Compare commits

...
6 Commits
10 changed files with 589 additions and 107 deletions
+4 -4
View File
@@ -276,7 +276,7 @@
CODE_SIGN_IDENTITY = "Apple Development";
CODE_SIGN_STYLE = Automatic;
COMBINE_HIDPI_IMAGES = YES;
CURRENT_PROJECT_VERSION = 38;
CURRENT_PROJECT_VERSION = 44;
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 = 1.1.26;
MARKETING_VERSION = 1.1.32;
ONLY_ACTIVE_ARCH = NO;
OTHER_SWIFT_FLAGS = "$(inherited) -D ENABLE_TCC_SPI";
PRODUCT_BUNDLE_IDENTIFIER = net.jamesbone.meetingnotes;
@@ -312,7 +312,7 @@
CODE_SIGN_IDENTITY = "Apple Development";
CODE_SIGN_STYLE = Automatic;
COMBINE_HIDPI_IMAGES = YES;
CURRENT_PROJECT_VERSION = 38;
CURRENT_PROJECT_VERSION = 44;
DEVELOPMENT_ASSET_PATHS = "\"meetingnotes/Preview Content\"";
DEVELOPMENT_TEAM = G9LVHZAJNX;
ENABLE_HARDENED_RUNTIME = YES;
@@ -326,7 +326,7 @@
"@executable_path/../Frameworks",
);
MACOSX_DEPLOYMENT_TARGET = 15.0;
MARKETING_VERSION = 1.1.26;
MARKETING_VERSION = 1.1.32;
ONLY_ACTIVE_ARCH = YES;
OTHER_SWIFT_FLAGS = "$(inherited) -D ENABLE_TCC_SPI";
PRODUCT_BUNDLE_IDENTIFIER = net.jamesbone.meetingnotes;
+334 -33
View File
@@ -20,6 +20,27 @@ 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 inputRateInference = "unavailable"
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.
@MainActor
final class AudioManager: NSObject, ObservableObject {
@@ -50,8 +71,9 @@ final class AudioManager: NSObject, ObservableObject {
private var micAudioURL: URL?
private var systemAudioURL: URL?
private var recordingStartedAt = Date()
private var systemDiagnostics = SystemCaptureDiagnostics()
private override init() {
override init() {
super.init()
observeAudioEngine()
}
@@ -67,6 +89,7 @@ final class AudioManager: NSObject, ObservableObject {
sessionID = UUID()
self.meetingID = meetingID
recordingStartedAt = Date()
systemDiagnostics = SystemCaptureDiagnostics()
do {
try prepareAudioFiles()
startMicrophoneTap()
@@ -79,6 +102,7 @@ final class AudioManager: NSObject, ObservableObject {
func stopRecordingAndTranscribe() async -> [TranscriptChunk] {
let completedMeetingID = meetingID
let completedSessionID = sessionID
let captureStartedAt = recordingStartedAt
let files = stopCaptureAndCloseFiles()
isProcessing = true
@@ -86,11 +110,20 @@ final class AudioManager: NSObject, ObservableObject {
isProcessing = false
}
repairHalfDurationSystemWAVIfNeeded(in: files)
let preRepairFileSummaries = files.map(audioFileSummary)
let repairApplied = repairHalfDurationSystemWAVIfNeeded(in: files)
let completedFiles = files.compactMap { $0 }
let audioFolder = preserveAudioFiles(completedFiles, meetingID: completedMeetingID)
let transcriptionFiles = preservedAudioFiles(files, in: audioFolder)
lastRecoveryAudioFolderName = audioFolder?.lastPathComponent
writeCaptureDiagnostics(
sessionID: completedSessionID,
captureStartedAt: captureStartedAt,
preRepairFileSummaries: preRepairFileSummaries,
repairedFiles: transcriptionFiles,
repairApplied: repairApplied,
audioFolder: audioFolder
)
if let mismatch = captureDurationMismatch(in: transcriptionFiles) {
let recoveryMessage = audioFolder == nil
? " The audio remains in the app's temporary folder."
