Compare commits

...
8 Commits
16 changed files with 487 additions and 79 deletions
+17 -11
View File
@@ -20,6 +20,9 @@ jobs:
runs-on: macos-15
env:
VERSION: ${{ inputs.version }}
APPLE_ID: ${{ secrets.APPLE_ID }}
APPLE_TEAM_ID: ${{ secrets.APPLE_TEAM_ID }}
APPLE_APP_PASSWORD: ${{ secrets.APPLE_APP_PASSWORD }}
SPARKLE_PRIVATE_KEY: ${{ secrets.SPARKLE_PRIVATE_KEY }}
steps:
- uses: actions/checkout@v7
@@ -31,7 +34,7 @@ jobs:
APPLE_CERTIFICATE_P12: ${{ secrets.APPLE_CERTIFICATE_P12 }}
APPLE_CERTIFICATE_PASSWORD: ${{ secrets.APPLE_CERTIFICATE_PASSWORD }}
run: |
for variable in APPLE_CERTIFICATE_P12 APPLE_CERTIFICATE_PASSWORD SPARKLE_PRIVATE_KEY; do
for variable in APPLE_CERTIFICATE_P12 APPLE_CERTIFICATE_PASSWORD APPLE_ID APPLE_TEAM_ID APPLE_APP_PASSWORD SPARKLE_PRIVATE_KEY; do
if [[ -z "${!variable:-}" ]]; then
echo "Missing GitHub Actions secret: $variable" >&2
exit 1
@@ -53,7 +56,7 @@ jobs:
fi
echo "SIGNING_IDENTITY=$signing_identity" >> "$GITHUB_ENV"
- name: Build and sign release
- name: Build, sign, and notarize release
timeout-minutes: 30
run: scripts/package_release.sh
@@ -70,13 +73,16 @@ jobs:
GH_TOKEN: ${{ github.token }}
run: |
tag="v$VERSION"
if git rev-parse "$tag" >/dev/null 2>&1; then
echo "Tag already exists: $tag" >&2
exit 1
if gh release view "$tag" >/dev/null 2>&1; then
gh release upload "$tag" \
"$RUNNER_TEMP/meetingnotes-release/release/Meetingnotes-$VERSION.zip" \
"$RUNNER_TEMP/meetingnotes-release/release/appcast.xml" \
--clobber
else
gh release create "$tag" \
"$RUNNER_TEMP/meetingnotes-release/release/Meetingnotes-$VERSION.zip" \
"$RUNNER_TEMP/meetingnotes-release/release/appcast.xml" \
--target "$GITHUB_SHA" \
--title "Meetingnotes $VERSION" \
--generate-notes
fi
gh release create "$tag" \
"$RUNNER_TEMP/meetingnotes-release/release/Meetingnotes-$VERSION.zip" \
"$RUNNER_TEMP/meetingnotes-release/release/appcast.xml" \
--target "$GITHUB_SHA" \
--title "Meetingnotes $VERSION" \
--generate-notes
+4 -4
View File
@@ -276,7 +276,7 @@
CODE_SIGN_IDENTITY = "Apple Development";
CODE_SIGN_STYLE = Automatic;
COMBINE_HIDPI_IMAGES = YES;
CURRENT_PROJECT_VERSION = 25;
CURRENT_PROJECT_VERSION = 32;
DEVELOPMENT_ASSET_PATHS = "\"meetingnotes/Preview Content\"";
DEVELOPMENT_TEAM = G9LVHZAJNX;
ENABLE_HARDENED_RUNTIME = YES;
@@ -290,7 +290,7 @@
"@executable_path/../Frameworks",
);
MACOSX_DEPLOYMENT_TARGET = 15.0;
MARKETING_VERSION = 1.1.13;
MARKETING_VERSION = 1.1.20;
ONLY_ACTIVE_ARCH = NO;
OTHER_SWIFT_FLAGS = "$(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 = 25;
CURRENT_PROJECT_VERSION = 32;
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 = 1.1.13;
MARKETING_VERSION = 1.1.20;
ONLY_ACTIVE_ARCH = YES;
OTHER_SWIFT_FLAGS = "$(inherited) -D ENABLE_TCC_SPI";
PRODUCT_BUNDLE_IDENTIFIER = net.jamesbone.meetingnotes;
+36 -42
View File
@@ -35,6 +35,7 @@ final class AudioManager: NSObject, ObservableObject {
private var audioEngine = AVAudioEngine()
private var sessionID = UUID()
private var meetingID = UUID()
private var processTap: ProcessTap?
