feat: actually implement system audio + refactor
This commit is contained in:
@@ -1,196 +0,0 @@
|
||||
// AudioManager.swift
|
||||
import AVFoundation
|
||||
import Foundation
|
||||
import SwiftUI
|
||||
import ScreenCaptureKit
|
||||
|
||||
class AudioManager: NSObject, ObservableObject {
|
||||
@Published var transcript = ""
|
||||
@Published var isRecording = false
|
||||
|
||||
private var stream: SCStream?
|
||||
private var webSocketTask: URLSessionWebSocketTask?
|
||||
private let deepgramURL = URL(string: "wss://api.deepgram.com/v1/listen?encoding=linear16&sample_rate=48000&channels=1&interim_results=true")!
|
||||
|
||||
func startRecording() {
|
||||
// Check if ScreenCaptureKit is available
|
||||
guard #available(macOS 13.0, *) else {
|
||||
print("ScreenCaptureKit requires macOS 13.0 or later")
|
||||
return
|
||||
}
|
||||
|
||||
Task {
|
||||
do {
|
||||
// Request permission for screen recording (includes system audio)
|
||||
let canRecord = await requestScreenRecordingPermission()
|
||||
if !canRecord {
|
||||
print("Screen recording permission denied")
|
||||
DispatchQueue.main.async {
|
||||
self.isRecording = false
|
||||
}
|
||||
return
|
||||
}
|
||||
|
||||
// Set up screen capture configuration
|
||||
let config = SCStreamConfiguration()
|
||||
config.capturesAudio = true
|
||||
config.captureMicrophone = true
|
||||
config.sampleRate = 48000
|
||||
config.channelCount = 1
|
||||
|
||||
// Create content filter for desktop audio capture
|
||||
let availableContent = try await SCShareableContent.current
|
||||
guard let display = availableContent.displays.first else {
|
||||
print("No displays available for capture")
|
||||
DispatchQueue.main.async {
|
||||
self.isRecording = false
|
||||
}
|
||||
return
|
||||
}
|
||||
let filter = SCContentFilter(display: display, excludingWindows: [])
|
||||
|
||||
// Create stream
|
||||
stream = SCStream(filter: filter, configuration: config, delegate: self)
|
||||
|
||||
// Add audio output
|
||||
try stream?.addStreamOutput(self, type: .audio, sampleHandlerQueue: DispatchQueue(label: "AudioQueue"))
|
||||
try stream?.addStreamOutput(self, type: .microphone, sampleHandlerQueue: DispatchQueue(label: "MicQueue"))
|
||||
|
||||
// Start capture
|
||||
try await stream?.startCapture()
|
||||
|
||||
DispatchQueue.main.async {
|
||||
self.isRecording = true
|
||||
}
|
||||
|
||||
// Connect to Deepgram
|
||||
connectToDeepgram()
|
||||
|
||||
} catch {
|
||||
print("Failed to start recording: \(error)")
|
||||
DispatchQueue.main.async {
|
||||
self.isRecording = false
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func stopRecording() {
|
||||
Task {
|
||||
do {
|
||||
try await stream?.stopCapture()
|
||||
stream = nil
|
||||
webSocketTask?.cancel(with: .normalClosure, reason: nil)
|
||||
webSocketTask = nil
|
||||
|
||||
DispatchQueue.main.async {
|
||||
self.isRecording = false
|
||||
}
|
||||
} catch {
|
||||
print("Failed to stop recording: \(error)")
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
private func requestScreenRecordingPermission() async -> Bool {
|
||||
// ScreenCaptureKit handles permission requests automatically
|
||||
// This will show a system dialog if needed
|
||||
do {
|
||||
let _ = try await SCShareableContent.current
|
||||
return true
|
||||
} catch {
|
||||
return false
|
||||
}
|
||||
}
|
||||
|
||||
private func connectToDeepgram() {
|
||||
let session = URLSession(configuration: .default)
|
||||
var request = URLRequest(url: deepgramURL)
|
||||
if let key = KeychainHelper.shared.get(forKey: "deepgramKey"), !key.isEmpty {
|
||||
request.addValue("Token \(key)", forHTTPHeaderField: "Authorization")
|
||||
} else {
|
||||
print("No Deepgram key found")
|
||||
return
|
||||
}
|
||||
webSocketTask = session.webSocketTask(with: request)
|
||||
webSocketTask?.resume()
|
||||
receiveMessage()
|
||||
}
|
||||
|
||||
private func sendAudioData(_ data: Data) {
|
||||
webSocketTask?.send(.data(data)) { error in
|
||||
if let error = error {
|
||||
print("Send error: \(error)")
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
private func receiveMessage() {
|
||||
webSocketTask?.receive { result in
|
||||
switch result {
|
||||
case .success(let message):
|
||||
switch message {
|
||||
case .string(let text):
|
||||
self.parseTranscription(text)
|
||||
case .data:
|
||||
break
|
||||
@unknown default:
|
||||
break
|
||||
}
|
||||
self.receiveMessage() // Continue receiving
|
||||
case .failure(let error):
|
||||
print("Receive error: \(error)")
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
private func parseTranscription(_ text: String) {
|
||||
guard let data = text.data(using: .utf8),
|
||||
let json = try? JSONSerialization.jsonObject(with: data) as? [String: Any],
|
||||
let type = json["type"] as? String, type == "Results",
|
||||
let channel = json["channel"] as? [String: Any],
|
||||
let alternatives = channel["alternatives"] as? [[String: Any]],
|
||||
let alt = alternatives.first,
|
||||
let transcript = alt["transcript"] as? String,
|
||||
!transcript.isEmpty else { return }
|
||||
|
||||
DispatchQueue.main.async {
|
||||
if let isFinal = json["is_final"] as? Bool, isFinal {
|
||||
self.transcript += transcript + " "
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// MARK: - SCStreamDelegate
|
||||
extension AudioManager: SCStreamDelegate {
|
||||
func stream(_ stream: SCStream, didStopWithError error: Error) {
|
||||
print("Stream stopped with error: \(error)")
|
||||
DispatchQueue.main.async {
|
||||
self.isRecording = false
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// MARK: - SCStreamOutput
|
||||
extension AudioManager: SCStreamOutput {
|
||||
func stream(_ stream: SCStream, didOutputSampleBuffer sampleBuffer: CMSampleBuffer, of type: SCStreamOutputType) {
|
||||
// Convert sample buffer to audio data and send to Deepgram
|
||||
guard let audioBuffer = CMSampleBufferGetDataBuffer(sampleBuffer) else {
|
||||
print("Failed to get data buffer from sample buffer")
|
||||
return
|
||||
}
|
||||
|
||||
var length: Int = 0
|
||||
var dataPointer: UnsafeMutablePointer<Int8>?
|
||||
let status = CMBlockBufferGetDataPointer(audioBuffer, atOffset: 0, lengthAtOffsetOut: nil, totalLengthOut: &length, dataPointerOut: &dataPointer)
|
||||
|
||||
guard status == noErr, let pointer = dataPointer, length > 0 else {
|
||||
print("Failed to get audio data pointer, status: \(status), length: \(length)")
|
||||
return
|
||||
}
|
||||
|
||||
let data = Data(bytes: pointer, count: length)
|
||||
sendAudioData(data)
|
||||
}
|
||||
}
|
||||
+17
-73
@@ -8,80 +8,24 @@
|
||||
import SwiftUI
|
||||
|
||||
struct ContentView: View {
|
||||
@StateObject private var audioManager = SimpleAudioManager()
|
||||
@State private var deepgramKey = ""
|
||||
@State private var openAIKey = ""
|
||||
@State private var userBlurb = ""
|
||||
@State private var systemPrompt = "Generate meeting notes from transcript and notes."
