength) * ratio)))
guard let outputBuffer = AVAudioPCMBuffer(pcmFormat: targetFormat, frameCapacity: capacity) else { return } guard let outputBuffer = AVAudioPCMBuffer(pcmFormat: targetFormat, frameCapacity: capacity) else {
if source == .system { systemDiagnostics.discardedCallbackCount += 1 }
return
}
var suppliedInput = false var suppliedInput = false
var conversionError: NSError? var conversionError: NSError?
let status = converter.convert(to: outputBuffer, error: &conversionError) { _, outputStatus in let status = converter.convert(to: outputBuffer, error: &conversionError) { _, outputStatus in
@@ -522,7 +599,10 @@ final class AudioManager: NSObject, ObservableObject {
outputStatus.pointee = .haveData outputStatus.pointee = .haveData
return inputBuffer return inputBuffer
} }
guard status != .error, conversionError == nil, outputBuffer.frameLength > 0 else { return } guard status != .error, conversionError == nil, outputBuffer.frameLength > 0 else {
if source == .system { systemDiagnostics.discardedCallbackCount += 1 }
return
}
do { do {
switch source { switch source {
@@ -530,6 +610,7 @@ final class AudioManager: NSObject, ObservableObject {
try micAudioFile?.write(from: outputBuffer) try micAudioFile?.write(from: outputBuffer)
case .system: case .system:
try systemAudioFile?.write(from: outputBuffer) try systemAudioFile?.write(from: outputBuffer)
systemDiagnostics.outputFrameCount += UInt64(outputBuffer.frameLength)
} }
} catch { } catch {
DispatchQueue.main.async { [weak self] in DispatchQueue.main.async { [weak self] in
@@ -623,15 +704,20 @@ final class AudioManager: NSObject, ObservableObject {
return String(format: "mic %.1fs, system %.1fs", micDuration, systemDuration) return String(format: "mic %.1fs, system %.1fs", micDuration, systemDuration)
} }
private func repairHalfDurationSystemWAVIfNeeded(in files: [URL?]) { private func repairHalfDurationSystemWAVIfNeeded(in files: [URL?]) -> Bool {
guard files.count >= 2, guard files.count >= 2,
let micDuration = audioDuration(at: files[0]), let micDuration = audioDuration(at: files[0]),
let systemURL = files[1], let systemURL = files[1],
let systemDuration = audioDuration(at: systemURL), let systemDuration = audioDuration(at: systemURL),
micDuration >= 60, micDuration >= 60,
systemURL.pathExtension.caseInsensitiveCompare("wav") == .orderedSame, systemURL.pathExtension.caseInsensitiveCompare("wav") == .orderedSame,
(0.48...0.52).contains(systemDuration / micDuration) else { return } (0.48...0.52).contains(systemDuration / micDuration) else { return false }
try? halveWAVSampleRate(at: systemURL) do {
try halveWAVSampleRate(at: systemURL)
return true
} catch {
return false
}
} }
private func halveWAVSampleRate(at url: URL) throws { private func halveWAVSampleRate(at url: URL) throws {
@@ -696,6 +782,112 @@ final class AudioManager: NSObject, ObservableObject {
file.processingFormat.sampleRate > 0 else { return nil } file.processingFormat.sampleRate > 0 else { return nil }
return Double(file.length) / file.processingFormat.sampleRate return Double(file.length) / file.processingFormat.sampleRate
} }
private func streamDescriptionSummary(_ description: AudioStreamBasicDescription?) -> String {
guard let description else { return "unavailable" }
return String(
format: "sampleRate=%.3f,formatID=%u,flags=%u,bytesPerPacket=%u,framesPerPacket=%u,bytesPerFrame=%u,channels=%u,bitsPerChannel=%u",
description.mSampleRate,
description.mFormatID,
description.mFormatFlags,
description.mBytesPerPacket,
description.mFramesPerPacket,
description.mBytesPerFrame,
description.mChannelsPerFrame,
description.mBitsPerChannel
)
}
private func audioFormatSummary(_ format: AVAudioFormat) -> String {
"sampleRate=\(format.sampleRate),channels=\(format.channelCount),commonFormat=\(format.commonFormat.rawValue),interleaved=\(format.isInterleaved)"
}
private func audioFileSummary(_ url: URL?) -> String {
guard let url else { return "missing" }
guard let file = try? AVAudioFile(forReading: url), file.processingFormat.sampleRate > 0 else {
return "\(url.lastPathComponent):unreadable"
}
let duration = Double(file.length) / file.processingFormat.sampleRate
return String(
format: "%@:sampleRate=%.3f,channels=%u,frames=%lld,duration=%.6f",
url.lastPathComponent,
file.processingFormat.sampleRate,
file.processingFormat.channelCount,
file.length,
duration
)
}
private func writeCaptureDiagnostics(
sessionID: UUID,
captureStartedAt: Date,
preRepairFileSummaries: [String],
repairedFiles: [URL?],
repairApplied: Bool,
audioFolder: URL?