private let permission = AudioRecordingPermission()
private let tapQueue = DispatchQueue(label: "io.meetingnotes.audiotap", qos: .userInitiated)
@@ -59,11 +60,12 @@ final class AudioManager: NSObject, ObservableObject {
NotificationCenter.default.removeObserver(self)
}
func startRecording() {
func startRecording(for meetingID: UUID) {
errorMessage = nil
lastRecoveryAudioFolderName = nil
cancelCapture(removeFiles: true)
sessionID = UUID()
self.meetingID = meetingID
recordingStartedAt = Date()
do {
try prepareAudioFiles()
@@ -76,7 +78,7 @@ final class AudioManager: NSObject, ObservableObject {
}
func stopRecordingAndTranscribe() async -> [TranscriptChunk] {
let completedSessionID = sessionID
let completedMeetingID = meetingID
let captureStartedAt = recordingStartedAt
let files = stopCaptureAndCloseFiles()
isProcessing = true
@@ -85,8 +87,8 @@ final class AudioManager: NSObject, ObservableObject {
}
let model = UserDefaultsManager.shared.transcriptionModel
async let micResult = transcribe(files[0], model: model)
async let systemResult = transcribe(files[1], model: model)
async let micResult = transcribe(files[0], model: model, diarization: false)
async let systemResult = transcribe(files[1], model: model, diarization: true)
let (micTranscription, systemTranscription) = await (micResult, systemResult)
let results = [micTranscription, systemTranscription]
@@ -97,16 +99,17 @@ final class AudioManager: NSObject, ObservableObject {
)
transcriptChunks = updated
let completedFiles = files.compactMap { $0 }
if failures.isEmpty {
lastRecoveryAudioFolderName = nil
removeAudioFiles(completedFiles)
} else {
let recoveryFolder = preserveAudioFiles(completedFiles, sessionID: completedSessionID)
lastRecoveryAudioFolderName = recoveryFolder?.lastPathComponent
let recoveryMessage = recoveryFolder == nil
let audioFolder = preserveAudioFiles(completedFiles, meetingID: completedMeetingID)
lastRecoveryAudioFolderName = audioFolder?.lastPathComponent
if !failures.isEmpty {
let retentionDays = UserDefaultsManager.shared.audioRetentionDays
let retentionUnit = retentionDays == 1 ? "day" : "days"
let recoveryMessage = audioFolder == nil
? " The audio remains in the app's temporary folder."
: " Audio was saved in Documents/Meetingnotes-Recovery/\(completedSessionID.uuidString)."
: " Audio was kept for \(retentionDays) \(retentionUnit). Use Show Audio Folder in the Meetingnotes menu to find it."
errorMessage = "Transcription failed for " + failures.joined(separator: "; ") + recoveryMessage
} else if audioFolder == nil, !completedFiles.isEmpty {
errorMessage = "The transcript completed, but Meetingnotes could not move the audio into its retention folder."
}
return updated
}
@@ -122,8 +125,8 @@ final class AudioManager: NSObject, ObservableObject {
let model = UserDefaultsManager.shared.transcriptionModel
let micURL = recoveryFiles.first(where: { $0.source == .mic })?.url
let systemURL = recoveryFiles.first(where: { $0.source == .system })?.url
async let micResult = transcribe(micURL, model: model)
async let systemResult = transcribe(systemURL, model: model)
async let micResult = transcribe(micURL, model: model, diarization: false)
async let systemResult = transcribe(systemURL, model: model, diarization: true)
let (micTranscription, systemTranscription) = await (micResult, systemResult)
let results = [micTranscription, systemTranscription]
let (chunks, failures) = buildTranscriptChunks(
@@ -146,10 +149,19 @@ final class AudioManager: NSObject, ObservableObject {
lastRecoveryAudioFolderName = nil
}
private func transcribe(_ fileURL: URL?, model: String) async -> Result<CoderAPIClient.Transcription, Error>? {
private func transcribe(
_ fileURL: URL?,
model: String,
diarization: Bool
) async -> Result<CoderAPIClient.Transcription, Error>? {
guard let fileURL else { return nil }
do {
return .success(try await CoderAPIClient.shared.transcribe(fileURL: fileURL, model: model))
return .success(try await CoderAPIClient.shared.transcribe(
fileURL: fileURL,
model: model,
diarization: diarization,
maxSpeakerCount: 4
))
} catch {
return .failure(error)
}
@@ -178,6 +190,7 @@ final class AudioManager: NSObject, ObservableObject {
updated.append(TranscriptChunk(
timestamp: captureStartedAt.addingTimeInterval(max(0, segment.start)),
source: source,
speaker: source == .system ? segment.speaker : nil,
text: text,
isFinal: true
))
@@ -341,12 +354,14 @@ final class AudioManager: NSObject, ObservableObject {
private func startTapIO(_ tap: ProcessTap) throws {
guard var description = tap.tapStreamDescription,
let inputFormat = AVAudioFormat(streamDescription: &description),
let targetFormat = systemAudioFile?.processingFormat else {
let targetFormat = systemAudioFile?.processingFormat,
let converter = AVAudioConverter(from: inputFormat, to: targetFormat) else {
throw NSError(domain: "AudioManager", code: -1, userInfo: [NSLocalizedDescriptionKey: "Unsupported system audio format"])
}
try tap.run(on: tapQueue) { [weak self] _, inputData, _, _, _ in
guard let self,
let converter = AVAudioConverter(from: inputFormat, to: targetFormat) else { return }
guard let self else { return }
// The tap queue is serial. Reusing the converter preserves its
// resampler state instead of discarding audio at every callback.