@@ -405,18 +438,76 @@ final class AudioManager: NSObject, ObservableObject {
advertisedInputFormat.sampleRate > 0 else {
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 converter: AVAudioConverter?
try tap.run(on: tapQueue) { [weak self] _, inputData, _, _, _ in
var pendingBuffer: AVAudioPCMBuffer?
var pendingInputTime: AudioTimeStamp?
var pendingNow: AudioTimeStamp?
try tap.run(on: tapQueue) { [weak self] inNow, inputData, inInputTime, _, _ in
guard let self else { return }
if inputFormat == nil {
guard let callbackFormat = self.inputFormat(
for: inputData,
sampleRate: advertisedInputFormat.sampleRate
),
let currentBuffer = self.copyAudioBuffer(from: inputData, format: callbackFormat) else {
self.systemDiagnostics.discardedCallbackCount += 1
return
}
guard let previousBuffer = pendingBuffer,
let previousInputTime = pendingInputTime,
let previousNow = pendingNow else {
pendingBuffer = currentBuffer
pendingInputTime = inInputTime.pointee
pendingNow = inNow.pointee
return
}
let selectedSampleRate = self.effectiveInputSampleRate(
advertisedSampleRate: advertisedInputFormat.sampleRate,
previousFrameLength: previousBuffer.frameLength,
previousTimestamp: previousInputTime,
currentTimestamp: inInputTime.pointee,
previousHostTimestamp: previousNow,
currentHostTimestamp: inNow.pointee
)
inputFormat = self.inputFormat(
for: inputData,
advertisedFormat: advertisedInputFormat
sampleRate: selectedSampleRate
)
if let inputFormat {
self.systemDiagnostics.selectedInputFormat = self.audioFormatSummary(inputFormat)
self.systemDiagnostics.selectedInputSampleRate = inputFormat.sampleRate
converter = AVAudioConverter(from: inputFormat, to: targetFormat)
}
guard let inputFormat, let converter else { return }
self.processAudioBuffer(
{ self.copyAudioBuffer(from: previousBuffer.audioBufferList, format: inputFormat) },
converter: converter,
targetFormat: targetFormat,
source: .system,
callbackTimestamp: previousInputTime,
ioTimestamp: pendingNow
)
pendingBuffer = nil
pendingInputTime = nil
pendingNow = nil
self.processAudioBuffer(
{ self.copyAudioBuffer(from: currentBuffer.audioBufferList, format: inputFormat) },
converter: converter,
targetFormat: targetFormat,
source: .system,
callbackTimestamp: inInputTime.pointee,
ioTimestamp: inNow.pointee
)
return
}
guard let inputFormat, let converter else { return }
// The tap queue is serial. Reusing the converter preserves its
@@ -425,7 +516,9 @@ final class AudioManager: NSObject, ObservableObject {
{ self.copyAudioBuffer(from: inputData, format: inputFormat) },
converter: converter,
targetFormat: targetFormat,
source: .system
source: .system,
callbackTimestamp: inInputTime.pointee,
ioTimestamp: inNow.pointee
)
} invalidationHandler: { [weak self] _ in
guard let self, self.isRecording else { return }
@@ -435,7 +528,7 @@ final class AudioManager: NSObject, ObservableObject {
private func inputFormat(
for inputData: UnsafePointer<AudioBufferList>,
advertisedFormat: AVAudioFormat
sampleRate: Double
) -> AVAudioFormat? {
let buffers = UnsafeMutableAudioBufferListPointer(
UnsafeMutablePointer(mutating: inputData)
@@ -443,38 +536,106 @@ final class AudioManager: NSObject, ObservableObject {
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.