|
||||
|
||||
var body: some View {
|
||||
VStack(spacing: 20) {
|
||||
Text("Notetaker Settings")
|
||||
.font(.title)
|
||||
|
||||
TextField("Deepgram API Key", text: $deepgramKey)
|
||||
TextField("OpenAI API Key", text: $openAIKey)
|
||||
TextField("About Yourself (for prompt)", text: $userBlurb)
|
||||
TextEditor(text: $systemPrompt)
|
||||
.frame(height: 100)
|
||||
|
||||
Button("Save") {
|
||||
saveSettings()
|
||||
}
|
||||
|
||||
Divider()
|
||||
|
||||
Toggle("Capture System Audio", isOn: $audioManager.captureSystemAudio)
|
||||
.disabled(audioManager.isRecording)
|
||||
|
||||
HStack {
|
||||
Button("Start Recording") {
|
||||
audioManager.startRecording()
|
||||
}
|
||||
.disabled(audioManager.isRecording)
|
||||
|
||||
Button("Stop Recording") {
|
||||
audioManager.stopRecording()
|
||||
}
|
||||
.disabled(!audioManager.isRecording)
|
||||
}
|
||||
|
||||
if audioManager.isRecording {
|
||||
Text(audioManager.captureSystemAudio ? "🔴 Recording microphone + system audio..." : "🔴 Recording microphone...")
|
||||
.foregroundColor(.red)
|
||||
}
|
||||
|
||||
Text("Live Transcript:")
|
||||
.font(.headline)
|
||||
|
||||
ScrollView {
|
||||
Text(audioManager.transcript.isEmpty ? "No transcript yet..." : audioManager.transcript)
|
||||
.frame(maxWidth: .infinity, alignment: .leading)
|
||||
.padding()
|
||||
}
|
||||
.frame(height: 150)
|
||||
.border(Color.gray)
|
||||
}
|
||||
.padding()
|
||||
.frame(minWidth: 500, minHeight: 600)
|
||||
.onAppear {
|
||||
loadSettings()
|
||||
}
|
||||
}
|
||||
@State private var showingSettings = false
|
||||
|
||||
func saveSettings() {
|
||||
_ = KeychainHelper.shared.save(deepgramKey, forKey: "deepgramKey")
|
||||
_ = KeychainHelper.shared.save(openAIKey, forKey: "openAIKey")
|
||||
_ = KeychainHelper.shared.save(userBlurb, forKey: "userBlurb")
|
||||
_ = KeychainHelper.shared.save(systemPrompt, forKey: "systemPrompt")
|
||||
print("Settings saved!")
|
||||
}
|
||||
|
||||
private func loadSettings() {
|
||||
deepgramKey = KeychainHelper.shared.get(forKey: "deepgramKey") ?? ""
|
||||
openAIKey = KeychainHelper.shared.get(forKey: "openAIKey") ?? ""
|
||||
userBlurb = KeychainHelper.shared.get(forKey: "userBlurb") ?? ""
|
||||
systemPrompt = KeychainHelper.shared.get(forKey: "systemPrompt") ?? "Generate meeting notes from transcript and notes."
|
||||
var body: some View {
|
||||
VStack {
|
||||
MeetingListView()
|
||||
.toolbar {
|
||||
ToolbarItem(placement: .automatic) {
|
||||
Button {
|
||||
showingSettings = true
|
||||
} label: {
|
||||
Label("Settings", systemImage: "gear")
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
.sheet(isPresented: $showingSettings) {
|
||||
SettingsView()
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@@ -4,5 +4,7 @@
|
||||
<dict>
|
||||
<key>NSAudioCaptureUsageDescription</key>
|
||||
<string>Notetaker needs access to capture system audio for transcription.</string>
|
||||
<key>NSScreenCaptureUsageDescription</key>
|
||||
<string>Notetaker needs screen recording permission to capture system audio for transcription.</string>
|
||||
</dict>
|
||||
</plist>
|
||||
|
||||
@@ -1,42 +0,0 @@
|
||||
// KeychainHelper.swift
|
||||
// Helper for secure storage
|
||||
|
||||
import Foundation
|
||||
import Security
|
||||
|
||||
class KeychainHelper {
|
||||
static let shared = KeychainHelper()
|
||||
|
||||
private init() {}
|
||||
|
||||
func save(_ value: String, forKey key: String) -> Bool {
|
||||
guard let data = value.data(using: .utf8) else { return false }
|
||||
|
||||
let query: [String: Any] = [
|
||||
kSecClass as String: kSecClassGenericPassword,
|
||||
kSecAttrAccount as String: key,
|
||||
kSecValueData as String: data,
|
||||
kSecAttrService as String: "owen.notetaker" // Replace with your bundle ID
|
||||
]
|
||||
|
||||
SecItemDelete(query as CFDictionary) // Delete if exists
|
||||
let status = SecItemAdd(query as CFDictionary, nil)
|
||||
return status == errSecSuccess
|
||||
}
|
||||
|
||||
func get(forKey key: String) -> String? {
|
||||
let query: [String: Any] = [
|
||||
kSecClass as String: kSecClassGenericPassword,
|
||||
kSecAttrAccount as String: key,
|
||||
kSecAttrService as String: "owen.notetaker",
|
||||
kSecMatchLimit as String: kSecMatchLimitOne,
|
||||
kSecReturnData as String: true
|
||||
]
|
||||
|
||||
var item: AnyObject?
|
||||
let status = SecItemCopyMatching(query as CFDictionary, &item)
|
||||
|
||||
guard status == errSecSuccess, let data = item as? Data else { return nil }
|
||||
return String(data: data, encoding: .utf8)
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,392 @@
|
||||
// AudioManager.swift
|
||||
// Unified audio manager for microphone and system audio capture
|
||||
|
||||
import AVFoundation
|
||||
import Foundation
|
||||
import SwiftUI
|
||||
import ScreenCaptureKit
|
||||
|
||||
// Distinguish which source (mic vs system) an audio buffer belongs to
|
||||
private enum AudioSource {
|
||||
case mic
|
||||
case system
|
||||
}
|
||||
|
||||
/// Manages audio capture from microphone and/or system audio and handles real-time transcription via Deepgram
|
||||
class AudioManager: NSObject, ObservableObject {
|
||||
@Published var transcript = ""
|
||||
@Published var isRecording = false
|
||||
@Published var captureSystemAudio = true
|
||||
|
||||
private var audioEngine = AVAudioEngine()
|
||||
private var micSocketTask: URLSessionWebSocketTask?
|
||||
private var systemSocketTask: URLSessionWebSocketTask?
|
||||
private let deepgramURL = URL(string: "wss://api.deepgram.com/v1/listen?encoding=linear16&sample_rate=16000&channels=1&interim_results=true")!
|
||||
|
||||
// ScreenCaptureKit properties
|
||||
private var stream: SCStream?
|
||||
|
||||
func startRecording() {
|
||||
print("Starting recording...")
|
||||
|
||||
// First ensure everything is stopped and cleaned up
|
||||
stopRecordingInternal()
|
||||
|
||||
if captureSystemAudio {
|
||||
// Start microphone capture in parallel with system audio
|
||||
startMicrophoneTap()
|
||||
// Start system audio capture asynchronously
|
||||
Task {
|
||||
await startSystemAudioCapture()
|
||||
}
|
||||
} else {
|
||||
startMicrophoneOnly()
|
||||
}
|
||||
}
|
||||
|
||||
private func stopRecordingInternal() {
|
||||
print("Internal cleanup...")
|
||||
|
||||
// Stop system audio capture
|
||||
if let stream = stream {
|
||||
stream.stopCapture()
|
||||
self.stream = nil
|
||||
print("System audio capture stopped")
|
||||
}
|
||||
|
||||
// Stop microphone capture
|
||||
cleanupAudioEngine()
|
||||
|
||||
// Close WebSocket
|
||||
micSocketTask?.cancel(with: .normalClosure, reason: nil)
|
||||
micSocketTask = nil
|
||||
systemSocketTask?.cancel(with: .normalClosure, reason: nil)
|
||||
systemSocketTask = nil
|
||||
|
||||
// Reset state
|
||||
DispatchQueue.main.async {
|
||||
self.isRecording = false
|
||||
}
|
||||
|
||||
print("Internal cleanup completed")
|
||||
}
|
||||
|
||||
private func startMicrophoneOnly() {
|
||||
print("Starting microphone-only recording...")