) {
guard let audioFolder else { return }
let callbackDuration = systemDiagnostics.firstCallbackAt.flatMap { first in
systemDiagnostics.lastCallbackAt.map { $0.timeIntervalSince(first) }
}
let sampleTimeDelta = systemDiagnostics.firstSampleTime.flatMap { first in
systemDiagnostics.lastSampleTime.map { $0 - first }
}
let hostTimeDelta = systemDiagnostics.firstHostTime.flatMap { first in
systemDiagnostics.lastHostTime.map { $0 >= first ? $0 - first : 0 }
}
let inputFrameDuration = systemDiagnostics.selectedInputSampleRate.flatMap { sampleRate in
sampleRate > 0 ? Double(systemDiagnostics.inputFrameCount) / sampleRate : nil
}
let outputFrameDuration = systemDiagnostics.targetSampleRate.flatMap { sampleRate in
sampleRate > 0 ? Double(systemDiagnostics.outputFrameCount) / sampleRate : nil
}
let observedInputRate = callbackDuration.flatMap { duration in
duration > 0 ? Double(systemDiagnostics.inputFrameCount) / duration : nil
}
let preRepairMic = preRepairFileSummaries.indices.contains(0) ? preRepairFileSummaries[0] : "missing"
let preRepairSystem = preRepairFileSummaries.indices.contains(1) ? preRepairFileSummaries[1] : "missing"
let postRepairMic = repairedFiles.indices.contains(0) ? audioFileSummary(repairedFiles[0]) : "missing"
let postRepairSystem = repairedFiles.indices.contains(1) ? audioFileSummary(repairedFiles[1]) : "missing"
let callbackDurationLine = callbackDuration.map { String(format: "callbackWallDuration=%.6f", $0) } ?? "callbackWallDuration=unavailable"
let inputFrameDurationLine = inputFrameDuration.map { String(format: "inputFrameDurationAtSelectedRate=%.6f", $0) } ?? "inputFrameDurationAtSelectedRate=unavailable"
let outputFrameDurationLine = outputFrameDuration.map { String(format: "outputFrameDurationAtTargetRate=%.6f", $0) } ?? "outputFrameDurationAtTargetRate=unavailable"
let observedInputRateLine = observedInputRate.map { String(format: "observedInputFramesPerSecond=%.3f", $0) } ?? "observedInputFramesPerSecond=unavailable"
let sampleTimeDeltaLine = sampleTimeDelta.map { String(format: "sampleTimeDelta=%.6f", $0) } ?? "sampleTimeDelta=unavailable"
let hostTimeDeltaLine = hostTimeDelta.map { "hostTimeDelta=\($0)" } ?? "hostTimeDelta=unavailable"
let lines: [String] = [
"Meetingnotes system capture diagnostics",
"sessionID=\(sessionID.uuidString)",
"createdAt=\(ISO8601DateFormatter().string(from: Date()))",
String(format: "captureWallDuration=%.6f", Date().timeIntervalSince(captureStartedAt)),
"tapAdvertisedFormat=\(systemDiagnostics.tapAdvertisedFormat)",
"aggregateInputFormat=\(systemDiagnostics.aggregateInputFormat)",
"selectedInputFormat=\(systemDiagnostics.selectedInputFormat)",
"targetFormat=\(systemDiagnostics.targetFormat)",
"firstBufferLayout=\(systemDiagnostics.firstBufferLayout ?? "unavailable")",
"callbackCount=\(systemDiagnostics.callbackCount)",
"inputFrameCount=\(systemDiagnostics.inputFrameCount)",
"outputFrameCount=\(systemDiagnostics.outputFrameCount)",