// HAL I/O proc samples use the canonical Float32 representation. The
// tap's stream description can advertise a different common format;
// using that to interpret the callback bytes can halve the frame count
// (for example, treating four-byte Float32 samples as eight-byte
// 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
return AVAudioFormat(
commonFormat: advertisedFormat.commonFormat,
sampleRate: advertisedFormat.sampleRate,
commonFormat: .pcmFormatFloat32,
sampleRate: sampleRate,
channels: AVAudioChannelCount(channelCount),
interleaved: isInterleaved
)
}
private func effectiveInputSampleRate(
advertisedSampleRate: Double,
previousFrameLength: AVAudioFrameCount,
previousTimestamp: AudioTimeStamp,
currentTimestamp: AudioTimeStamp,
previousHostTimestamp: AudioTimeStamp,
currentHostTimestamp: AudioTimeStamp
) -> Double {
let sampleTimeDelta = currentTimestamp.mSampleTime - previousTimestamp.mSampleTime
guard advertisedSampleRate > 0, previousFrameLength > 0 else {
systemDiagnostics.inputRateInference = "advertised (invalid advertised rate or frame count)"
return advertisedSampleRate
}
guard currentHostTimestamp.mHostTime > previousHostTimestamp.mHostTime else {
systemDiagnostics.inputRateInference = "advertised (host timestamps unavailable)"
return advertisedSampleRate
}
let hostTimeDelta = currentHostTimestamp.mHostTime - previousHostTimestamp.mHostTime
let hostNanoseconds = AudioConvertHostTimeToNanos(hostTimeDelta)
guard hostNanoseconds > 0 else {
systemDiagnostics.inputRateInference = "advertised (host-time conversion failed)"
return advertisedSampleRate
}
let inferredSampleRate = Double(previousFrameLength) * 1_000_000_000 / Double(hostNanoseconds)
let plausibleRange = (advertisedSampleRate * 0.25)...(advertisedSampleRate * 1.25)
guard inferredSampleRate.isFinite, plausibleRange.contains(inferredSampleRate) else {
systemDiagnostics.inputRateInference = String(
format: "advertised (invalid host inference %.3f from %u frames / %llu ns; sampleTimeDelta=%.3f)",
inferredSampleRate,
previousFrameLength,
hostNanoseconds,
sampleTimeDelta
)
return advertisedSampleRate
}
let commonSampleRates: [Double] = [8_000, 11_025, 12_000, 16_000, 22_050, 24_000, 32_000, 44_100, 48_000, 88_200, 96_000]
let selectedSampleRate = commonSampleRates
.min(by: { abs($0 - inferredSampleRate) < abs($1 - inferredSampleRate) })
.flatMap { abs($0 - inferredSampleRate) / $0 <= 0.01 ? $0 : nil }
?? inferredSampleRate
systemDiagnostics.inputRateInference = String(
format: "source=hostTime,advertised=%.3f,inferred=%.3f,selected=%.3f,previousFrames=%u,hostNanoseconds=%llu,sampleTimeDelta=%.3f",
advertisedSampleRate,
inferredSampleRate,
selectedSampleRate,
previousFrameLength,
hostNanoseconds,
sampleTimeDelta
)
return selectedSampleRate
}
private func copyAudioBuffer(
from inputData: UnsafePointer<AudioBufferList>,
format: AVAudioFormat
) -> 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(
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(
ownedBuffer.mutableAudioBufferList
)
@@ -497,20 +658,44 @@ final class AudioManager: NSObject, ObservableObject {
_ inputBufferProvider: () -> AVAudioPCMBuffer?,
converter: AVAudioConverter,
targetFormat: AVAudioFormat,
source: AudioSource
source: AudioSource,
callbackTimestamp: AudioTimeStamp? = nil,
ioTimestamp: AudioTimeStamp? = nil
) {
// The system callback copies its borrowed Core Audio memory while this
// lock prevents teardown, then conversion operates on the owned copy.
audioFileLock.lock()
defer { audioFileLock.unlock() }
guard isAcceptingAudio,
let inputBuffer = inputBufferProvider(),
inputBuffer.frameLength > 0 else { return }
guard isAcceptingAudio else { return }
if source == .system {
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)
let ratio = targetFormat.sampleRate / inputBuffer.format.sampleRate
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 conversionError: NSError?