|
||||
|
||||
// Ensure audio engine is stopped and cleaned up
|
||||
cleanupAudioEngine()
|
||||
|
||||
// Set up audio engine for microphone capture
|
||||
let inputNode = audioEngine.inputNode
|
||||
let recordingFormat = inputNode.outputFormat(forBus: 0)
|
||||
|
||||
// Convert to the format Deepgram expects: 16kHz, 16-bit, mono
|
||||
let targetFormat = AVAudioFormat(commonFormat: .pcmFormatInt16,
|
||||
sampleRate: 16000,
|
||||
channels: 1,
|
||||
interleaved: false)!
|
||||
|
||||
let converter = AVAudioConverter(from: recordingFormat, to: targetFormat)!
|
||||
|
||||
inputNode.installTap(onBus: 0, bufferSize: 1024, format: recordingFormat) { [weak self] (buffer, time) in
|
||||
self?.processAudioBuffer(buffer, converter: converter, targetFormat: targetFormat, source: .mic)
|
||||
}
|
||||
|
||||
audioEngine.prepare()
|
||||
|
||||
do {
|
||||
try audioEngine.start()
|
||||
DispatchQueue.main.async {
|
||||
self.isRecording = true
|
||||
}
|
||||
connectToDeepgram(source: .mic)
|
||||
print("Microphone recording started")
|
||||
} catch {
|
||||
print("Failed to start audio engine: \(error)")
|
||||
}
|
||||
}
|
||||
|
||||
/// Starts a microphone tap without creating a new Deepgram connection (used when also capturing system audio)
|
||||
private func startMicrophoneTap() {
|
||||
// Ensure audio engine is stopped and cleaned up
|
||||
cleanupAudioEngine()
|
||||
|
||||
let inputNode = audioEngine.inputNode
|
||||
let recordingFormat = inputNode.outputFormat(forBus: 0)
|
||||
|
||||
let targetFormat = AVAudioFormat(commonFormat: .pcmFormatInt16,
|
||||
sampleRate: 16000,
|
||||
channels: 1,
|
||||
interleaved: false)!
|
||||
|
||||
let converter = AVAudioConverter(from: recordingFormat, to: targetFormat)!
|
||||
|
||||
inputNode.installTap(onBus: 0, bufferSize: 1024, format: recordingFormat) { [weak self] buffer, _ in
|
||||
// Debug mic RMS
|
||||
if let ch = buffer.floatChannelData?[0] {
|
||||
let frameCount = Int(buffer.frameLength)
|
||||
let samples = UnsafeBufferPointer(start: ch, count: frameCount)
|
||||
let rms = sqrt(samples.map { $0 * $0 }.reduce(0, +) / Float(frameCount))
|
||||
print("🎤 Mic RMS: \(rms)")
|
||||
}
|
||||
self?.processAudioBuffer(buffer, converter: converter, targetFormat: targetFormat, source: .mic)
|
||||
}
|
||||
|
||||
audioEngine.prepare()
|
||||
|
||||
do {
|
||||
try audioEngine.start()
|
||||
connectToDeepgram(source: .mic)
|
||||
} catch {
|
||||
print("Failed to start microphone tap: \(error)")
|
||||
}
|
||||
}
|
||||
|
||||
private func cleanupAudioEngine() {
|
||||
if audioEngine.isRunning {
|
||||
audioEngine.stop()
|
||||
}
|
||||
|
||||
// Reset the audio engine - this safely removes all taps
|
||||
audioEngine.reset()
|
||||
}
|
||||
|
||||
private func startSystemAudioCapture() async {
|
||||
print("🎧 Starting system audio capture...")
|
||||
|
||||
do {
|
||||
// Request screen capture permission
|
||||
let content = try await SCShareableContent.excludingDesktopWindows(true, onScreenWindowsOnly: true)
|
||||
|
||||
// Exclude self to avoid feedback
|
||||
let excludedApps = content.applications.filter { app in
|
||||
Bundle.main.bundleIdentifier == app.bundleIdentifier
|
||||
}
|
||||
|
||||
guard let display = content.displays.first else {
|
||||
print("❌ No display found")
|
||||
startMicrophoneOnly()
|
||||
return
|
||||
}
|
||||
|
||||
// Create filter
|
||||
let filter = SCContentFilter(display: display, excludingApplications: excludedApps, exceptingWindows: [])
|
||||
|
||||
// Configure stream
|
||||
let configuration = SCStreamConfiguration()
|
||||
configuration.width = 2 // Minimal video settings
|
||||
configuration.height = 2
|
||||
configuration.minimumFrameInterval = CMTime(value: 1, timescale: CMTimeScale.max)
|
||||
configuration.capturesAudio = true
|
||||
configuration.sampleRate = 48000
|
||||
configuration.channelCount = 2
|
||||
|
||||
// Create stream
|
||||
let stream = SCStream(filter: filter, configuration: configuration, delegate: self)
|
||||
|
||||
// Add stream output for audio processing
|
||||
try stream.addStreamOutput(self, type: .audio, sampleHandlerQueue: .global(qos: .userInitiated))
|
||||
// Add a minimal screen output so SCStream doesn't complain about missing video output
|
||||
try stream.addStreamOutput(self, type: .screen, sampleHandlerQueue: .global(qos: .userInitiated))
|
||||
|
||||
// Start capture
|
||||
try await stream.startCapture()
|
||||
|
||||
// Store reference
|
||||
self.stream = stream
|
||||
|
||||
DispatchQueue.main.async {
|
||||
self.isRecording = true
|
||||
}
|
||||
|
||||
connectToDeepgram(source: .system)
|
||||
print("✅ System audio capture started successfully")
|
||||
|
||||
} catch {
|
||||
print("❌ Failed to start system audio capture: \(error)")
|
||||
|
||||
if case SCStreamError.userDeclined = error {
|
||||
print("📍 Permission denied. User needs to enable screen recording in System Settings.")
|
||||
}
|
||||
|
||||
// Fallback to microphone only
|
||||
startMicrophoneOnly()
|
||||
}
|
||||
}
|
||||
|
||||
func stopRecording() {
|
||||
print("Stopping recording...")
|
||||
|
||||
// Stop system audio capture
|
||||
if let stream = stream {
|
||||
stream.stopCapture()
|
||||
self.stream = nil
|
||||
}
|
||||
|
||||
// Stop microphone capture
|
||||
cleanupAudioEngine()
|
||||
|
||||
// Close WebSocket
|
||||
micSocketTask?.cancel(with: .normalClosure, reason: nil)
|
||||
micSocketTask = nil
|
||||
systemSocketTask?.cancel(with: .normalClosure, reason: nil)
|
||||
systemSocketTask = nil
|
||||
|
||||
DispatchQueue.main.async {
|
||||
self.isRecording = false
|
||||
}
|
||||
|
||||
print("Recording stopped")
|
||||
}
|
||||
|
||||
private func processAudioBuffer(_ buffer: AVAudioPCMBuffer, converter: AVAudioConverter, targetFormat: AVAudioFormat, source: AudioSource) {
|
||||
var processBuffer = buffer
|
||||
|
||||
// Convert to target format (16kHz int16 mono) in a single step – AVAudioConverter will handle resampling and downmixing
|
||||
let outputFrameCapacity = AVAudioFrameCount(Double(processBuffer.frameLength) * targetFormat.sampleRate / processBuffer.format.sampleRate)
|
||||
guard let outputBuffer = AVAudioPCMBuffer(pcmFormat: targetFormat, frameCapacity: outputFrameCapacity) else {
|
||||
return
|
||||
}
|
||||
|
||||
var error: NSError?
