Compare commits

...
6 Commits
12 changed files with 460 additions and 82 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 = 32; CURRENT_PROJECT_VERSION = 38;
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 = 1.1.20; MARKETING_VERSION = 1.1.26;
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 = 32; CURRENT_PROJECT_VERSION = 38;
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 = 1.1.20; MARKETING_VERSION = 1.1.26;
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;
+2 -1
View File
@@ -22,7 +22,8 @@ Implemented:
- Meeting search functionality - Meeting search functionality
- Abilty to edit system prompt - Abilty to edit system prompt
- Select any compatible Coder model for transcription and note generation - Select any compatible Coder model for transcription and note generation
- Automatic start and stop from MuteDeck through a compatible local API - Automatic start and stop from MuteDeck through a compatible local API, with a 10-second reconnect grace period
- Merge an automatically split continuation back into its previous meeting
- Auto updates - Auto updates
- Text formatting - Text formatting
- Different note templates - Different note templates
+194 -18
View File
@@ -86,9 +86,22 @@ final class AudioManager: NSObject, ObservableObject {
isProcessing = false 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 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 (micTranscription, systemTranscription) = await (micResult, systemResult)
let results = [micTranscription, systemTranscription] let results = [micTranscription, systemTranscription]
@@ -98,9 +111,6 @@ final class AudioManager: NSObject, ObservableObject {
existingChunks: transcriptChunks.filter(\.isFinal) existingChunks: transcriptChunks.filter(\.isFinal)
) )
transcriptChunks = updated transcriptChunks = updated
let completedFiles = files.compactMap { $0 }
let audioFolder = preserveAudioFiles(completedFiles, meetingID: completedMeetingID)
lastRecoveryAudioFolderName = audioFolder?.lastPathComponent
if !failures.isEmpty { if !failures.isEmpty {
let retentionDays = UserDefaultsManager.shared.audioRetentionDays let retentionDays = UserDefaultsManager.shared.audioRetentionDays
let retentionUnit = retentionDays == 1 ? "day" : "days" let retentionUnit = retentionDays == 1 ? "day" : "days"
@@ -123,17 +133,29 @@ final class AudioManager: NSObject, ObservableObject {
isProcessing = true isProcessing = true
defer { isProcessing = false } defer { isProcessing = false }
let model = UserDefaultsManager.shared.transcriptionModel let model = UserDefaultsManager.shared.transcriptionModel
let micURL = recoveryFiles.first(where: { $0.source == .mic })?.url let sessions = Dictionary(grouping: recoveryFiles) { recoverySessionKey(for: $0.url) }
let systemURL = recoveryFiles.first(where: { $0.source == .system })?.url .values
.map { files in
(files: files, startedAt: recoveryCaptureStartedAt(for: files, fallback: captureStartedAt))
}
.sorted { $0.startedAt < $1.startedAt }
var chunks: [TranscriptChunk] = []
var failures: [String] = []
for session in sessions {
let micURL = session.files.first(where: { $0.source == .mic })?.url
let systemURL = session.files.first(where: { $0.source == .system })?.url
async let micResult = transcribe(micURL, model: model, diarization: false) async let micResult = transcribe(micURL, model: model, diarization: false)
async let systemResult = transcribe(systemURL, model: model, diarization: true) async let systemResult = transcribe(systemURL, model: model, diarization: true)