"discardedCallbackCount=\(systemDiagnostics.discardedCallbackCount)",
callbackDurationLine,
inputFrameDurationLine,
outputFrameDurationLine,
observedInputRateLine,
sampleTimeDeltaLine,
hostTimeDeltaLine,
"repairApplied=\(repairApplied)",
"preRepairMic=\(preRepairMic)",
"preRepairSystem=\(preRepairSystem)",
"postRepairMic=\(postRepairMic)",
"postRepairSystem=\(postRepairSystem)"
]
let diagnosticsURL = audioFolder.appendingPathComponent("audio-diagnostics-\(sessionID.uuidString).txt")
try? lines.joined(separator: "\n").appending("\n").write(
to: diagnosticsURL,
atomically: true,
encoding: .utf8
)
}
private func resetAudioLevels() { private func resetAudioLevels() {
micAudioLevel = 0 micAudioLevel = 0
@@ -96,7 +96,11 @@ class RecordingSessionManager: ObservableObject {
let chunks = await audioManager.stopRecordingAndTranscribe() let chunks = await audioManager.stopRecordingAndTranscribe()
activeRecordingTranscriptChunks = chunks activeRecordingTranscriptChunks = chunks
activeRecordingTranscriptChunksUpdated = chunks activeRecordingTranscriptChunksUpdated = chunks
updateActiveMeetingTranscript(meetingId: meetingId, chunks: chunks) updateActiveMeetingTranscript(
meetingId: meetingId,
chunks: chunks,
transcriptionError: audioManager.errorMessage
)
activeMeetingId = nil activeMeetingId = nil
activeRecordingTranscriptChunks = [] activeRecordingTranscriptChunks = []
return chunks return chunks
@@ -117,7 +121,11 @@ class RecordingSessionManager: ObservableObject {
audioManager.lastRecoveryAudioFolderName audioManager.lastRecoveryAudioFolderName
} }
private func updateActiveMeetingTranscript(meetingId: UUID, chunks: [TranscriptChunk]) { private func updateActiveMeetingTranscript(
meetingId: UUID,
chunks: [TranscriptChunk],
transcriptionError: String? = nil
) {
// Load all meetings // Load all meetings
var meetings = LocalStorageManager.shared.loadMeetings() var meetings = LocalStorageManager.shared.loadMeetings()
@@ -127,6 +135,7 @@ class RecordingSessionManager: ObservableObject {
if let recoveryAudioFolderName = lastRecoveryAudioFolderName { if let recoveryAudioFolderName = lastRecoveryAudioFolderName {
meetings[index].recoveryAudioFolderName = recoveryAudioFolderName meetings[index].recoveryAudioFolderName = recoveryAudioFolderName
} }
meetings[index].transcriptionError = transcriptionError
// Save the updated meeting // Save the updated meeting
let success = LocalStorageManager.shared.saveMeeting(meetings[index]) let success = LocalStorageManager.shared.saveMeeting(meetings[index])
+3
View File
@@ -99,6 +99,7 @@ struct Meeting: Codable, Identifiable, Hashable {
var generatedNotes: String var generatedNotes: String
var templateId: UUID? // Add property to track per-meeting template var templateId: UUID? // Add property to track per-meeting template
var recoveryAudioFolderName: String? var recoveryAudioFolderName: String?
var transcriptionError: String?