|
||||
let status = converter.convert(to: outputBuffer, error: &error) { _, outStatus in
|
||||
outStatus.pointee = .haveData
|
||||
return processBuffer
|
||||
}
|
||||
|
||||
guard status == .haveData, error == nil else {
|
||||
return
|
||||
}
|
||||
|
||||
// Convert to Data for Deepgram
|
||||
guard let channelData = outputBuffer.int16ChannelData?[0] else {
|
||||
return
|
||||
}
|
||||
|
||||
let frameCount = Int(outputBuffer.frameLength)
|
||||
let data = Data(bytes: channelData, count: frameCount * 2)
|
||||
|
||||
sendAudioData(data, source: source)
|
||||
}
|
||||
|
||||
private func connectToDeepgram(source: AudioSource) {
|
||||
guard let key = KeychainHelper.shared.get(forKey: "deepgramKey"), !key.isEmpty else {
|
||||
print("❌ No Deepgram key found")
|
||||
return
|
||||
}
|
||||
|
||||
let session = URLSession(configuration: .default)
|
||||
var request = URLRequest(url: deepgramURL)
|
||||
request.addValue("Token \(key)", forHTTPHeaderField: "Authorization")
|
||||
|
||||
let task = session.webSocketTask(with: request)
|
||||
task.resume()
|
||||
|
||||
switch source {
|
||||
case .mic:
|
||||
micSocketTask = task
|
||||
case .system:
|
||||
systemSocketTask = task
|
||||
}
|
||||
|
||||
receiveMessage(for: source)
|
||||
print("🌐 Connected to Deepgram (\(source))")
|
||||
}
|
||||
|
||||
private func receiveMessage(for source: AudioSource) {
|
||||
let task: URLSessionWebSocketTask? = (source == .mic) ? micSocketTask : systemSocketTask
|
||||
task?.receive { [weak self] result in
|
||||
switch result {
|
||||
case .success(let message):
|
||||
switch message {
|
||||
case .string(let text):
|
||||
self?.parseTranscription(text, source: source)
|
||||
case .data:
|
||||
break
|
||||
@unknown default:
|
||||
break
|
||||
}
|
||||
self?.receiveMessage(for: source) // continue loop
|
||||
case .failure(let error):
|
||||
print("❌ Receive error (\(source)): \(error)")
|
||||
// Attempt reconnect if still recording
|
||||
DispatchQueue.main.asyncAfter(deadline: .now() + 2) {
|
||||
if self?.isRecording == true {
|
||||
self?.connectToDeepgram(source: source)
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
private func parseTranscription(_ text: String, source: AudioSource) {
|
||||
guard let data = text.data(using: .utf8),
|
||||
let json = try? JSONSerialization.jsonObject(with: data) as? [String: Any],
|
||||
let type = json["type"] as? String, type == "Results",
|
||||
let channel = json["channel"] as? [String: Any],
|
||||
let alternatives = channel["alternatives"] as? [[String: Any]],
|
||||
let alt = alternatives.first,
|
||||
let transcriptText = alt["transcript"] as? String,
|
||||
!transcriptText.isEmpty else { return }
|
||||
|
||||
let prefix = (source == .mic) ? "[MIC]" : "[SYS]"
|
||||
DispatchQueue.main.async {
|
||||
if let isFinal = json["is_final"] as? Bool, isFinal {
|
||||
self.transcript += "\(prefix) \(transcriptText) "
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
private func sendAudioData(_ data: Data, source: AudioSource) {
|
||||
let task: URLSessionWebSocketTask? = (source == .mic) ? micSocketTask : systemSocketTask
|
||||
|
||||
guard let socket = task, socket.state == .running else { return }
|
||||
|
||||
socket.send(.data(data)) { error in
|
||||
if let error = error {
|
||||
print("❌ Send error (\(source)): \(error)")
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// MARK: - SCStreamDelegate & SCStreamOutput
|
||||
extension AudioManager: SCStreamDelegate, SCStreamOutput {
|
||||
func stream(_ stream: SCStream, didOutputSampleBuffer sampleBuffer: CMSampleBuffer, of type: SCStreamOutputType) {
|
||||
guard type == .audio else { return }
|
||||
guard sampleBuffer.isValid else { return }
|
||||
|
||||
// Convert CMSampleBuffer to AVAudioPCMBuffer
|
||||
guard let pcmBuffer = sampleBuffer.asPCMBuffer else { return }
|
||||
|
||||
// Create converter for Deepgram format
|
||||
let targetFormat = AVAudioFormat(commonFormat: .pcmFormatInt16,
|
||||
sampleRate: 16000,
|
||||
channels: 1,
|
||||
interleaved: false)!
|
||||
|
||||
guard let converter = AVAudioConverter(from: pcmBuffer.format, to: targetFormat) else { return }
|
||||
|
||||
processAudioBuffer(pcmBuffer, converter: converter, targetFormat: targetFormat, source: .system)
|
||||
}
|
||||
|
||||
func stream(_ stream: SCStream, didStopWithError error: Error) {
|
||||
print("❌ Stream stopped with error: \(error)")
|
||||
DispatchQueue.main.async {
|
||||
self.stream = nil
|
||||
self.isRecording = false
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// MARK: - CMSampleBuffer Extension
|
||||
extension CMSampleBuffer {
|
||||
var asPCMBuffer: AVAudioPCMBuffer? {
|
||||
try? self.withAudioBufferList { audioBufferList, _ -> AVAudioPCMBuffer? in
|
||||
guard let absd = self.formatDescription?.audioStreamBasicDescription else { return nil }
|
||||
guard let format = AVAudioFormat(standardFormatWithSampleRate: absd.mSampleRate, channels: absd.mChannelsPerFrame) else { return nil }
|
||||
return AVAudioPCMBuffer(pcmFormat: format, bufferListNoCopy: audioBufferList.unsafePointer)
|
||||
}
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,70 @@
|
||||
// KeychainHelper.swift
|
||||
// Secure storage helper for API keys and sensitive data
|
||||
|
||||
import Foundation
|
||||
import Security
|
||||
|
||||
/// Manages secure storage of sensitive data using the macOS Keychain
|
||||
class KeychainHelper {
|
||||
static let shared = KeychainHelper()
|
||||
|
||||
private let serviceName = "owen.notetaker"
|
||||
|
||||
private init() {}
|
||||
|
||||
/// Saves a string value to the keychain
|
||||
/// - Parameters:
|
||||
/// - value: The string value to save
|
||||
/// - key: The key to save the value under
|
||||
/// - Returns: True if the save was successful, false otherwise
|
||||
func save(_ value: String, forKey key: String) -> Bool {
|
||||
guard let data = value.data(using: .utf8) else { return false }
|
||||
|
||||
let query: [String: Any] = [
|
||||
kSecClass as String: kSecClassGenericPassword,
|
||||
kSecAttrAccount as String: key,
|
||||
kSecValueData as String: data,
|
||||
kSecAttrService as String: serviceName
|
||||
]
|
||||
|
||||
// Delete any existing item
|
||||
SecItemDelete(query as CFDictionary)
|
||||
|
||||
// Add the new item
|
||||
let status = SecItemAdd(query as CFDictionary, nil)
|
||||
return status == errSecSuccess
|
||||
}
|
||||
|
||||
/// Retrieves a string value from the keychain
|
||||
/// - Parameter key: The key to retrieve the value for
|
||||
/// - Returns: The string value if found, nil otherwise
|
||||
func get(forKey key: String) -> String? {
|
||||
let query: [String: Any] = [
|
||||
kSecClass as String: kSecClassGenericPassword,
|
||||
kSecAttrAccount as String: key,
|
||||
kSecAttrService as String: serviceName,
|
||||
kSecMatchLimit as String: kSecMatchLimitOne,
|
||||
kSecReturnData as String: true
|
||||
]
|
||||
|
||||
var item: AnyObject?