let (micTranscription, systemTranscription) = await (micResult, systemResult) let (micTranscription, systemTranscription) = await (micResult, systemResult)
let results = [micTranscription, systemTranscription] let result = buildTranscriptChunks(
let (chunks, failures) = buildTranscriptChunks( from: [micTranscription, systemTranscription],
from: results, captureStartedAt: session.startedAt,
captureStartedAt: captureStartedAt, existingChunks: chunks
existingChunks: []
) )
chunks = result.0
failures.append(contentsOf: result.1)
}
if !failures.isEmpty { if !failures.isEmpty {
throw RecoveryTranscriptionError.requestFailed(failures.joined(separator: "; ")) throw RecoveryTranscriptionError.requestFailed(failures.joined(separator: "; "))
@@ -144,6 +166,28 @@ final class AudioManager: NSObject, ObservableObject {
return chunks return chunks
} }
private func recoverySessionKey(for url: URL) -> String {
let name = url.deletingPathExtension().lastPathComponent
if name.hasSuffix("-mic") { return String(name.dropLast(4)) }
if name.hasSuffix("-system") { return String(name.dropLast(7)) }
return name
}
private func recoveryCaptureStartedAt(
for files: [(url: URL, source: AudioSource)],
fallback: Date
) -> Date {
let estimatedStarts = files.compactMap { file -> Date? in
guard let duration = audioDuration(at: file.url),
let values = try? file.url.resourceValues(forKeys: [.contentModificationDateKey, .creationDateKey]),
let finishedAt = values.contentModificationDate ?? values.creationDate else {
return nil
}
return finishedAt.addingTimeInterval(-duration)
}
return estimatedStarts.min() ?? fallback
}
func cancelRecording() { func cancelRecording() {
cancelCapture(removeFiles: true) cancelCapture(removeFiles: true)
lastRecoveryAudioFolderName = nil lastRecoveryAudioFolderName = nil
@@ -209,15 +253,18 @@ final class AudioManager: NSObject, ObservableObject {
private func prepareAudioFiles() throws { private func prepareAudioFiles() throws {
let settings: [String: Any] = [ let settings: [String: Any] = [
AVFormatIDKey: kAudioFormatMPEG4AAC, AVFormatIDKey: kAudioFormatLinearPCM,
AVSampleRateKey: 16_000, AVSampleRateKey: 16_000,
AVNumberOfChannelsKey: 1, AVNumberOfChannelsKey: 1,
AVEncoderBitRateKey: 48_000 AVLinearPCMBitDepthKey: 16,
AVLinearPCMIsFloatKey: false,
AVLinearPCMIsBigEndianKey: false,
AVLinearPCMIsNonInterleaved: false
] ]
let base = FileManager.default.temporaryDirectory let base = FileManager.default.temporaryDirectory
let id = sessionID.uuidString 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( let newMicAudioFile = try AVAudioFile(
forWriting: micURL, forWriting: micURL,
settings: settings, settings: settings,
@@ -353,13 +400,25 @@ final class AudioManager: NSObject, ObservableObject {
private func startTapIO(_ tap: ProcessTap) throws { private func startTapIO(_ tap: ProcessTap) throws {
guard var description = tap.tapStreamDescription, guard var description = tap.tapStreamDescription,
let inputFormat = AVAudioFormat(streamDescription: &description), let advertisedInputFormat = AVAudioFormat(streamDescription: &description),
let targetFormat = systemAudioFile?.processingFormat, 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"]) 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 try tap.run(on: tapQueue) { [weak self] _, inputData, _, _, _ in
guard let self else { return } 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 // The tap queue is serial. Reusing the converter preserves its
// resampler state instead of discarding audio at every callback. // resampler state instead of discarding audio at every callback.