// MARK: - Data versioning // MARK: - Data versioning
/// Version of this Meeting record on disk. Useful for migration. /// Version of this Meeting record on disk. Useful for migration.
var dataVersion: Int var dataVersion: Int
@@ -113,6 +114,7 @@ struct Meeting: Codable, Identifiable, Hashable {
generatedNotes: String = "", generatedNotes: String = "",
templateId: UUID? = nil, templateId: UUID? = nil,
recoveryAudioFolderName: String? = nil, recoveryAudioFolderName: String? = nil,
transcriptionError: String? = nil,
dataVersion: Int = Meeting.currentDataVersion) { dataVersion: Int = Meeting.currentDataVersion) {
self.id = id self.id = id
self.date = date self.date = date
@@ -122,6 +124,7 @@ struct Meeting: Codable, Identifiable, Hashable {
self.generatedNotes = generatedNotes self.generatedNotes = generatedNotes
self.templateId = templateId self.templateId = templateId
self.recoveryAudioFolderName = recoveryAudioFolderName self.recoveryAudioFolderName = recoveryAudioFolderName
self.transcriptionError = transcriptionError
self.dataVersion = dataVersion self.dataVersion = dataVersion
} }
+13 -6
View File
@@ -63,6 +63,10 @@ final class ProcessTap {
@ObservationIgnored @ObservationIgnored
private(set) var tapStreamDescription: AudioStreamBasicDescription? private(set) var tapStreamDescription: AudioStreamBasicDescription?
@ObservationIgnored @ObservationIgnored
private(set) var tapAdvertisedStreamDescription: AudioStreamBasicDescription?
@ObservationIgnored
private(set) var aggregateInputStreamDescription: AudioStreamBasicDescription?
@ObservationIgnored
private var invalidationHandler: InvalidationHandler? private var invalidationHandler: InvalidationHandler?
@ObservationIgnored @ObservationIgnored
@@ -138,11 +142,11 @@ final class ProcessTap {
tapDescription = CATapDescription(stereoMixdownOfProcesses: [process.objectID]) tapDescription = CATapDescription(stereoMixdownOfProcesses: [process.objectID])
logger.debug("Configuring tap for single process objectID: \(process.objectID)") logger.debug("Configuring tap for single process objectID: \(process.objectID)")
case .systemAudio: case .systemAudio:
// Keep the HAL tap's buffer layout consistent with the default // The transcription file is mono, so ask Core Audio for a mono
// output stream. AudioManager performs the stereo-to-mono mix when // mixdown at the source. This avoids interpreting a stereo HAL
// it converts the captured audio to the 16 kHz transcription file. // buffer as half as many frames before the 16 kHz conversion.
tapDescription = CATapDescription(stereoGlobalTapButExcludeProcesses: []) tapDescription = CATapDescription(monoGlobalTapButExcludeProcesses: [])
logger.debug("Configuring a stereo global system audio tap.") logger.info("Configuring a mono global system audio tap.")
} }
tapDescription.uuid = UUID() tapDescription.uuid = UUID()
@@ -249,6 +253,7 @@ final class ProcessTap {
do { do {
logger.debug("Attempting to read audio tap stream basic description for tapID #\(tapID)...") logger.debug("Attempting to read audio tap stream basic description for tapID #\(tapID)...")
let advertisedDescription = try tapID.readAudioTapStreamBasicDescription() let advertisedDescription = try tapID.readAudioTapStreamBasicDescription()
self.tapAdvertisedStreamDescription = advertisedDescription
// The aggregate device may adapt the tap to the active hardware's // The aggregate device may adapt the tap to the active hardware's
// sample rate. Its input stream is the format actually delivered // sample rate. Its input stream is the format actually delivered
@@ -256,7 +261,9 @@ final class ProcessTap {
// advertisement. Using the latter can halve the written duration // advertisement. Using the latter can halve the written duration
// when, for example, a 48 kHz tap is delivered at 24 kHz. // when, for example, a 48 kHz tap is delivered at 24 kHz.
do { do {
self.tapStreamDescription = try aggregateDeviceID.readInputStreamBasicDescription() let aggregateDescription = try aggregateDeviceID.readInputStreamBasicDescription()
self.aggregateInputStreamDescription = aggregateDescription
self.tapStreamDescription = aggregateDescription
logger.info("Using aggregate input stream description: \(String(describing: self.tapStreamDescription), privacy: .public); tap advertised: \(String(describing: advertisedDescription), privacy: .public)") logger.info("Using aggregate input stream description: \(String(describing: self.tapStreamDescription), privacy: .public); tap advertised: \(String(describing: advertisedDescription), privacy: .public)")
} catch { } catch {
self.tapStreamDescription = advertisedDescription self.tapStreamDescription = advertisedDescription
+17 -2
View File
@@ -142,12 +142,23 @@ class MeetingViewModel: ObservableObject {
.dropFirst() .dropFirst()
.sink { [weak self] updatedChunks in .sink { [weak self] updatedChunks in
guard let self = self else { return } guard let self = self else { return }
// Only update if this meeting is the active recording // Stop publishes the final chunks after capture has ended but
if recordingSessionManager.isRecordingMeeting(self.meeting.id) { // before the active meeting is cleared.