|
||||
let status = SecItemCopyMatching(query as CFDictionary, &item)
|
||||
|
||||
guard status == errSecSuccess, let data = item as? Data else { return nil }
|
||||
return String(data: data, encoding: .utf8)
|
||||
}
|
||||
|
||||
/// Deletes a value from the keychain
|
||||
/// - Parameter key: The key to delete
|
||||
/// - Returns: True if the deletion was successful, false otherwise
|
||||
func delete(forKey key: String) -> Bool {
|
||||
let query: [String: Any] = [
|
||||
kSecClass as String: kSecClassGenericPassword,
|
||||
kSecAttrAccount as String: key,
|
||||
kSecAttrService as String: serviceName
|
||||
]
|
||||
|
||||
let status = SecItemDelete(query as CFDictionary)
|
||||
return status == errSecSuccess
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,112 @@
|
||||
// LocalStorageManager.swift
|
||||
// Handles local storage of meetings and app data
|
||||
|
||||
import Foundation
|
||||
|
||||
/// Manages local file storage for meetings and app data
|
||||
class LocalStorageManager {
|
||||
static let shared = LocalStorageManager()
|
||||
|
||||
private let documentsDirectory: URL
|
||||
private let meetingsDirectory: URL
|
||||
|
||||
private init() {
|
||||
// Get the app's documents directory
|
||||
documentsDirectory = FileManager.default.urls(for: .documentDirectory,
|
||||
in: .userDomainMask).first!
|
||||
|
||||
// Create meetings subdirectory
|
||||
meetingsDirectory = documentsDirectory.appendingPathComponent("Meetings")
|
||||
|
||||
// Ensure directory exists
|
||||
try? FileManager.default.createDirectory(at: meetingsDirectory,
|
||||
withIntermediateDirectories: true)
|
||||
}
|
||||
|
||||
// MARK: - Meeting Management
|
||||
|
||||
/// Saves a meeting to local storage
|
||||
/// - Parameter meeting: The meeting to save
|
||||
/// - Returns: True if successful, false otherwise
|
||||
func saveMeeting(_ meeting: Meeting) -> Bool {
|
||||
let fileURL = meetingsDirectory.appendingPathComponent("\(meeting.id.uuidString).json")
|
||||
|
||||
do {
|
||||
let encoder = JSONEncoder()
|
||||
encoder.outputFormatting = .prettyPrinted
|
||||
encoder.dateEncodingStrategy = .iso8601
|
||||
|
||||
let data = try encoder.encode(meeting)
|
||||
try data.write(to: fileURL)
|
||||
|
||||
print("✅ Saved meeting: \(meeting.id)")
|
||||
return true
|
||||
} catch {
|
||||
print("❌ Failed to save meeting: \(error)")
|
||||
return false
|
||||
}
|
||||
}
|
||||
|
||||
/// Loads all meetings from local storage
|
||||
/// - Returns: Array of meetings, sorted by date (newest first)
|
||||
func loadMeetings() -> [Meeting] {
|
||||
do {
|
||||
let fileURLs = try FileManager.default.contentsOfDirectory(at: meetingsDirectory,
|
||||
includingPropertiesForKeys: nil)
|
||||
.filter { $0.pathExtension == "json" }
|
||||
|
||||
let decoder = JSONDecoder()
|
||||
decoder.dateDecodingStrategy = .iso8601
|
||||
|
||||
let meetings = fileURLs.compactMap { url -> Meeting? in
|
||||
guard let data = try? Data(contentsOf: url),
|
||||
let meeting = try? decoder.decode(Meeting.self, from: data) else {
|
||||
print("⚠️ Failed to decode meeting at: \(url)")
|
||||
return nil
|
||||
}
|
||||
return meeting
|
||||
}
|
||||
|
||||
return meetings.sorted { $0.date > $1.date }
|
||||
} catch {
|
||||
print("❌ Failed to load meetings: \(error)")
|
||||
return []
|
||||
}
|
||||
}
|
||||
|
||||
/// Deletes a meeting from local storage
|
||||
/// - Parameter meeting: The meeting to delete
|
||||
/// - Returns: True if successful, false otherwise
|
||||
func deleteMeeting(_ meeting: Meeting) -> Bool {
|
||||
let fileURL = meetingsDirectory.appendingPathComponent("\(meeting.id.uuidString).json")
|
||||
|
||||
do {
|
||||
try FileManager.default.removeItem(at: fileURL)
|
||||
print("✅ Deleted meeting: \(meeting.id)")
|
||||
return true
|
||||
} catch {
|
||||
print("❌ Failed to delete meeting: \(error)")
|
||||
return false
|
||||
}
|
||||
}
|
||||
|
||||
// MARK: - Settings Management
|
||||
|
||||
/// Saves non-sensitive settings to local storage
|
||||
/// - Parameter settings: The settings to save (sensitive data should use Keychain)
|
||||
func saveSettings(_ settings: Settings) -> Bool {
|
||||
// For now, all settings are stored in Keychain
|
||||
// This method is here for future non-sensitive settings
|
||||
return true
|
||||
}
|
||||
|
||||
/// Gets the app's documents directory URL
|
||||
var documentsDirectoryURL: URL {
|
||||
documentsDirectory
|
||||
}
|
||||
|
||||
/// Gets the meetings directory URL
|
||||
var meetingsDirectoryURL: URL {
|
||||
meetingsDirectory
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,21 @@
|
||||
import Foundation
|
||||
|
||||
struct Meeting: Codable, Identifiable, Hashable {
|
||||
let id: UUID
|
||||
let date: Date
|
||||
var transcript: String
|
||||
var userNotes: String
|
||||
var generatedNotes: String
|
||||
|
||||
init(id: UUID = UUID(),
|
||||
date: Date = Date(),
|
||||
transcript: String = "",
|
||||
userNotes: String = "",
|
||||
generatedNotes: String = "") {
|
||||
self.id = id
|
||||
self.date = date
|
||||
self.transcript = transcript
|
||||
self.userNotes = userNotes
|
||||
self.generatedNotes = generatedNotes
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,20 @@
|
||||
import Foundation
|
||||
|
||||
struct Settings: Codable {
|
||||
var deepgramKey: String
|
||||
var openAIKey: String
|
||||
var userBlurb: String
|
||||
var systemPrompt: String
|
||||
|
||||
static let defaultSystemPrompt = "Generate concise meeting notes from the transcript and any additional notes provided. Focus on key points, action items, and decisions made."
|
||||
|
||||
init(deepgramKey: String = "",
|
||||
openAIKey: String = "",
|
||||
userBlurb: String = "",
|
||||
systemPrompt: String = Settings.defaultSystemPrompt) {
|
||||
self.deepgramKey = deepgramKey
|
||||
self.openAIKey = openAIKey
|
||||
self.userBlurb = userBlurb
|
||||
self.systemPrompt = systemPrompt
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,59 @@
|
||||
// NotesGenerator.swift
|
||||
// Handles AI-powered note generation using OpenAI
|
||||
|
||||
import Foundation
|
||||
|
||||
/// Generates meeting notes using OpenAI API
|
||||
class NotesGenerator {
|
||||
static let shared = NotesGenerator()
|
||||
|
||||
private init() {}
|
||||
|
||||
/// Generates meeting notes from transcript and user notes
|
||||
/// - Parameters:
|
||||
/// - transcript: The meeting transcript
|
||||
/// - userNotes: Additional notes from the user
|
||||
/// - userBlurb: Information about the user for context
|
||||
/// - systemPrompt: The system prompt for AI generation
|
||||
/// - Returns: Generated meeting notes
|
||||
func generateNotes(transcript: String,
|
||||
userNotes: String,
|
||||
userBlurb: String,
|
||||
systemPrompt: String) async throws -> String {
|
||||
|
||||
// TODO: Implement OpenAI API integration
|
||||
// For now, return a placeholder
|
||||
|
||||
// Simulate API delay
|
||||
try await Task.sleep(nanoseconds: 2_000_000_000)
|
||||
|
||||
// Placeholder response
|
||||
return """
|
||||
# Meeting Notes
|
||||
|
||||
## Summary
|
||||
This is a placeholder for AI-generated meeting notes.
|
||||
|
||||
## Key Points
|
||||
- Transcript length: \(transcript.count) characters
|
||||
- User notes provided: \(!userNotes.isEmpty)
|
||||
|
||||
## Action Items
|
||||
- Implement OpenAI API integration
|
||||
- Add proper prompt engineering
|
||||
|
||||
## Next Steps
|
||||
Once OpenAI API is integrated, this will provide comprehensive meeting notes based on the transcript and any additional context provided.