self.processAudioBuffer( self.processAudioBuffer(
@@ -374,6 +433,30 @@ 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( private func copyAudioBuffer(
from inputData: UnsafePointer<AudioBufferList>, from inputData: UnsafePointer<AudioBufferList>,
format: AVAudioFormat format: AVAudioFormat
@@ -521,6 +604,99 @@ final class AudioManager: NSObject, ObservableObject {
LocalStorageManager.shared.preserveAudioFiles(urls, for: meetingID) 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() { private func resetAudioLevels() {
micAudioLevel = 0 micAudioLevel = 0
systemAudioLevel = 0 systemAudioLevel = 0
+96 -42
View File
@@ -152,6 +152,98 @@ class LocalStorageManager {
} }
} }
func mergeMeeting(_ continuation: Meeting, into previous: Meeting) -> Meeting? {
guard continuation.id != previous.id, continuation.date >= previous.date else { return nil }
let previousFolder = recoveryDirectory.appendingPathComponent(previous.id.uuidString, isDirectory: true)
let continuationFolder = recoveryAudioFolder(named: continuation.id.uuidString)
var copiedAudioURLs: [URL] = []
var createdPreviousFolder = false
if let continuationFolder {
if !FileManager.default.fileExists(atPath: previousFolder.path) {
do {
try FileManager.default.createDirectory(at: previousFolder, withIntermediateDirectories: true)
createdPreviousFolder = true
} catch {
return nil
}
}
for audioFile in recoveryAudioFiles(in: continuationFolder) {
let destination = previousFolder.appendingPathComponent(audioFile.url.lastPathComponent)
if FileManager.default.fileExists(atPath: destination.path) {
guard FileManager.default.contentsEqual(
atPath: audioFile.url.path,
andPath: destination.path
) else {
rollbackMergedAudio(copiedAudioURLs, removeFolder: createdPreviousFolder ? previousFolder : nil)
return nil
}
continue
}
do {
try FileManager.default.copyItem(at: audioFile.url, to: destination)
copiedAudioURLs.append(destination)
} catch {
rollbackMergedAudio(copiedAudioURLs, removeFolder: createdPreviousFolder ? previousFolder : nil)
return nil
}
}
}
var merged = previous
var seenChunkIDs = Set<UUID>()
merged.transcriptChunks = (previous.transcriptChunks + continuation.transcriptChunks)
.filter { seenChunkIDs.insert($0.id).inserted }
.sorted {
if $0.timestamp == $1.timestamp {
return $0.id.uuidString < $1.id.uuidString
}
return $0.timestamp < $1.timestamp
}
merged.userNotes = mergedText(previous.userNotes, continuation.userNotes)
merged.generatedNotes = mergedText(previous.generatedNotes, continuation.generatedNotes)
merged.templateId = previous.templateId ?? continuation.templateId
if !recoveryAudioFiles(in: previousFolder).isEmpty {
merged.recoveryAudioFolderName = previous.id.uuidString
}
guard saveMeeting(merged) else {
rollbackMergedAudio(copiedAudioURLs, removeFolder: createdPreviousFolder ? previousFolder : nil)
return nil
}
guard deleteMeeting(continuation) else {
_ = saveMeeting(previous)
rollbackMergedAudio(copiedAudioURLs, removeFolder: createdPreviousFolder ? previousFolder : nil)
return nil
}
if let continuationFolder {
try? FileManager.default.removeItem(at: continuationFolder)
}
return merged
}
private func mergedText(_ first: String, _ second: String) -> String {
let first = first.trimmingCharacters(in: .whitespacesAndNewlines)
let second = second.trimmingCharacters(in: .whitespacesAndNewlines)
guard !first.isEmpty else { return second }
guard !second.isEmpty, second != first else { return first }
return first + "\n\n---\n\n" + second
}
private func rollbackMergedAudio(_ copiedURLs: [URL], removeFolder folder: URL?) {
for url in copiedURLs {
try? FileManager.default.removeItem(at: url)
}
if let folder, recoveryAudioFiles(in: folder).isEmpty {
try? FileManager.default.removeItem(at: folder)
}
}
// MARK: - Recovery Audio // MARK: - Recovery Audio
func preserveAudioFiles(_ urls: [URL], for meetingID: UUID) -> URL? { func preserveAudioFiles(_ urls: [URL], for meetingID: UUID) -> URL? {
@@ -256,50 +348,12 @@ class LocalStorageManager {
} }
func findRecoveryAudioFolder(for meeting: Meeting) -> URL? { 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 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) { func deleteRecoveryAudioFolder(_ folder: URL) {
@@ -124,6 +124,9 @@ class RecordingSessionManager: ObservableObject {
// Find and update the active meeting // Find and update the active meeting
if let index = meetings.firstIndex(where: { $0.id == meetingId }) { if let index = meetings.firstIndex(where: { $0.id == meetingId }) {
meetings[index].transcriptChunks = chunks meetings[index].transcriptChunks = chunks
if let recoveryAudioFolderName = lastRecoveryAudioFolderName {
meetings[index].recoveryAudioFolderName = recoveryAudioFolderName
}
// Save the updated meeting // Save the updated meeting
let success = LocalStorageManager.shared.saveMeeting(meetings[index]) let success = LocalStorageManager.shared.saveMeeting(meetings[index])
@@ -116,6 +116,47 @@ extension AudioObjectID {
try read(kAudioTapPropertyFormat, defaultValue: AudioStreamBasicDescription()) try read(kAudioTapPropertyFormat, defaultValue: AudioStreamBasicDescription())
} }
/// Reads the virtual format of the first input stream exposed by this device.