if recordingSessionManager.activeMeetingId == self.meeting.id {
self.meeting.transcriptChunks = updatedChunks self.meeting.transcriptChunks = updatedChunks
} }
} }
.store(in: &cancellables) .store(in: &cancellables)
NotificationCenter.default.publisher(for: .meetingSaved)
.compactMap { $0.object as? Meeting }
.filter { [weak self] in $0.id == self?.meeting.id }
.sink { [weak self] savedMeeting in
guard let self, !self.isDeleted, self.meeting != savedMeeting else { return }
self.meeting = savedMeeting
self.refreshRecoveryAudioFolder()
}
.store(in: &cancellables)
@@ -215,6 +226,7 @@ class MeetingViewModel: ObservableObject {
let chunks = await recordingSessionManager.stopRecording() let chunks = await recordingSessionManager.stopRecording()
meeting.transcriptChunks = chunks meeting.transcriptChunks = chunks
meeting.recoveryAudioFolderName = recordingSessionManager.lastRecoveryAudioFolderName meeting.recoveryAudioFolderName = recordingSessionManager.lastRecoveryAudioFolderName
meeting.transcriptionError = recordingSessionManager.errorMessage
refreshRecoveryAudioFolder() refreshRecoveryAudioFolder()
saveMeeting() saveMeeting()
if !meeting.formattedTranscript.isEmpty { if !meeting.formattedTranscript.isEmpty {
@@ -238,6 +250,7 @@ class MeetingViewModel: ObservableObject {
) )
meeting.transcriptChunks = chunks meeting.transcriptChunks = chunks
meeting.recoveryAudioFolderName = recoveryAudioFolderURL.lastPathComponent meeting.recoveryAudioFolderName = recoveryAudioFolderURL.lastPathComponent
meeting.transcriptionError = nil
selectedTab = .transcript selectedTab = .transcript
guard saveMeeting() else { guard saveMeeting() else {
@@ -247,6 +260,8 @@ class MeetingViewModel: ObservableObject {
await generateNotes() await generateNotes()
} catch { } catch {
errorMessage = error.localizedDescription errorMessage = error.localizedDescription
meeting.transcriptionError = error.localizedDescription
_ = saveMeeting()
print("Retry transcription failed: \(error)") print("Retry transcription failed: \(error)")
} }
} }
+91 -14
View File
@@ -6,6 +6,8 @@ struct MeetingListView: View {
@StateObject private var recordingSessionManager = RecordingSessionManager.shared @StateObject private var recordingSessionManager = RecordingSessionManager.shared
@State private var selectedMeeting: Meeting? @State private var selectedMeeting: Meeting?
@State private var navigationPath = NavigationPath() @State private var navigationPath = NavigationPath()
@State private var recordingFailureMessage: String?
@State private var failedMeetingID: UUID?