|
||||
"""
|
||||
}
|
||||
|
||||
/// Validates if OpenAI API key is configured
|
||||
/// - Returns: True if API key exists, false otherwise
|
||||
func isConfigured() -> Bool {
|
||||
guard let key = KeychainHelper.shared.get(forKey: "openAIKey"),
|
||||
!key.isEmpty else {
|
||||
return false
|
||||
}
|
||||
return true
|
||||
}
|
||||
}
|
||||
@@ -1,298 +0,0 @@
|
||||
// SimpleAudioManager.swift
|
||||
// Audio manager for microphone and system audio capture
|
||||
|
||||
import AVFoundation
|
||||
import Foundation
|
||||
import SwiftUI
|
||||
import ScreenCaptureKit
|
||||
|
||||
class SimpleAudioManager: NSObject, ObservableObject {
|
||||
@Published var transcript = ""
|
||||
@Published var isRecording = false
|
||||
@Published var captureSystemAudio = true
|
||||
|
||||
private var audioEngine = AVAudioEngine()
|
||||
private var systemAudioEngine = AVAudioEngine()
|
||||
private var webSocketTask: URLSessionWebSocketTask?
|
||||
private let deepgramURL = URL(string: "wss://api.deepgram.com/v1/listen?encoding=linear16&sample_rate=16000&channels=1&interim_results=true")!
|
||||
|
||||
// ScreenCaptureKit properties
|
||||
private var stream: SCStream?
|
||||
private var streamConfiguration: SCStreamConfiguration?
|
||||
|
||||
func startRecording() {
|
||||
print("Starting recording...")
|
||||
|
||||
if captureSystemAudio {
|
||||
startSystemAudioCapture()
|
||||
} else {
|
||||
startMicrophoneOnly()
|
||||
}
|
||||
}
|
||||
|
||||
private func startMicrophoneOnly() {
|
||||
print("Starting microphone-only recording...")
|
||||
|
||||
// Set up audio engine for microphone capture
|
||||
let inputNode = audioEngine.inputNode
|
||||
let recordingFormat = inputNode.outputFormat(forBus: 0)
|
||||
|
||||
// Convert to the format Deepgram expects: 16kHz, 16-bit, mono
|
||||
let targetFormat = AVAudioFormat(commonFormat: .pcmFormatInt16,
|
||||
sampleRate: 16000,
|
||||
channels: 1,
|
||||
interleaved: false)!
|
||||
|
||||
let converter = AVAudioConverter(from: recordingFormat, to: targetFormat)!
|
||||
|
||||
inputNode.installTap(onBus: 0, bufferSize: 1024, format: recordingFormat) { [weak self] (buffer, time) in
|
||||
self?.processAudioBuffer(buffer, converter: converter, targetFormat: targetFormat)
|
||||
}
|
||||
|
||||
audioEngine.prepare()
|
||||
|
||||
do {
|
||||
try audioEngine.start()
|
||||
DispatchQueue.main.async {
|
||||
self.isRecording = true
|
||||
}
|
||||
connectToDeepgram()
|
||||
print("Microphone recording started")
|
||||
} catch {
|
||||
print("Failed to start audio engine: \(error)")
|
||||
}
|
||||
}
|
||||
|
||||
private func startSystemAudioCapture() {
|
||||
print("Starting system audio capture...")
|
||||
|
||||
Task {
|
||||
do {
|
||||
// Request screen recording permission
|
||||
let canRecord = try await SCShareableContent.excludingDesktopWindows(false, onScreenWindowsOnly: true)
|
||||
|
||||
// Get the main display
|
||||
guard let display = canRecord.displays.first else {
|
||||
print("❌ No display found")
|
||||
return
|
||||
}
|
||||
|
||||
// Configure stream for audio only
|
||||
let configuration = SCStreamConfiguration()
|
||||
configuration.capturesAudio = true
|
||||
configuration.excludesCurrentProcessAudio = false
|
||||
configuration.sampleRate = 16000
|
||||
configuration.channelCount = 1
|
||||
|
||||
// Create content filter (we only want audio, but need to include display)
|
||||
let filter = SCContentFilter(display: display, excludingWindows: [])
|
||||
|
||||
// Create and start stream
|
||||
let stream = SCStream(filter: filter, configuration: configuration, delegate: self)
|
||||
|
||||
try await stream.startCapture()
|
||||
|
||||
self.stream = stream
|
||||
self.streamConfiguration = configuration
|
||||
|
||||
DispatchQueue.main.async {
|
||||
self.isRecording = true
|
||||
}
|
||||
|
||||
connectToDeepgram()
|
||||
print("System audio capture started")
|
||||
|
||||
} catch {
|
||||
print("❌ Failed to start system audio capture: \(error)")
|
||||
// Fallback to microphone only
|
||||
startMicrophoneOnly()
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func stopRecording() {
|
||||
print("Stopping recording...")
|
||||
|
||||
// Stop system audio capture
|
||||
if let stream = stream {
|
||||
Task {
|
||||
try? await stream.stopCapture()
|
||||
}
|
||||
self.stream = nil
|
||||
}
|
||||
|
||||
// Stop microphone capture
|
||||
if audioEngine.isRunning {
|
||||
audioEngine.stop()
|
||||
audioEngine.inputNode.removeTap(onBus: 0)
|
||||
}
|
||||
|
||||
// Close WebSocket
|
||||
webSocketTask?.cancel(with: .normalClosure, reason: nil)
|
||||
webSocketTask = nil
|
||||
|
||||
DispatchQueue.main.async {
|
||||
self.isRecording = false
|
||||
}
|
||||
print("Recording stopped")
|
||||
}
|
||||
|
||||
private func processAudioBuffer(_ buffer: AVAudioPCMBuffer, converter: AVAudioConverter, targetFormat: AVAudioFormat) {
|
||||
// Create output buffer for conversion
|
||||
let outputFrameCapacity = AVAudioFrameCount(Double(buffer.frameLength) * targetFormat.sampleRate / buffer.format.sampleRate)
|
||||
guard let outputBuffer = AVAudioPCMBuffer(pcmFormat: targetFormat, frameCapacity: outputFrameCapacity) else {
|
||||
print("❌ Failed to create output buffer")
|
||||
return
|
||||
}
|
||||
|
||||
var error: NSError?
|
||||
let status = converter.convert(to: outputBuffer, error: &error) { _, outStatus in
|
||||
outStatus.pointee = .haveData
|
||||
return buffer
|
||||
}
|
||||
|
||||
guard status == .haveData, error == nil else {
|
||||
print("❌ Audio conversion failed: \(error?.localizedDescription ?? "Unknown error")")
|
||||
return
|
||||
}
|
||||
|
||||
// Convert to Data for Deepgram
|
||||
guard let channelData = outputBuffer.int16ChannelData?[0] else {
|
||||
print("❌ No channel data available")
|
||||
return
|
||||
}
|
||||
|
||||
let frameCount = Int(outputBuffer.frameLength)
|
||||
let data = Data(bytes: channelData, count: frameCount * 2)
|
||||
|
||||
print("🎵 Sending audio data: \(data.count) bytes")
|
||||
sendAudioData(data)
|
||||
}
|
||||
|
||||
private func connectToDeepgram() {
|
||||
guard let key = KeychainHelper.shared.get(forKey: "deepgramKey"), !key.isEmpty else {
|
||||
print("❌ No Deepgram key found")
|
||||
return
|
||||
}
|
||||
|
||||
print("🔑 Using Deepgram key: \(String(key.prefix(8)))...")
|
||||
|
||||
let session = URLSession(configuration: .default)
|
||||
var request = URLRequest(url: deepgramURL)
|
||||
request.addValue("Token \(key)", forHTTPHeaderField: "Authorization")
|
||||
|
||||
webSocketTask = session.webSocketTask(with: request)
|
||||
webSocketTask?.resume()
|
||||
receiveMessage()
|
||||
print("🌐 Attempting to connect to Deepgram...")