let status = converter.convert(to: outputBuffer, error: &conversionError) { _, outputStatus in
@@ -522,7 +707,10 @@ final class AudioManager: NSObject, ObservableObject {
outputStatus.pointee = .haveData
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 {
switch source {
@@ -530,6 +718,7 @@ final class AudioManager: NSObject, ObservableObject {
try micAudioFile?.write(from: outputBuffer)
case .system:
try systemAudioFile?.write(from: outputBuffer)
systemDiagnostics.outputFrameCount += UInt64(outputBuffer.frameLength)
}
} catch {
DispatchQueue.main.async { [weak self] in
@@ -623,15 +812,20 @@ final class AudioManager: NSObject, ObservableObject {
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,
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)
(0.48...0.52).contains(systemDuration / micDuration) else { return false }
do {
try halveWAVSampleRate(at: systemURL)
return true
} catch {
return false
}
}
private func halveWAVSampleRate(at url: URL) throws {
@@ -697,6 +891,113 @@ final class AudioManager: NSObject, ObservableObject {
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)",
"inputRateInference=\(systemDiagnostics.inputRateInference)",
"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() {
micAudioLevel = 0
systemAudioLevel = 0
@@ -170,11 +170,13 @@ class LocalStorageManager {
}
}
for audioFile in recoveryAudioFiles(in: continuationFolder) {
let destination = previousFolder.appendingPathComponent(audioFile.url.lastPathComponent)
let recoveryFiles = recoveryAudioFiles(in: continuationFolder).map(\.url)
+ recoveryDiagnosticFiles(in: continuationFolder)
for recoveryFile in recoveryFiles {
let destination = previousFolder.appendingPathComponent(recoveryFile.lastPathComponent)
if FileManager.default.fileExists(atPath: destination.path) {
guard FileManager.default.contentsEqual(
atPath: audioFile.url.path,
atPath: recoveryFile.path,
andPath: destination.path
) else {
rollbackMergedAudio(copiedAudioURLs, removeFolder: createdPreviousFolder ? previousFolder : nil)
@@ -184,7 +186,7 @@ class LocalStorageManager {
}
do {
try FileManager.default.copyItem(at: audioFile.url, to: destination)
try FileManager.default.copyItem(at: recoveryFile, to: destination)
copiedAudioURLs.append(destination)
} catch {
rollbackMergedAudio(copiedAudioURLs, removeFolder: createdPreviousFolder ? previousFolder : nil)
@@ -206,6 +208,7 @@ class LocalStorageManager {
merged.userNotes = mergedText(previous.userNotes, continuation.userNotes)
merged.generatedNotes = mergedText(previous.generatedNotes, continuation.generatedNotes)
merged.templateId = previous.templateId ?? continuation.templateId
merged.transcriptionError = previous.transcriptionError ?? continuation.transcriptionError
if !recoveryAudioFiles(in: previousFolder).isEmpty {
merged.recoveryAudioFolderName = previous.id.uuidString
}
@@ -347,6 +350,23 @@ class LocalStorageManager {
}
}
private func recoveryDiagnosticFiles(in folder: URL) -> [URL] {
guard let files = try? FileManager.default.contentsOfDirectory(
at: folder,
includingPropertiesForKeys: [.isRegularFileKey],
options: [.skipsHiddenFiles]
) else {
return []
}
return files.filter { url in
let values = try? url.resourceValues(forKeys: [.isRegularFileKey])
return values?.isRegularFile == true
&& url.pathExtension.caseInsensitiveCompare("txt") == .orderedSame
&& url.lastPathComponent.lowercased().hasPrefix("audio-diagnostics-")
}
}
func findRecoveryAudioFolder(for meeting: Meeting) -> URL? {
let canonicalName = meeting.id.uuidString
guard meeting.recoveryAudioFolderName == nil
@@ -2,6 +2,12 @@ import Foundation
import SwiftUI
import Combine
struct RecordingCompletion {
let chunks: [TranscriptChunk]
let recoveryAudioFolderName: String?
let transcriptionError: String?
}
/// Manages recording sessions at the app level to persist across navigation
@MainActor
class RecordingSessionManager: ObservableObject {
@@ -14,9 +20,11 @@ class RecordingSessionManager: ObservableObject {
@Published var errorMessage: String?