self.processAudioBuffer(
{ self.copyAudioBuffer(from: inputData, format: inputFormat) },
converter: converter,
@@ -502,29 +517,8 @@ final class AudioManager: NSObject, ObservableObject {
for url in urls { try? FileManager.default.removeItem(at: url) }
}
private func preserveAudioFiles(_ urls: [URL], sessionID: UUID) -> URL? {
guard !urls.isEmpty,
let documents = FileManager.default.urls(for: .documentDirectory, in: .userDomainMask).first else {
return nil
}
let folder = documents
.appendingPathComponent("Meetingnotes-Recovery", isDirectory: true)
.appendingPathComponent(sessionID.uuidString, isDirectory: true)
do {
try FileManager.default.createDirectory(at: folder, withIntermediateDirectories: true)
} catch {
return nil
}
var preservedCount = 0
for url in urls {
do {
try FileManager.default.moveItem(at: url, to: folder.appendingPathComponent(url.lastPathComponent))
preservedCount += 1
} catch {
continue
}
}
return preservedCount > 0 ? folder : nil
private func preserveAudioFiles(_ urls: [URL], meetingID: UUID) -> URL? {
LocalStorageManager.shared.preserveAudioFiles(urls, for: meetingID)
}
private func resetAudioLevels() {
@@ -2,6 +2,7 @@
// Handles local storage of meetings and app data
import Foundation
import AppKit
/// Manages local file storage for meetings and app data
class LocalStorageManager {
@@ -28,7 +29,7 @@ class LocalStorageManager {
// Create templates subdirectory
templatesDirectory = documentsDirectory.appendingPathComponent("Templates")
recoveryDirectory = documentsDirectory.appendingPathComponent("Meetingnotes-Recovery")
recoveryDirectory = documentsDirectory.appendingPathComponent("Meetingnotes Audio")
// Ensure directories exist
try? FileManager.default.createDirectory(at: meetingsDirectory,
@@ -37,6 +38,8 @@ class LocalStorageManager {
withIntermediateDirectories: true)
try? FileManager.default.createDirectory(at: recoveryDirectory,
withIntermediateDirectories: true)
migrateLegacyRecoveryAudio()
purgeExpiredAudioFolders()
}
// MARK: - Meeting Management
@@ -151,6 +154,67 @@ class LocalStorageManager {
// MARK: - Recovery Audio
func preserveAudioFiles(_ urls: [URL], for meetingID: UUID) -> URL? {
guard !urls.isEmpty else { return nil }
purgeExpiredAudioFolders()
let folder = recoveryDirectory.appendingPathComponent(meetingID.uuidString, isDirectory: true)
do {
try FileManager.default.createDirectory(at: folder, withIntermediateDirectories: true)
} catch {
return nil
}
var preservedCount = 0
for url in urls {
let destination = folder.appendingPathComponent(url.lastPathComponent)
do {
if FileManager.default.fileExists(atPath: destination.path) {
try FileManager.default.removeItem(at: destination)
}
try FileManager.default.moveItem(at: url, to: destination)
preservedCount += 1
} catch {
continue
}
}
return preservedCount > 0 ? folder : nil
}
func purgeExpiredAudioFolders(now: Date = Date()) {
guard let folders = try? FileManager.default.contentsOfDirectory(
at: recoveryDirectory,
includingPropertiesForKeys: [.isDirectoryKey],
options: [.skipsHiddenFiles]
) else { return }
let retentionInterval = TimeInterval(UserDefaultsManager.shared.audioRetentionDays) * 24 * 60 * 60
let expirationDate = now.addingTimeInterval(-retentionInterval)
for folder in folders {
guard (try? folder.resourceValues(forKeys: [.isDirectoryKey]).isDirectory) == true else { continue }
let audioFiles = recoveryAudioFiles(in: folder)
let newestDate = audioFiles.compactMap { file -> Date? in
let values = try? file.url.resourceValues(forKeys: [.contentModificationDateKey, .creationDateKey])
return values?.contentModificationDate ?? values?.creationDate
}.max()
guard let newestDate else {
try? FileManager.default.removeItem(at: folder)
continue
}
if newestDate < expirationDate {
try? FileManager.default.removeItem(at: folder)
}
}
}
@MainActor
func showAudioFolderInFinder(_ folder: URL? = nil) {
let target = folder ?? recoveryDirectory
try? FileManager.default.createDirectory(at: target, withIntermediateDirectories: true)
NSWorkspace.shared.open(target)
}
func recoveryAudioFolder(named name: String) -> URL? {
let trimmedName = name.trimmingCharacters(in: .whitespacesAndNewlines)
guard !trimmedName.isEmpty,
@@ -245,6 +309,22 @@ class LocalStorageManager {
try? FileManager.default.removeItem(at: folder)
}
private func migrateLegacyRecoveryAudio() {
let legacyDirectory = documentsDirectory.appendingPathComponent("Meetingnotes-Recovery", isDirectory: true)
guard let folders = try? FileManager.default.contentsOfDirectory(
at: legacyDirectory,
includingPropertiesForKeys: [.isDirectoryKey],
options: [.skipsHiddenFiles]
) else { return }
for folder in folders {
let destination = recoveryDirectory.appendingPathComponent(folder.lastPathComponent, isDirectory: true)
guard !FileManager.default.fileExists(atPath: destination.path) else { continue }
try? FileManager.default.moveItem(at: folder, to: destination)
}
try? FileManager.default.removeItem(at: legacyDirectory)
}
/// Imports meeting JSON files from a folder selected by the user.