var body: some View { var body: some View {
NavigationSplitView { NavigationSplitView {
@@ -23,6 +25,38 @@ struct MeetingListView: View {
.background(Color.clear) .background(Color.clear)
} }
} }
.onReceive(recordingSessionManager.$activeMeetingId.compactMap { $0 }) { meetingID in
if let meeting = LocalStorageManager.shared.loadMeetings().first(where: { $0.id == meetingID }) {
selectedMeeting = meeting
}
}
.onReceive(NotificationCenter.default.publisher(for: .meetingSaved)) { notification in
guard let savedMeeting = notification.object as? Meeting,
selectedMeeting?.id == savedMeeting.id else { return }
selectedMeeting = savedMeeting
}
.onReceive(recordingSessionManager.$errorMessage.compactMap { $0 }) { message in
failedMeetingID = recordingSessionManager.activeMeetingId
recordingFailureMessage = message
}
.alert("Transcription Failed", isPresented: Binding(
get: { recordingFailureMessage != nil },
set: { if !$0 { recordingFailureMessage = nil } }
)) {
if failedMeetingID != nil {
Button("View Meeting") {
selectFailedMeeting()
recordingFailureMessage = nil
recordingSessionManager.errorMessage = nil
}
}
Button("OK") {
recordingFailureMessage = nil
recordingSessionManager.errorMessage = nil
}
} message: {
Text(recordingFailureMessage ?? "")
}
} }
private var sidebarContent: some View { private var sidebarContent: some View {
@@ -161,6 +195,14 @@ struct MeetingListView: View {
return DayGroup(day: dayString, date: date, meetings: meetings.sorted { $0.date > $1.date }) return DayGroup(day: dayString, date: date, meetings: meetings.sorted { $0.date > $1.date })
}.sorted { $0.date > $1.date } }.sorted { $0.date > $1.date }
} }
private func selectFailedMeeting() {
guard let failedMeetingID,
let meeting = LocalStorageManager.shared.loadMeetings().first(where: { $0.id == failedMeetingID }) else {
return
}
selectedMeeting = meeting
}
} }
struct DayGroup { struct DayGroup {
@@ -182,7 +224,14 @@ struct MeetingRowView: View {
.foregroundColor(.red) .foregroundColor(.red)
.font(.headline) .font(.headline)
} }
Text(meeting.title.isEmpty ? "Untitled meeting" : meeting.title) if meeting.transcriptionError != nil {
Image(systemName: "exclamationmark.triangle.fill")
.foregroundColor(.orange)
.accessibilityLabel("Transcription failed")
}
Text(meeting.title.isEmpty
? (meeting.transcriptionError == nil ? "Untitled meeting" : "Transcription failed")
: meeting.title)
.font(.headline) .font(.headline)
.lineLimit(1) .lineLimit(1)
} }
@@ -542,24 +591,52 @@ struct MeetingDetailContentView: View {
} }
private var transcriptView: some View { private var transcriptView: some View {
ScrollView { VStack(alignment: .leading, spacing: 10) {
if viewModel.meeting.collapsedTranscriptChunks.isEmpty { if let transcriptionError = viewModel.meeting.transcriptionError {
Text("Transcript will appear here...") HStack(alignment: .top, spacing: 10) {
.frame(maxWidth: .infinity, alignment: .leading) Image(systemName: "exclamationmark.triangle.fill")
.padding() .foregroundColor(.orange)
.foregroundColor(.secondary) VStack(alignment: .leading, spacing: 4) {
} else { Text("Transcription failed")
LazyVStack(alignment: .leading, spacing: 4) { .font(.headline)
ForEach(viewModel.meeting.collapsedTranscriptChunks) { chunk in Text(transcriptionError)
CollapsedTranscriptChunkView(chunk: chunk) .font(.caption)
.foregroundColor(.secondary)
}
Spacer()
if viewModel.recoveryAudioFolderURL != nil {
Button("Retry") {
viewModel.retryTranscription()
}
.disabled(!viewModel.canRetryTranscription)
} }
} }
.padding() .padding()
.background(Color.orange.opacity(0.08))
.cornerRadius(8)
} }
ScrollView {
if viewModel.meeting.collapsedTranscriptChunks.isEmpty {
Text(viewModel.meeting.transcriptionError == nil
? "Transcript will appear here..."
: "No transcript was produced. The recording was kept and can be retried.")
.frame(maxWidth: .infinity, alignment: .leading)
.padding()
.foregroundColor(.secondary)
} else {
LazyVStack(alignment: .leading, spacing: 4) {
ForEach(viewModel.meeting.collapsedTranscriptChunks) { chunk in
CollapsedTranscriptChunkView(chunk: chunk)
}
}
.padding()
}
}
.frame(maxHeight: .infinity)
.background(Color.gray.opacity(0.05))
.cornerRadius(8)
} }
.frame(maxHeight: .infinity)
.background(Color.gray.opacity(0.05))
.cornerRadius(8)
} }
private var enhancedNotesView: some View { private var enhancedNotesView: some View {