@Published var activeRecordingTranscriptChunksUpdated: [TranscriptChunk] = []
private let audioManager = AudioManager.shared
private var audioManager = AudioManager.shared
private var cancellables = Set<AnyCancellable>()
private var audioManagerCancellables = Set<AnyCancellable>()
private let transcriptUpdateSubject = PassthroughSubject<[TranscriptChunk], Never>()
private var processingMeetingIds = Set<UUID>()
// Store transcript chunks for the active recording session
private var activeRecordingTranscriptChunks: [TranscriptChunk] = []
@@ -27,24 +35,19 @@ class RecordingSessionManager: ObservableObject {
}
private func setupAudioManagerBindings() {
audioManagerCancellables.removeAll()
// Bind to audio manager state
audioManager.$isRecording
.sink { [weak self] isRecording in
self?.isRecording = isRecording
}
.store(in: &cancellables)
audioManager.$isProcessing
.sink { [weak self] isProcessing in
self?.isProcessing = isProcessing
}
.store(in: &cancellables)
.store(in: &audioManagerCancellables)
audioManager.$errorMessage
.sink { [weak self] errorMessage in
self?.errorMessage = errorMessage
}
.store(in: &cancellables)
.store(in: &audioManagerCancellables)
// When transcript chunks change, store them for the active recording and send to debouncer
audioManager.$transcriptChunks
@@ -55,7 +58,7 @@ class RecordingSessionManager: ObservableObject {
self.transcriptUpdateSubject.send(newChunks)
}
.store(in: &cancellables)
.store(in: &audioManagerCancellables)
}
private func setupDebouncedSaving() {
@@ -84,22 +87,40 @@ class RecordingSessionManager: ObservableObject {
audioManager.startRecording(for: meetingId)
}
func stopRecording() async -> [TranscriptChunk] {
func stopRecording() async -> RecordingCompletion {
print("🛑 Stopping recording for meeting: \(activeMeetingId?.uuidString ?? "unknown")")
guard let meetingId = activeMeetingId else {
audioManager.cancelRecording()
recordingStartedAt = nil
return []
return RecordingCompletion(chunks: [], recoveryAudioFolderName: nil, transcriptionError: nil)
}
recordingStartedAt = nil
let chunks = await audioManager.stopRecordingAndTranscribe()
activeRecordingTranscriptChunks = chunks
activeRecordingTranscriptChunksUpdated = chunks
updateActiveMeetingTranscript(meetingId: meetingId, chunks: chunks)
let completedAudioManager = audioManager
audioManager = AudioManager()
setupAudioManagerBindings()
activeMeetingId = nil
activeRecordingTranscriptChunks = []
return chunks
recordingStartedAt = nil
processingMeetingIds.insert(meetingId)
isProcessing = true
defer {
processingMeetingIds.remove(meetingId)
isProcessing = !processingMeetingIds.isEmpty
}
let chunks = await completedAudioManager.stopRecordingAndTranscribe()
let completion = RecordingCompletion(
chunks: chunks,
recoveryAudioFolderName: completedAudioManager.lastRecoveryAudioFolderName,
transcriptionError: completedAudioManager.errorMessage
)
updateActiveMeetingTranscript(
meetingId: meetingId,
chunks: chunks,
recoveryAudioFolderName: completion.recoveryAudioFolderName,
transcriptionError: completion.transcriptionError
)
return completion
}
func cancelRecording() {
@@ -113,20 +134,26 @@ class RecordingSessionManager: ObservableObject {
return isRecording && activeMeetingId == meetingId
}
var lastRecoveryAudioFolderName: String? {
audioManager.lastRecoveryAudioFolderName
func isProcessingMeeting(_ meetingId: UUID) -> Bool {
processingMeetingIds.contains(meetingId)
}
private func updateActiveMeetingTranscript(meetingId: UUID, chunks: [TranscriptChunk]) {
private func updateActiveMeetingTranscript(
meetingId: UUID,
chunks: [TranscriptChunk],
recoveryAudioFolderName: String? = nil,
transcriptionError: String? = nil
) {
// Load all meetings
var meetings = LocalStorageManager.shared.loadMeetings()
// Find and update the active meeting
if let index = meetings.firstIndex(where: { $0.id == meetingId }) {
meetings[index].transcriptChunks = chunks
if let recoveryAudioFolderName = lastRecoveryAudioFolderName {
if let recoveryAudioFolderName {
meetings[index].recoveryAudioFolderName = recoveryAudioFolderName
}
meetings[index].transcriptionError = transcriptionError
// Save the updated meeting
let success = LocalStorageManager.shared.saveMeeting(meetings[index])
+3
View File
@@ -99,6 +99,7 @@ struct Meeting: Codable, Identifiable, Hashable {
var generatedNotes: String
var templateId: UUID? // Add property to track per-meeting template
var recoveryAudioFolderName: String?
var transcriptionError: String?