func importMeetings(from directory: URL) throws -> MeetingImportResult {
let didStartAccess = directory.startAccessingSecurityScopedResource()
@@ -81,7 +81,7 @@ class RecordingSessionManager: ObservableObject {
activeMeetingId = meetingId
recordingStartedAt = Date()
audioManager.startRecording()
audioManager.startRecording(for: meetingId)
}
func stopRecording() async -> [TranscriptChunk] {
@@ -23,6 +23,7 @@ class UserDefaultsManager {
static let transcriptionModel = "transcriptionModel"
static let muteDeckAPIEnabled = "muteDeckAPIEnabled"
static let muteDeckAPIPort = "muteDeckAPIPort"
static let audioRetentionDays = "audioRetentionDays"
}
// MARK: - User Blurb
@@ -78,7 +79,7 @@ class UserDefaultsManager {
}
var transcriptionModel: String {
get { userDefaults.string(forKey: Keys.transcriptionModel) ?? "groq/whisper-large-v3-turbo" }
get { userDefaults.string(forKey: Keys.transcriptionModel) ?? "local-parakeet/parakeet-tdt-0.6b-v3" }
set { userDefaults.set(newValue, forKey: Keys.transcriptionModel) }
}
@@ -94,4 +95,12 @@ class UserDefaultsManager {
}
set { userDefaults.set(newValue, forKey: Keys.muteDeckAPIPort) }
}
var audioRetentionDays: Int {
get {
guard userDefaults.object(forKey: Keys.audioRetentionDays) != nil else { return 3 }
return min(max(userDefaults.integer(forKey: Keys.audioRetentionDays), 1), 365)
}
set { userDefaults.set(min(max(newValue, 1), 365), forKey: Keys.audioRetentionDays) }
}
}
+10
View File
@@ -46,6 +46,10 @@ struct MeetingnotesApp: App {
.commands {
CommandGroup(after: .appInfo) {
CheckForUpdatesView(updater: updaterController.updater)
Divider()
Button("Show Audio Folder") {
LocalStorageManager.shared.showAudioFolderInFinder()
}
}
}
@@ -79,6 +83,12 @@ private struct MeetingnotesMenu: View {
Label("Open Meetingnotes", systemImage: "macwindow")
}
Button {
LocalStorageManager.shared.showAudioFolderInFinder()
} label: {
Label("Show Audio Folder", systemImage: "folder")
}
Divider()
Button {
+21 -4
View File
@@ -48,30 +48,46 @@ struct TranscriptChunk: Codable, Identifiable, Hashable {
let id: UUID
let timestamp: Date
let source: AudioSource
let speaker: Int?
let text: String
let isFinal: Bool
init(id: UUID = UUID(), timestamp: Date = Date(), source: AudioSource, text: String, isFinal: Bool = false) {
init(id: UUID = UUID(), timestamp: Date = Date(), source: AudioSource, speaker: Int? = nil, text: String, isFinal: Bool = false) {
self.id = id
self.timestamp = timestamp
self.source = source
self.speaker = speaker
self.text = text
self.isFinal = isFinal
}
var displayName: String {
if source == .mic { return "Me" }
if let speaker { return "Speaker \(speaker)" }
return source.displayName
}
}
struct CollapsedTranscriptChunk: Identifiable {
let id: UUID
let timestamp: Date
let source: AudioSource
let speaker: Int?