///
/// An aggregate device can adapt a tap to the active output hardware. Its
/// input stream format is therefore the format delivered to the I/O proc,
/// which can differ from the tap object's originally advertised format.
func readInputStreamBasicDescription() throws -> AudioStreamBasicDescription {
var streamsAddress = AudioObjectPropertyAddress(
mSelector: kAudioDevicePropertyStreams,
mScope: kAudioObjectPropertyScopeGlobal,
mElement: kAudioObjectPropertyElementMain
)
var dataSize: UInt32 = 0
var status = AudioObjectGetPropertyDataSize(self, &streamsAddress, 0, nil, &dataSize)
guard status == noErr else {
throw "Error reading device stream list size: \(status)"
}
var streamIDs = [AudioObjectID](
repeating: .unknown,
count: Int(dataSize) / MemoryLayout<AudioObjectID>.size
)
status = AudioObjectGetPropertyData(self, &streamsAddress, 0, nil, &dataSize, &streamIDs)
guard status == noErr else {
throw "Error reading device stream list: \(status)"
}
for streamID in streamIDs {
let direction: UInt32 = try streamID.read(
kAudioStreamPropertyDirection,
defaultValue: 0
)
guard direction == 1 else { continue }
return try streamID.read(
kAudioStreamPropertyVirtualFormat,
defaultValue: AudioStreamBasicDescription()
)
}
throw "Device has no input stream."
}
private func requireSystemObject() throws { private func requireSystemObject() throws {
if self != .system { throw "Only supported for the system object." } if self != .system { throw "Only supported for the system object." }
} }
+14 -2
View File
@@ -248,8 +248,20 @@ 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)...")
self.tapStreamDescription = try tapID.readAudioTapStreamBasicDescription() let advertisedDescription = try tapID.readAudioTapStreamBasicDescription()
logger.debug("Successfully read tap stream description: \(String(describing: self.tapStreamDescription))")
// The aggregate device may adapt the tap to the active hardware's
// sample rate. Its input stream is the format actually delivered
// to the I/O proc, so use that rather than the tap's pre-aggregate
// 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()
logger.info("Using aggregate input stream description: \(String(describing: self.tapStreamDescription), privacy: .public); tap advertised: \(String(describing: advertisedDescription), privacy: .public)")
} catch {
self.tapStreamDescription = advertisedDescription
logger.warning("Could not read aggregate input stream format; using tap format: \(error, privacy: .public)")
}
} catch { } catch {
logger.error("Failed to read audio tap stream basic description for tapID #\(tapID): \(error)") logger.error("Failed to read audio tap stream basic description for tapID #\(tapID): \(error)")
throw error // Propagate error throw error // Propagate error
@@ -275,6 +275,13 @@ final class CoderAPIClient {
let framesPerChunk = preserveSpeakerIdentity let framesPerChunk = preserveSpeakerIdentity
? max(1, input.length) ? max(1, input.length)
: AVAudioFramePosition(format.sampleRate * transcriptionChunkDuration) : 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 chunks: [AudioChunk] = []
var frameOffset: AVAudioFramePosition = 0 var frameOffset: AVAudioFramePosition = 0
+19 -2
View File
@@ -200,7 +200,9 @@ private final class LocalRecordingController {
static let shared = LocalRecordingController() static let shared = LocalRecordingController()
private let recordingManager = RecordingSessionManager.shared private let recordingManager = RecordingSessionManager.shared
private let stopGraceNanoseconds: UInt64 = 10_000_000_000
private var isStopping = false private var isStopping = false
private var pendingStopTask: Task<Void, Never>?