// MARK: - Data versioning
/// Version of this Meeting record on disk. Useful for migration.
var dataVersion: Int
@@ -113,6 +114,7 @@ struct Meeting: Codable, Identifiable, Hashable {
generatedNotes: String = "",
templateId: UUID? = nil,
recoveryAudioFolderName: String? = nil,
transcriptionError: String? = nil,
dataVersion: Int = Meeting.currentDataVersion) {
self.id = id
self.date = date
@@ -122,6 +124,7 @@ struct Meeting: Codable, Identifiable, Hashable {
self.generatedNotes = generatedNotes
self.templateId = templateId
self.recoveryAudioFolderName = recoveryAudioFolderName
self.transcriptionError = transcriptionError
self.dataVersion = dataVersion
}
+13 -6
View File
@@ -63,6 +63,10 @@ final class ProcessTap {
@ObservationIgnored
private(set) var tapStreamDescription: AudioStreamBasicDescription?
@ObservationIgnored
private(set) var tapAdvertisedStreamDescription: AudioStreamBasicDescription?
@ObservationIgnored
private(set) var aggregateInputStreamDescription: AudioStreamBasicDescription?
@ObservationIgnored
private var invalidationHandler: InvalidationHandler?
@ObservationIgnored
@@ -138,11 +142,11 @@ final class ProcessTap {
tapDescription = CATapDescription(stereoMixdownOfProcesses: [process.objectID])
logger.debug("Configuring tap for single process objectID: \(process.objectID)")
case .systemAudio:
// Keep the HAL tap's buffer layout consistent with the default
// output stream. AudioManager performs the stereo-to-mono mix when
// it converts the captured audio to the 16 kHz transcription file.
tapDescription = CATapDescription(stereoGlobalTapButExcludeProcesses: [])
logger.debug("Configuring a stereo global system audio tap.")
// The transcription file is mono, so ask Core Audio for a mono
// mixdown at the source. This avoids interpreting a stereo HAL
// buffer as half as many frames before the 16 kHz conversion.
tapDescription = CATapDescription(monoGlobalTapButExcludeProcesses: [])
logger.info("Configuring a mono global system audio tap.")
}
tapDescription.uuid = UUID()
@@ -249,6 +253,7 @@ final class ProcessTap {
do {
logger.debug("Attempting to read audio tap stream basic description for tapID #\(tapID)...")
let advertisedDescription = try tapID.readAudioTapStreamBasicDescription()
self.tapAdvertisedStreamDescription = advertisedDescription
// The aggregate device may adapt the tap to the active hardware's
// 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
// when, for example, a 48 kHz tap is delivered at 24 kHz.
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)")
} catch {
self.tapStreamDescription = advertisedDescription
+35 -2
View File
@@ -30,6 +30,7 @@ enum CoderAPIError: LocalizedError {
case missingModel(String)
case invalidResponse
case serviceError(Int, String)
case audioPreparationFailed(String, String)
var errorDescription: String? {
switch self {
@@ -43,6 +44,8 @@ enum CoderAPIError: LocalizedError {
return "Coder returned an invalid response."