let combinedText: String
init(id: UUID = UUID(), timestamp: Date, source: AudioSource, combinedText: String) {
init(id: UUID = UUID(), timestamp: Date, source: AudioSource, speaker: Int? = nil, combinedText: String) {
self.id = id
self.timestamp = timestamp
self.source = source
self.speaker = speaker
self.combinedText = combinedText
}
var displayName: String {
if source == .mic { return "Me" }
if let speaker { return "Speaker \(speaker)" }
return source.displayName
}
}
struct Meeting: Codable, Identifiable, Hashable {
@@ -115,7 +131,7 @@ struct Meeting: Codable, Identifiable, Hashable {
var transcript: String {
return transcriptChunks
.filter { $0.isFinal }
.map { "[\($0.source.rawValue)] \($0.text)" }
.map { "[\($0.displayName)] \($0.text)" }
.joined(separator: " ")
}
@@ -124,7 +140,7 @@ struct Meeting: Codable, Identifiable, Hashable {
let finalChunks = transcriptChunks.filter { $0.isFinal }
return finalChunks.map { chunk in
"[\(TranscriptTimestampFormatter.string(from: chunk.timestamp))] \(chunk.source.copyPrefix): \(chunk.text)"
"[\(TranscriptTimestampFormatter.string(from: chunk.timestamp))] \(chunk.displayName): \(chunk.text)"
}.joined(separator: "\n")
}
@@ -135,6 +151,7 @@ struct Meeting: Codable, Identifiable, Hashable {
id: chunk.id,
timestamp: chunk.timestamp,
source: chunk.source,
speaker: chunk.speaker,
combinedText: chunk.text
)
}
+5
View File
@@ -54,6 +54,11 @@ struct Settings: Codable {
set { UserDefaultsManager.shared.muteDeckAPIPort = newValue }
}
var audioRetentionDays: Int {
get { UserDefaultsManager.shared.audioRetentionDays }
set { UserDefaultsManager.shared.audioRetentionDays = newValue }
}
// System prompt default loading
static func defaultSystemPrompt() -> String {
guard let path = Bundle.main.path(forResource: "DefaultSystemPrompt", ofType: "txt"),
+5 -2
View File
@@ -138,8 +138,11 @@ final class ProcessTap {
tapDescription = CATapDescription(stereoMixdownOfProcesses: [process.objectID])
logger.debug("Configuring tap for single process objectID: \(process.objectID)")
case .systemAudio:
tapDescription = CATapDescription(monoGlobalTapButExcludeProcesses: [])
logger.debug("Configuring a global system audio tap.")
// 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.")
}
tapDescription.uuid = UUID()
+222 -5
View File
@@ -1,3 +1,4 @@
import AVFoundation
import Foundation
struct CoderModel: Codable, Identifiable, Hashable {
@@ -20,6 +21,7 @@ struct CoderModel: Codable, Identifiable, Hashable {
var supportsChat: Bool { capabilities.isEmpty || capabilities.contains("chat") }
var supportsTranscription: Bool { capabilities.contains("audio_transcription") }
var supportsSpeakerDiarization: Bool { capabilities.contains("speaker_diarization") }
}
enum CoderAPIError: LocalizedError {
@@ -53,6 +55,7 @@ final class CoderAPIClient {
let start: TimeInterval
let end: TimeInterval
let text: String
let speaker: Int?
}
let text: String
@@ -69,10 +72,25 @@ final class CoderAPIClient {
}
private struct TranscriptionResponse: Decodable {
struct Word: Decodable {
let word: String
let start: TimeInterval
let end: TimeInterval
let speaker: Int?
}
let text: String
let segments: [Transcription.Segment]?
let words: [Word]?