|
||||
}
|
||||
|
||||
private func sendAudioData(_ data: Data) {
|
||||
guard webSocketTask?.state == .running else {
|
||||
print("⚠️ WebSocket not connected, skipping audio data")
|
||||
return
|
||||
}
|
||||
|
||||
webSocketTask?.send(.data(data)) { error in
|
||||
if let error = error {
|
||||
print("❌ Send error: \(error)")
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
private func receiveMessage() {
|
||||
webSocketTask?.receive { [weak self] result in
|
||||
switch result {
|
||||
case .success(let message):
|
||||
switch message {
|
||||
case .string(let text):
|
||||
print("📝 Received: \(text)")
|
||||
self?.parseTranscription(text)
|
||||
case .data:
|
||||
break
|
||||
@unknown default:
|
||||
break
|
||||
}
|
||||
self?.receiveMessage() // Continue receiving
|
||||
case .failure(let error):
|
||||
print("❌ Receive error: \(error)")
|
||||
// Try to reconnect after a delay
|
||||
DispatchQueue.main.asyncAfter(deadline: .now() + 2.0) {
|
||||
if self?.isRecording == true {
|
||||
print("🔄 Attempting to reconnect...")
|
||||
self?.connectToDeepgram()
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
private func parseTranscription(_ text: String) {
|
||||
guard let data = text.data(using: .utf8),
|
||||
let json = try? JSONSerialization.jsonObject(with: data) as? [String: Any],
|
||||
let type = json["type"] as? String, type == "Results",
|
||||
let channel = json["channel"] as? [String: Any],
|
||||
let alternatives = channel["alternatives"] as? [[String: Any]],
|
||||
let alt = alternatives.first,
|
||||
let transcript = alt["transcript"] as? String,
|
||||
!transcript.isEmpty else { return }
|
||||
|
||||
DispatchQueue.main.async {
|
||||
if let isFinal = json["is_final"] as? Bool, isFinal {
|
||||
self.transcript += transcript + " "
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// MARK: - SCStreamDelegate
|
||||
extension SimpleAudioManager: SCStreamDelegate {
|
||||
func stream(_ stream: SCStream, didOutputSampleBuffer sampleBuffer: CMSampleBuffer, of type: SCStreamOutputType) {
|
||||
guard type == .audio else { return }
|
||||
|
||||
// Convert CMSampleBuffer to AVAudioPCMBuffer
|
||||
guard let formatDescription = CMSampleBufferGetFormatDescription(sampleBuffer) else { return }
|
||||
let audioStreamBasicDescription = CMAudioFormatDescriptionGetStreamBasicDescription(formatDescription)
|
||||
|
||||
guard let audioStreamBasicDescription = audioStreamBasicDescription else { return }
|
||||
|
||||
let format = AVAudioFormat(streamDescription: audioStreamBasicDescription)
|
||||
guard let format = format else { return }
|
||||
|
||||
let frameCount = CMSampleBufferGetNumSamples(sampleBuffer)
|
||||
guard let buffer = AVAudioPCMBuffer(pcmFormat: format, frameCapacity: AVAudioFrameCount(frameCount)) else { return }
|
||||
|
||||
// Copy audio data
|
||||
var blockBuffer: CMBlockBuffer?
|
||||
var audioBufferListOut = AudioBufferList()
|
||||
var audioBufferListSize = MemoryLayout<AudioBufferList>.size
|
||||
|
||||
let status = CMSampleBufferGetAudioBufferListWithRetainedBlockBuffer(
|
||||
sampleBuffer,
|
||||
bufferListSizeNeededOut: &audioBufferListSize,
|
||||
bufferListOut: &audioBufferListOut,
|
||||
bufferListSize: audioBufferListSize,
|
||||
blockBufferAllocator: nil,
|
||||
blockBufferMemoryAllocator: nil,
|
||||
flags: 0,
|
||||
blockBufferOut: &blockBuffer
|
||||
)
|
||||
|
||||
guard status == noErr else { return }
|
||||
|
||||
buffer.frameLength = AVAudioFrameCount(frameCount)
|
||||
|
||||
// Convert to target format and send
|
||||
let targetFormat = AVAudioFormat(commonFormat: .pcmFormatInt16,
|
||||
sampleRate: 16000,
|
||||
channels: 1,
|
||||
interleaved: false)!
|
||||
|
||||
let converter = AVAudioConverter(from: format, to: targetFormat)!
|
||||
processAudioBuffer(buffer, converter: converter, targetFormat: targetFormat)
|
||||
}
|
||||
|
||||
func stream(_ stream: SCStream, didStopWithError error: Error) {
|
||||
print("❌ Stream stopped with error: \(error)")
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,35 @@
|
||||
import Foundation
|
||||
import SwiftUI
|
||||
|
||||
@MainActor
|
||||
class MeetingListViewModel: ObservableObject {
|
||||
@Published var meetings: [Meeting] = []
|
||||
@Published var isLoading = false
|
||||
@Published var errorMessage: String?
|
||||
|
||||
init() {
|
||||
loadMeetings()
|
||||
}
|
||||
|
||||
func loadMeetings() {
|
||||
isLoading = true
|
||||
errorMessage = nil
|
||||
|
||||
DispatchQueue.main.async { [weak self] in
|
||||
self?.meetings = LocalStorageManager.shared.loadMeetings()
|
||||
self?.isLoading = false
|
||||
}
|
||||
}
|
||||
|
||||
func deleteMeeting(_ meeting: Meeting) {
|
||||
meetings.removeAll { $0.id == meeting.id }
|
||||
_ = LocalStorageManager.shared.deleteMeeting(meeting)
|
||||
}
|
||||
|
||||
func createNewMeeting() -> Meeting {
|
||||
let newMeeting = Meeting()
|
||||
meetings.insert(newMeeting, at: 0)
|
||||
_ = LocalStorageManager.shared.saveMeeting(newMeeting)
|
||||
return newMeeting
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,85 @@
|
||||
import Foundation
|
||||
import SwiftUI
|
||||
import Combine
|
||||
|
||||
@MainActor
|
||||
class MeetingViewModel: ObservableObject {
|
||||
@Published var meeting: Meeting
|
||||
@Published var isGeneratingNotes = false
|
||||
@Published var errorMessage: String?
|
||||
@Published var isRecording = false
|
||||
|
||||
private let audioManager = AudioManager()
|
||||
private var cancellables = Set<AnyCancellable>()
|
||||
|
||||
init(meeting: Meeting = Meeting()) {
|
||||
self.meeting = meeting
|
||||
|
||||
// Update meeting transcript when audio manager transcript changes
|
||||
audioManager.$transcript
|
||||
.sink { [weak self] newTranscript in
|
||||
self?.meeting.transcript = newTranscript
|
||||
}
|
||||
.store(in: &cancellables)
|
||||
|
||||
// Update isRecording when audio manager recording state changes
|
||||
audioManager.$isRecording
|
||||
.sink { [weak self] isRecording in
|
||||
self?.isRecording = isRecording
|
||||
}
|
||||
.store(in: &cancellables)
|
||||
}
|
||||
|
||||
var captureSystemAudio: Bool {
|
||||
get { audioManager.captureSystemAudio }
|
||||
set { audioManager.captureSystemAudio = newValue }
|
||||
}
|
||||
|
||||
func startRecording() {
|
||||
audioManager.startRecording()
|
||||
}
|
||||
|
||||
func stopRecording() {
|
||||
audioManager.stopRecording()
|
||||
saveMeeting()
|
||||
}
|
||||
|
||||
func generateNotes() async {
|
||||
isGeneratingNotes = true
|
||||
errorMessage = nil
|
||||
|
||||
do {
|
||||
// Load settings for generation
|
||||
let userBlurb = KeychainHelper.shared.get(forKey: "userBlurb") ?? ""
|
||||
let systemPrompt = KeychainHelper.shared.get(forKey: "systemPrompt") ?? Settings.defaultSystemPrompt
|
||||
|
||||
// Generate notes
|
||||
meeting.generatedNotes = try await NotesGenerator.shared.generateNotes(
|
||||
transcript: meeting.transcript,
|
||||
userNotes: meeting.userNotes,
|
||||
userBlurb: userBlurb,
|
||||
systemPrompt: systemPrompt
|
||||
)
|
||||
|
||||
saveMeeting()
|
||||
} catch {
|
||||
errorMessage = "Failed to generate notes: \(error.localizedDescription)"
|
||||
}
|
||||
|
||||
isGeneratingNotes = false
|
||||
}
|
||||
|
||||
func saveMeeting() {
|
||||
_ = LocalStorageManager.shared.saveMeeting(meeting)
|
||||
}
|
||||
|
||||
func copyTranscript() {
|
||||
NSPasteboard.general.clearContents()
|
||||
NSPasteboard.general.setString(meeting.transcript, forType: .string)
|
||||
}
|
||||
|
||||
func copyNotes() {
|
||||
NSPasteboard.general.clearContents()
|
||||
NSPasteboard.general.setString(meeting.generatedNotes, forType: .string)
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,51 @@
|
||||
import Foundation
|
||||
import SwiftUI
|
||||
|
||||
@MainActor
|
||||
class SettingsViewModel: ObservableObject {
|
||||
@Published var settings: Settings
|
||||
@Published var isSaving = false
|
||||
@Published var saveSuccessful = false
|
||||
@Published var errorMessage: String?