case .serviceError(let status, let message):
return "Coder request failed (\(status)): \(message)"
case .audioPreparationFailed(let filename, let message):
return "Could not prepare \(filename) for transcription: \(message)"
}
}
}
@@ -179,7 +182,12 @@ final class CoderAPIClient {
let selectedModel = model.trimmingCharacters(in: .whitespacesAndNewlines)
guard !selectedModel.isEmpty else { throw CoderAPIError.missingModel("transcription") }
let apiKey = try requiredAPIKey(KeychainHelper.shared.getCoderAPIKey() ?? "")
let chunks = try makeAudioChunks(from: fileURL, preserveSpeakerIdentity: diarization)
let chunks: [AudioChunk]
do {
chunks = try makeAudioChunks(from: fileURL, preserveSpeakerIdentity: diarization)
} catch {
throw CoderAPIError.audioPreparationFailed(fileURL.lastPathComponent, error.localizedDescription)
}
defer {
for chunk in chunks where chunk.isTemporary {
try? FileManager.default.removeItem(at: chunk.url)
@@ -260,13 +268,38 @@ final class CoderAPIClient {
let size = attributes[.size] as? NSNumber {
request.setValue(size.stringValue, forHTTPHeaderField: "Content-Length")
}
let (data, response) = try await transcriptionSession.upload(for: request, fromFile: bodyURL)
let (data, response) = try await uploadTranscription(request: request, bodyURL: bodyURL)
try validate(response: response, data: data)
let decoded = try JSONDecoder().decode(TranscriptionResponse.self, from: data)
let segments = decoded.segments ?? segments(from: decoded.words ?? [])
return Transcription(text: decoded.text, segments: segments)
}
private func uploadTranscription(request: URLRequest, bodyURL: URL) async throws -> (Data, URLResponse) {
let retryDelays: [UInt64] = [3, 10]
for attempt in 0...retryDelays.count {
do {
return try await transcriptionSession.upload(for: request, fromFile: bodyURL)
} catch {
guard attempt < retryDelays.count, isRetryableTranscriptionError(error) else {
throw error
}
try await Task.sleep(nanoseconds: retryDelays[attempt] * 1_000_000_000)
}
}
throw CoderAPIError.invalidResponse
}
private func isRetryableTranscriptionError(_ error: Error) -> Bool {
guard let urlError = error as? URLError else { return false }
switch urlError.code {
case .networkConnectionLost, .cannotConnectToHost, .timedOut:
return true
default:
return false
}
}
private func makeAudioChunks(from fileURL: URL, preserveSpeakerIdentity: Bool) throws -> [AudioChunk] {
let input = try AVAudioFile(forReading: fileURL)
let format = input.processingFormat
+13 -13
View File
@@ -182,13 +182,13 @@ private enum LocalAPIRouter {
}
private enum LocalRecordingError: LocalizedError {
case processing
case stopping
case saveFailed
var errorDescription: String? {
switch self {
case .processing:
return "The previous meeting is still processing."
case .stopping:
return "The current recording is still stopping."
case .saveFailed:
return "Could not create a meeting for this recording."
}
@@ -208,12 +208,12 @@ private final class LocalRecordingController {
func statusPayload() -> [String: Any] {
let state: String
if isStopping || recordingManager.isProcessing {
state = "processing"
} else if recordingManager.isRecording {
if recordingManager.isRecording {
state = "recording"
} else if recordingManager.activeMeetingId != nil {
state = "starting"
} else if isStopping || recordingManager.isProcessing {
state = "processing"
} else {
state = "idle"
}
@@ -245,8 +245,8 @@ private final class LocalRecordingController {
isStopping = false
return statusPayload()
}
if isStopping || recordingManager.isProcessing {
throw LocalRecordingError.processing
if isStopping {
throw LocalRecordingError.stopping
}
if recordingManager.activeMeetingId != nil {
return statusPayload()
@@ -295,15 +295,16 @@ private final class LocalRecordingController {
private func finishRecording() async {
pendingStopTask = nil
let meetingID = recordingManager.activeMeetingId
let chunks = await recordingManager.stopRecording()
isStopping = false
let completion = await recordingManager.stopRecording()
guard let meetingID,
var meeting = LocalStorageManager.shared.loadMeetings().first(where: { $0.id == meetingID }) else {
isStopping = false
return
}
meeting.transcriptChunks = chunks
meeting.recoveryAudioFolderName = recordingManager.lastRecoveryAudioFolderName
meeting.transcriptChunks = completion.chunks
meeting.recoveryAudioFolderName = completion.recoveryAudioFolderName
meeting.transcriptionError = completion.transcriptionError
let templates = LocalStorageManager.shared.loadTemplates()
if meeting.templateId == nil {
meeting.templateId = LocalStorageManager.shared.preferredTemplateID(in: templates)
@@ -352,7 +353,6 @@ private final class LocalRecordingController {
NotificationCenter.default.post(name: .meetingSaved, object: meeting)
}
}
isStopping = false
}
}
+21 -7
View File
@@ -47,7 +47,7 @@ class MeetingViewModel: ObservableObject {
var isProcessing: Bool {
return isRetryingTranscription ||
(recordingSessionManager.isProcessing && recordingSessionManager.activeMeetingId == meeting.id)
recordingSessionManager.isProcessingMeeting(meeting.id)
}
var canRetryTranscription: Bool {
@@ -137,18 +137,27 @@ class MeetingViewModel: ObservableObject {
self.meeting.transcriptChunks = recordingSessionManager.getTranscriptChunks(for: meeting.id)
}
// Listen for final transcript updates for this meeting.