}
private struct AudioChunk {
let url: URL
let offset: TimeInterval
let isTemporary: Bool
}
private let transcriptionChunkDuration: TimeInterval = 3 * 60
private let transcriptionSession: URLSession
private init() {
@@ -151,15 +169,85 @@ final class CoderAPIClient {
}
}
func transcribe(fileURL: URL, model: String, language: String = "en") async throws -> Transcription {
func transcribe(
fileURL: URL,
model: String,
language: String = "en",
diarization: Bool = false,
maxSpeakerCount: Int = 4
) async throws -> Transcription {
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)
defer {
for chunk in chunks where chunk.isTemporary {
try? FileManager.default.removeItem(at: chunk.url)
}
}
var textParts: [String] = []
var segments: [Transcription.Segment] = []
var lastNormalizedText = ""
var consecutiveDuplicateCount = 0
for chunk in chunks {
let transcription = try await transcribeChunk(
chunk.url,
model: selectedModel,
language: language,
apiKey: apiKey,
diarization: diarization,
maxSpeakerCount: maxSpeakerCount
)
if transcription.segments.isEmpty {
let text = transcription.text.trimmingCharacters(in: .whitespacesAndNewlines)
if !text.isEmpty {
textParts.append(text)
segments.append(.init(start: chunk.offset, end: chunk.offset, text: text, speaker: nil))
}
continue
}
for segment in transcription.segments {
let text = segment.text.trimmingCharacters(in: .whitespacesAndNewlines)
guard !text.isEmpty else { continue }
let normalized = text.lowercased()
if normalized == lastNormalizedText {
consecutiveDuplicateCount += 1
} else {
lastNormalizedText = normalized
consecutiveDuplicateCount = 1
}
guard consecutiveDuplicateCount <= 2 else { continue }
textParts.append(text)
segments.append(.init(
start: segment.start + chunk.offset,
end: segment.end + chunk.offset,
text: text,
speaker: segment.speaker
))
}
}
return Transcription(text: textParts.joined(separator: "\n"), segments: segments)
}
private func transcribeChunk(
_ fileURL: URL,
model: String,
language: String,
apiKey: String,
diarization: Bool,
maxSpeakerCount: Int
) async throws -> Transcription {
let boundary = "Meetingnotes-\(UUID().uuidString)"
let bodyURL = try makeMultipartBody(
audioURL: fileURL,
model: selectedModel,
model: model,
language: language,
diarization: diarization,
maxSpeakerCount: maxSpeakerCount,
boundary: boundary
)
defer { try? FileManager.default.removeItem(at: bodyURL) }
@@ -175,7 +263,125 @@ final class CoderAPIClient {
let (data, response) = try await transcriptionSession.upload(for: request, fromFile: bodyURL)
try validate(response: response, data: data)
let decoded = try JSONDecoder().decode(TranscriptionResponse.self, from: data)
return Transcription(text: decoded.text, segments: decoded.segments ?? [])
let segments = decoded.segments ?? segments(from: decoded.words ?? [])
return Transcription(text: decoded.text, segments: segments)
}
private func makeAudioChunks(from fileURL: URL, preserveSpeakerIdentity: Bool) throws -> [AudioChunk] {
let input = try AVAudioFile(forReading: fileURL)
let format = input.processingFormat
guard format.sampleRate > 0 else { throw CoderAPIError.invalidResponse }
let framesPerChunk = preserveSpeakerIdentity
? max(1, input.length)
: AVAudioFramePosition(format.sampleRate * transcriptionChunkDuration)
var chunks: [AudioChunk] = []
var frameOffset: AVAudioFramePosition = 0
do {
while frameOffset < input.length {
let frameCount = min(framesPerChunk, input.length - frameOffset)
let chunkURL = FileManager.default.temporaryDirectory
.appendingPathComponent("meetingnotes-transcription-\(UUID().uuidString).wav")
try writeAudioChunk(
from: input,
frameCount: frameCount,
format: format,
to: chunkURL
)
chunks.append(AudioChunk(
url: chunkURL,
offset: Double(frameOffset) / format.sampleRate,
isTemporary: true
))
frameOffset += frameCount
}
return chunks
} catch {
for chunk in chunks {
try? FileManager.default.removeItem(at: chunk.url)
}
throw error
}
}
private func writeAudioChunk(
from input: AVAudioFile,
frameCount: AVAudioFramePosition,
format: AVAudioFormat,
to outputURL: URL
) throws {
let settings: [String: Any] = [
AVFormatIDKey: kAudioFormatLinearPCM,
AVSampleRateKey: format.sampleRate,
AVNumberOfChannelsKey: format.channelCount,
AVLinearPCMBitDepthKey: 16,
AVLinearPCMIsFloatKey: false,
AVLinearPCMIsBigEndianKey: false,
AVLinearPCMIsNonInterleaved: false
]
let output = try AVAudioFile(
forWriting: outputURL,
settings: settings,
commonFormat: format.commonFormat,
interleaved: format.isInterleaved
)
var remaining = frameCount
while remaining > 0 {
let requestedFrames = AVAudioFrameCount(min(remaining, 8_192))
guard let buffer = AVAudioPCMBuffer(pcmFormat: format, frameCapacity: requestedFrames) else {
throw CoderAPIError.invalidResponse
}
try input.read(into: buffer, frameCount: requestedFrames)
guard buffer.frameLength > 0 else { break }
try output.write(from: buffer)
remaining -= AVAudioFramePosition(buffer.frameLength)
}
}
private func segments(from words: [TranscriptionResponse.Word]) -> [Transcription.Segment] {
var result: [Transcription.Segment] = []
var currentWords: [String] = []
var currentStart: TimeInterval?
var currentEnd: TimeInterval = 0
var currentSpeaker: Int?