private init() {} private init() {}
@@ -237,6 +239,12 @@ private final class LocalRecordingController {
} }
func startRecording() throws -> [String: Any] { func startRecording() throws -> [String: Any] {
if let pendingStopTask {
pendingStopTask.cancel()
self.pendingStopTask = nil
isStopping = false
return statusPayload()
}
if isStopping || recordingManager.isProcessing { if isStopping || recordingManager.isProcessing {
throw LocalRecordingError.processing throw LocalRecordingError.processing
} }
@@ -258,14 +266,22 @@ private final class LocalRecordingController {
return statusPayload() return statusPayload()
} }
isStopping = true isStopping = true
Task { [weak self] in pendingStopTask = Task { [weak self] in
try? await Task.sleep(nanoseconds: self?.stopGraceNanoseconds ?? 0)
guard !Task.isCancelled else { return }
await self?.finishRecording() await self?.finishRecording()
} }
return statusPayload() return statusPayload()
} }
func cancelRecording() -> [String: Any] { func cancelRecording() -> [String: Any] {
guard !isStopping else { return statusPayload() } if let pendingStopTask {
pendingStopTask.cancel()
self.pendingStopTask = nil
isStopping = false
} else if isStopping {
return statusPayload()
}
let meetingID = recordingManager.activeMeetingId let meetingID = recordingManager.activeMeetingId
recordingManager.cancelRecording() recordingManager.cancelRecording()
if let meetingID, if let meetingID,
@@ -277,6 +293,7 @@ private final class LocalRecordingController {
} }
private func finishRecording() async { private func finishRecording() async {
pendingStopTask = nil
let meetingID = recordingManager.activeMeetingId let meetingID = recordingManager.activeMeetingId
let chunks = await recordingManager.stopRecording() let chunks = await recordingManager.stopRecording()
guard let meetingID, guard let meetingID,
@@ -67,6 +67,28 @@ class MeetingListViewModel: ObservableObject {
_ = LocalStorageManager.shared.deleteMeeting(meeting) _ = LocalStorageManager.shared.deleteMeeting(meeting)
} }
func previousMeeting(for meeting: Meeting) -> Meeting? {
meetings
.filter { $0.id != meeting.id && $0.date < meeting.date }
.max { $0.date < $1.date }
}
func mergeIntoPrevious(_ meeting: Meeting) -> Meeting? {
guard let previous = previousMeeting(for: meeting),
let merged = LocalStorageManager.shared.mergeMeeting(meeting, into: previous) else {
errorMessage = "The meetings could not be merged. Their original records and audio were kept."
return nil
}
meetings.removeAll { $0.id == meeting.id || $0.id == previous.id }
meetings.append(merged)
meetings.sort { $0.date > $1.date }
NotificationCenter.default.post(name: .meetingSaved, object: merged)
NotificationCenter.default.post(name: .meetingDeleted, object: meeting)
PostHogSDK.shared.capture("meetings_merged")
return merged
}
func createNewMeeting() -> Meeting { func createNewMeeting() -> Meeting {
let newMeeting = Meeting(templateId: LocalStorageManager.shared.preferredTemplateID()) let newMeeting = Meeting(templateId: LocalStorageManager.shared.preferredTemplateID())
meetings.insert(newMeeting, at: 0) meetings.insert(newMeeting, at: 0)
@@ -254,9 +254,7 @@ class MeetingViewModel: ObservableObject {
private func refreshRecoveryAudioFolder() { private func refreshRecoveryAudioFolder() {
recoveryAudioFolderURL = LocalStorageManager.shared.findRecoveryAudioFolder(for: meeting) recoveryAudioFolderURL = LocalStorageManager.shared.findRecoveryAudioFolder(for: meeting)
if let recoveryAudioFolderURL { meeting.recoveryAudioFolderName = recoveryAudioFolderURL?.lastPathComponent
meeting.recoveryAudioFolderName = recoveryAudioFolderURL.lastPathComponent
}
} }
func showAudioInFinder() { func showAudioInFinder() {
+50 -3
View File
@@ -83,10 +83,19 @@ struct MeetingListView: View {
NavigationStack(path: $navigationPath) { NavigationStack(path: $navigationPath) {