// Listen for transcript updates emitted while this meeting is recording.
recordingSessionManager.$activeRecordingTranscriptChunksUpdated
.dropFirst()
.sink { [weak self] updatedChunks in
guard let self = self else { return }
// Only update if this meeting is the active recording
if recordingSessionManager.isRecordingMeeting(self.meeting.id) {
if recordingSessionManager.activeMeetingId == self.meeting.id {
self.meeting.transcriptChunks = updatedChunks
}
}
.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)
// Auto-save when meeting properties change
@@ -212,9 +221,11 @@ class MeetingViewModel: ObservableObject {
func stopRecording() {
isStartingRecording = true
Task {
let chunks = await recordingSessionManager.stopRecording()
meeting.transcriptChunks = chunks
meeting.recoveryAudioFolderName = recordingSessionManager.lastRecoveryAudioFolderName
let completion = await recordingSessionManager.stopRecording()
meeting.transcriptChunks = completion.chunks
meeting.recoveryAudioFolderName = completion.recoveryAudioFolderName
meeting.transcriptionError = completion.transcriptionError
errorMessage = completion.transcriptionError
refreshRecoveryAudioFolder()
saveMeeting()
if !meeting.formattedTranscript.isEmpty {
@@ -238,6 +249,7 @@ class MeetingViewModel: ObservableObject {
)
meeting.transcriptChunks = chunks
meeting.recoveryAudioFolderName = recoveryAudioFolderURL.lastPathComponent
meeting.transcriptionError = nil
selectedTab = .transcript
guard saveMeeting() else {
@@ -247,6 +259,8 @@ class MeetingViewModel: ObservableObject {
await generateNotes()
} catch {
errorMessage = error.localizedDescription
meeting.transcriptionError = error.localizedDescription
_ = saveMeeting()
print("Retry transcription failed: \(error)")
}
}
+91 -14
View File
@@ -6,6 +6,8 @@ struct MeetingListView: View {
@StateObject private var recordingSessionManager = RecordingSessionManager.shared
@State private var selectedMeeting: Meeting?
@State private var navigationPath = NavigationPath()
@State private var recordingFailureMessage: String?
@State private var failedMeetingID: UUID?
var body: some View {
NavigationSplitView {
@@ -23,6 +25,38 @@ struct MeetingListView: View {
.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 {
@@ -161,6 +195,14 @@ struct MeetingListView: View {
return DayGroup(day: dayString, date: date, meetings: meetings.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 {
@@ -182,7 +224,14 @@ struct MeetingRowView: View {
.foregroundColor(.red)
.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)
.lineLimit(1)
}
@@ -542,24 +591,52 @@ struct MeetingDetailContentView: View {
}
private var transcriptView: some View {
ScrollView {
if viewModel.meeting.collapsedTranscriptChunks.isEmpty {
Text("Transcript will appear here...")
.frame(maxWidth: .infinity, alignment: .leading)
.padding()
.foregroundColor(.secondary)
} else {
LazyVStack(alignment: .leading, spacing: 4) {
ForEach(viewModel.meeting.collapsedTranscriptChunks) { chunk in
CollapsedTranscriptChunkView(chunk: chunk)
VStack(alignment: .leading, spacing: 10) {
if let transcriptionError = viewModel.meeting.transcriptionError {
HStack(alignment: .top, spacing: 10) {
Image(systemName: "exclamationmark.triangle.fill")
.foregroundColor(.orange)
VStack(alignment: .leading, spacing: 4) {
Text("Transcription failed")
.font(.headline)
Text(transcriptionError)
.font(.caption)
.foregroundColor(.secondary)
}
Spacer()
if viewModel.recoveryAudioFolderURL != nil {
Button("Retry") {
viewModel.retryTranscription()
}
.disabled(!viewModel.canRetryTranscription)
}
}
.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 {