func flush() {
guard let start = currentStart, !currentWords.isEmpty else { return }
result.append(.init(
start: start,
end: currentEnd,
text: currentWords.joined(separator: " "),
speaker: currentSpeaker
))
currentWords.removeAll(keepingCapacity: true)
currentStart = nil
currentEnd = 0
currentSpeaker = nil
}
for word in words {
let text = word.word.trimmingCharacters(in: .whitespacesAndNewlines)
guard !text.isEmpty else { continue }
let speakerChanged = currentStart != nil && word.speaker != currentSpeaker
let longPause = currentStart != nil && word.start - currentEnd > 1.5
if speakerChanged || longPause { flush() }
if currentStart == nil {
currentStart = word.start
currentSpeaker = word.speaker
}
currentWords.append(text)
currentEnd = word.end
let sentenceEnded = text.last.map { ".!?".contains($0) } ?? false
let duration = currentEnd - (currentStart ?? currentEnd)
if currentWords.count >= 40 || (sentenceEnded && (currentWords.count >= 12 || duration >= 8)) {
flush()
}
}
flush()
return result
}
private func endpoint(baseURL: String, path: String) throws -> URL {
@@ -207,7 +413,14 @@ final class CoderAPIClient {
}
}
private func makeMultipartBody(audioURL: URL, model: String, language: String, boundary: String) throws -> URL {
private func makeMultipartBody(
audioURL: URL,
model: String,
language: String,
diarization: Bool,
maxSpeakerCount: Int,
boundary: String
) throws -> URL {
let bodyURL = FileManager.default.temporaryDirectory.appendingPathComponent("meetingnotes-upload-\(UUID().uuidString).body")
_ = FileManager.default.createFile(atPath: bodyURL.path, contents: nil)
let output = try FileHandle(forWritingTo: bodyURL)
@@ -219,7 +432,11 @@ final class CoderAPIClient {
try write("--\(boundary)\r\nContent-Disposition: form-data; name=\"model\"\r\n\r\n\(model)\r\n")
try write("--\(boundary)\r\nContent-Disposition: form-data; name=\"language\"\r\n\r\n\(language)\r\n")
try write("--\(boundary)\r\nContent-Disposition: form-data; name=\"response_format\"\r\n\r\nverbose_json\r\n")
try write("--\(boundary)\r\nContent-Disposition: form-data; name=\"file\"; filename=\"\(audioURL.lastPathComponent)\"\r\nContent-Type: audio/mp4\r\n\r\n")
if diarization {
try write("--\(boundary)\r\nContent-Disposition: form-data; name=\"diarization\"\r\n\r\ntrue\r\n")
try write("--\(boundary)\r\nContent-Disposition: form-data; name=\"max_speaker_count\"\r\n\r\n\(maxSpeakerCount)\r\n")
}
try write("--\(boundary)\r\nContent-Disposition: form-data; name=\"file\"; filename=\"\(audioURL.lastPathComponent)\"\r\nContent-Type: audio/wav\r\n\r\n")
let input = try FileHandle(forReadingFrom: audioURL)
defer { try? input.close() }
while let chunk = try input.read(upToCount: 1 << 20), !chunk.isEmpty {
@@ -244,10 +244,6 @@ class MeetingViewModel: ObservableObject {
throw CocoaError(.fileWriteUnknown)
}
LocalStorageManager.shared.deleteRecoveryAudioFolder(recoveryAudioFolderURL)
meeting.recoveryAudioFolderName = nil
self.recoveryAudioFolderURL = nil
saveMeeting()
await generateNotes()
} catch {
errorMessage = error.localizedDescription
@@ -262,6 +258,11 @@ class MeetingViewModel: ObservableObject {
meeting.recoveryAudioFolderName = recoveryAudioFolderURL.lastPathComponent
}
}
func showAudioInFinder() {
guard let recoveryAudioFolderURL else { return }
LocalStorageManager.shared.showAudioFolderInFinder(recoveryAudioFolderURL)
}
func loadTemplates() {
templates = LocalStorageManager.shared.loadTemplates()
@@ -71,6 +71,7 @@ class SettingsViewModel: ObservableObject {
// via computed properties when they're modified
let coderSaved = KeychainHelper.shared.saveCoderAPIKey(settings.coderAPIKey)
LocalAPIServer.shared.applyConfiguration()
LocalStorageManager.shared.purgeExpiredAudioFolders()
if showMessage {
if coderSaved {
+8 -2
View File
@@ -208,12 +208,12 @@ struct CollapsedTranscriptChunkView: View {
.font(.caption)
.foregroundColor(chunk.source == .mic ? .blue : .orange)
Text(chunk.source.displayName)
Text(chunk.displayName)
.font(.caption)
.fontWeight(.medium)
.foregroundColor(chunk.source == .mic ? .blue : .orange)
}
.frame(width: 50, alignment: .leading)
.frame(width: 78, alignment: .leading)
// Transcript text
Text(chunk.combinedText)
@@ -267,6 +267,12 @@ struct MeetingDetailContentView: View {
}
.disabled(!viewModel.canRetryTranscription)
Button {
viewModel.showAudioInFinder()
} label: {
Label("Show Audio in Finder", systemImage: "folder")
}
Divider()
}
+29
View File
@@ -58,6 +58,18 @@ struct SettingsView: View {
Text(model.displayName).tag(model.id)
}
}
if let selectedModel = viewModel.coderModels.first(where: { $0.id == viewModel.settings.transcriptionModel }) {
if selectedModel.supportsSpeakerDiarization {
Label("Remote participants are labeled Speaker 14; your microphone is labeled Me.", systemImage: "person.2.wave.2")
.font(.caption)
.foregroundColor(.secondary)
} else {
Label("Remote participants are labeled Them; your microphone is labeled Me.", systemImage: "person.2")
.font(.caption)
.foregroundColor(.secondary)
}
}
}
Text("The token is stored locally in Keychain. Audio and note generation are sent only to this Coder service.")