Group { Group {
if let selectedMeeting = selectedMeeting { if let selectedMeeting = selectedMeeting {
MeetingDetailContentView(meeting: selectedMeeting, onDelete: { MeetingDetailContentView(
meeting: selectedMeeting,
mergeCandidate: viewModel.previousMeeting(for: selectedMeeting),
onMerge: { meeting in
guard let merged = viewModel.mergeIntoPrevious(meeting) else { return false }
self.selectedMeeting = merged
return true
},
onDelete: {
// When a meeting is deleted from the detail view, clear the selection // When a meeting is deleted from the detail view, clear the selection
self.selectedMeeting = nil self.selectedMeeting = nil
}) }
)
.id(selectedMeeting.id) // Force recreation when selection changes .id(selectedMeeting.id) // Force recreation when selection changes
} else { } else {
ContentUnavailableView( ContentUnavailableView(
@@ -229,12 +238,22 @@ struct MeetingDetailContentView: View {
@StateObject private var viewModel: MeetingViewModel @StateObject private var viewModel: MeetingViewModel
@StateObject private var recordingSessionManager = RecordingSessionManager.shared @StateObject private var recordingSessionManager = RecordingSessionManager.shared
@State private var showDeleteAlert = false @State private var showDeleteAlert = false
@State private var showMergeAlert = false
@State private var isEditing = false @State private var isEditing = false
@State private var showCopyConfirmation = false @State private var showCopyConfirmation = false
let mergeCandidate: Meeting?
let onMerge: (Meeting) -> Bool
let onDelete: () -> Void let onDelete: () -> Void
init(meeting: Meeting, onDelete: @escaping () -> Void) { init(
meeting: Meeting,
mergeCandidate: Meeting?,
onMerge: @escaping (Meeting) -> Bool,
onDelete: @escaping () -> Void
) {
self._viewModel = StateObject(wrappedValue: MeetingViewModel(meeting: meeting)) self._viewModel = StateObject(wrappedValue: MeetingViewModel(meeting: meeting))
self.mergeCandidate = mergeCandidate
self.onMerge = onMerge
self.onDelete = onDelete self.onDelete = onDelete
} }
@@ -276,6 +295,17 @@ struct MeetingDetailContentView: View {
Divider() Divider()
} }
if mergeCandidate != nil {
Button {
showMergeAlert = true
} label: {
Label("Merge into Previous Meeting", systemImage: "arrow.triangle.merge")
}
.disabled(recordingSessionManager.isRecording || recordingSessionManager.isProcessing)
Divider()
}
Button("Delete Meeting", role: .destructive) { Button("Delete Meeting", role: .destructive) {
showDeleteAlert = true showDeleteAlert = true
} }
@@ -461,6 +491,23 @@ struct MeetingDetailContentView: View {
} message: { } message: {
Text("Are you sure you want to delete this meeting? This action cannot be undone.") Text("Are you sure you want to delete this meeting? This action cannot be undone.")
} }
.alert("Merge into Previous Meeting?", isPresented: $showMergeAlert) {
Button("Merge", role: .destructive) {
// Prevent the disappearing detail view from auto-saving the
// continuation after the storage layer removes it.
viewModel.isDeleted = true
if !onMerge(viewModel.meeting) {
viewModel.isDeleted = false
viewModel.errorMessage = "The meetings could not be merged. Their original records and audio were kept."
}
}
Button("Cancel", role: .cancel) { }
} message: {
let previousTitle = mergeCandidate?.title.isEmpty == false
? mergeCandidate?.title ?? "the previous meeting"
: "the previous meeting"
Text("This combines this meeting's transcript, notes, and saved audio into \"\(previousTitle)\", keeps the earlier meeting's title and time, then removes this continuation.")
}
.onDisappear { .onDisappear {
// A failed recording may still be empty. Keep it until the user // A failed recording may still be empty. Keep it until the user
// explicitly deletes it so app updates cannot erase history. // explicitly deletes it so app updates cannot erase history.