@@ -128,6 +140,23 @@ struct SettingsView: View {
Text("Meeting Storage")
.font(.headline)
LabeledContent("Audio retention") {
Stepper(
value: $viewModel.settings.audioRetentionDays,
in: 1...365
) {
Text("\(viewModel.settings.audioRetentionDays) \(viewModel.settings.audioRetentionDays == 1 ? "day" : "days")")
.monospacedDigit()
.frame(minWidth: 70, alignment: .trailing)
}
}
Button {
LocalStorageManager.shared.showAudioFolderInFinder()
} label: {
Label("Show Audio Folder", systemImage: "folder")
}
Button {
showingMeetingImporter = true
} label: {
+32 -2
View File
@@ -14,6 +14,9 @@ APP_PATH="$DERIVED_DATA/Build/Products/Release/$APP_NAME.app"
required_variables=(
VERSION
SIGNING_IDENTITY
APPLE_ID
APPLE_TEAM_ID
APPLE_APP_PASSWORD
SPARKLE_PRIVATE_KEY
GITHUB_REPOSITORY
)
@@ -80,6 +83,33 @@ ARCHIVE_NAME="$APP_NAME-$VERSION.zip"
ARCHIVE_PATH="$RELEASE_DIR/$ARCHIVE_NAME"
ditto -c -k --sequesterRsrc --keepParent "$APP_PATH" "$ARCHIVE_PATH"
NOTARY_RESULT="$BUILD_ROOT/notary-result.json"
xcrun notarytool submit "$ARCHIVE_PATH" \
--apple-id "$APPLE_ID" \
--team-id "$APPLE_TEAM_ID" \
--password "$APPLE_APP_PASSWORD" \
--wait \
--timeout 20m \
--output-format json > "$NOTARY_RESULT"
NOTARY_STATUS=$(plutil -extract status raw -o - "$NOTARY_RESULT")
if [[ "$NOTARY_STATUS" != "Accepted" ]]; then
submission_id=$(plutil -extract id raw -o - "$NOTARY_RESULT")
xcrun notarytool log "$submission_id" \
--apple-id "$APPLE_ID" \
--team-id "$APPLE_TEAM_ID" \
--password "$APPLE_APP_PASSWORD" || true
echo "Apple notarization failed with status: $NOTARY_STATUS" >&2
exit 1
fi
xcrun stapler staple "$APP_PATH"
xcrun stapler validate "$APP_PATH"
spctl --assess --type execute --verbose=2 "$APP_PATH"
rm -f "$ARCHIVE_PATH"
ditto -c -k --sequesterRsrc --keepParent "$APP_PATH" "$ARCHIVE_PATH"
GENERATE_APPCAST=$(find "$DERIVED_DATA/SourcePackages/artifacts" -type f -name generate_appcast -print -quit)
if [[ -z "$GENERATE_APPCAST" ]]; then
echo "Sparkle generate_appcast tool was not found" >&2
@@ -97,8 +127,8 @@ grep -q "$DOWNLOAD_URL$ARCHIVE_NAME" "$RELEASE_DIR/appcast.xml"
grep -q 'sparkle:edSignature=' "$RELEASE_DIR/appcast.xml"
if [[ -n "${GITHUB_STEP_SUMMARY:-}" ]]; then
printf 'Built and Developer ID-signed Meetingnotes %s. The GitHub release is ready to publish.\n' \
printf 'Built, Developer ID-signed, notarized, and stapled Meetingnotes %s. The GitHub release is ready to publish.\n' \
"$VERSION" >> "$GITHUB_STEP_SUMMARY"
fi
echo "Signed release artifacts are ready in $RELEASE_DIR"
echo "Signed and notarized release artifacts are ready in $RELEASE_DIR"