|
||||
|
||||
init() {
|
||||
self.settings = Settings()
|
||||
loadSettings()
|
||||
}
|
||||
|
||||
func loadSettings() {
|
||||
// Load API keys from Keychain
|
||||
settings.deepgramKey = KeychainHelper.shared.get(forKey: "deepgramKey") ?? ""
|
||||
settings.openAIKey = KeychainHelper.shared.get(forKey: "openAIKey") ?? ""
|
||||
settings.userBlurb = KeychainHelper.shared.get(forKey: "userBlurb") ?? ""
|
||||
settings.systemPrompt = KeychainHelper.shared.get(forKey: "systemPrompt") ?? Settings.defaultSystemPrompt
|
||||
}
|
||||
|
||||
func saveSettings() {
|
||||
isSaving = true
|
||||
errorMessage = nil
|
||||
saveSuccessful = false
|
||||
|
||||
// Save to Keychain
|
||||
let deepgramSaved = KeychainHelper.shared.save(settings.deepgramKey, forKey: "deepgramKey")
|
||||
let openAISaved = KeychainHelper.shared.save(settings.openAIKey, forKey: "openAIKey")
|
||||
let blurbSaved = KeychainHelper.shared.save(settings.userBlurb, forKey: "userBlurb")
|
||||
let promptSaved = KeychainHelper.shared.save(settings.systemPrompt, forKey: "systemPrompt")
|
||||
|
||||
if deepgramSaved && openAISaved && blurbSaved && promptSaved {
|
||||
saveSuccessful = true
|
||||
// Show success briefly
|
||||
DispatchQueue.main.asyncAfter(deadline: .now() + 2) { [weak self] in
|
||||
self?.saveSuccessful = false
|
||||
}
|
||||
} else {
|
||||
errorMessage = "Failed to save some settings. Please try again."
|
||||
}
|
||||
|
||||
isSaving = false
|
||||
}
|
||||
|
||||
func resetToDefaults() {
|
||||
settings.systemPrompt = Settings.defaultSystemPrompt
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,136 @@
|
||||
import SwiftUI
|
||||
|
||||
struct MeetingDetailView: View {
|
||||
@StateObject private var viewModel: MeetingViewModel
|
||||
|
||||
init(meeting: Meeting) {
|
||||
self._viewModel = StateObject(wrappedValue: MeetingViewModel(meeting: meeting))
|
||||
}
|
||||
|
||||
var body: some View {
|
||||
VStack(spacing: 20) {
|
||||
// Recording Controls
|
||||
VStack(spacing: 12) {
|
||||
Toggle("Capture System Audio", isOn: $viewModel.captureSystemAudio)
|
||||
.disabled(viewModel.isRecording)
|
||||
|
||||
HStack {
|
||||
Button {
|
||||
viewModel.startRecording()
|
||||
} label: {
|
||||
Label("Start Recording", systemImage: "record.circle")
|
||||
}
|
||||
.disabled(viewModel.isRecording)
|
||||
.buttonStyle(.borderedProminent)
|
||||
|
||||
Button {
|
||||
viewModel.stopRecording()
|
||||
} label: {
|
||||
Label("Stop Recording", systemImage: "stop.circle")
|
||||
}
|
||||
.disabled(!viewModel.isRecording)
|
||||
.buttonStyle(.bordered)
|
||||
}
|
||||
|
||||
if viewModel.isRecording {
|
||||
Label("Recording...", systemImage: "dot.radiowaves.left.and.right")
|
||||
.foregroundColor(.red)
|
||||
}
|
||||
}
|
||||
.padding()
|
||||
.background(Color.gray.opacity(0.1))
|
||||
.cornerRadius(10)
|
||||
|
||||
// Transcript and Notes
|
||||
HStack(spacing: 20) {
|
||||
// Transcript Column
|
||||
VStack(alignment: .leading, spacing: 8) {
|
||||
HStack {
|
||||
Text("Live Transcript")
|
||||
.font(.headline)
|
||||
Spacer()
|
||||
Button {
|
||||
viewModel.copyTranscript()
|
||||
} label: {
|
||||
Label("Copy", systemImage: "doc.on.doc")
|
||||
}
|
||||
}
|
||||
|
||||
ScrollView {
|
||||
Text(viewModel.meeting.transcript.isEmpty ? "Transcript will appear here..." : viewModel.meeting.transcript)
|
||||
.frame(maxWidth: .infinity, alignment: .leading)
|
||||
.padding()
|
||||
}
|
||||
.frame(maxHeight: .infinity)
|
||||
.background(Color.gray.opacity(0.05))
|
||||
.cornerRadius(8)
|
||||
}
|
||||
|
||||
// Notes Column
|
||||
VStack(alignment: .leading, spacing: 8) {
|
||||
Text("Your Notes")
|
||||
.font(.headline)
|
||||
|
||||
TextEditor(text: $viewModel.meeting.userNotes)
|
||||
.font(.body)
|
||||
.scrollContentBackground(.hidden)
|
||||
.background(Color.gray.opacity(0.05))
|
||||
.cornerRadius(8)
|
||||
}
|
||||
}
|
||||
.frame(maxHeight: 300)
|
||||
|
||||
// Generated Notes Section
|
||||
VStack(alignment: .leading, spacing: 8) {
|
||||
HStack {
|
||||
Text("Generated Notes")
|
||||
.font(.headline)
|
||||
Spacer()
|
||||
|
||||
if viewModel.isGeneratingNotes {
|
||||
ProgressView()
|
||||
.scaleEffect(0.7)
|
||||
} else {
|
||||
Button {
|
||||
Task {
|
||||
await viewModel.generateNotes()
|
||||
}
|
||||
} label: {
|
||||
Label("Generate", systemImage: "sparkles")
|
||||
}
|
||||
.disabled(viewModel.meeting.transcript.isEmpty)
|
||||
|
||||
Button {
|
||||
viewModel.copyNotes()
|
||||
} label: {
|
||||
Label("Copy", systemImage: "doc.on.doc")
|
||||
}
|
||||
.disabled(viewModel.meeting.generatedNotes.isEmpty)
|
||||
}
|
||||
}
|
||||
|
||||
TextEditor(text: $viewModel.meeting.generatedNotes)
|
||||
.font(.body)
|
||||
.scrollContentBackground(.hidden)
|
||||
.background(Color.gray.opacity(0.05))
|
||||
.cornerRadius(8)
|
||||
.frame(minHeight: 150)
|
||||
}
|
||||
}
|
||||
.padding()
|
||||
.navigationTitle("Meeting Notes")
|
||||
.alert("Error", isPresented: .constant(viewModel.errorMessage != nil)) {
|
||||
Button("OK") {
|
||||
viewModel.errorMessage = nil
|
||||
}
|
||||
} message: {
|
||||
Text(viewModel.errorMessage ?? "")
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
#Preview {
|
||||
NavigationStack {
|
||||
MeetingDetailView(meeting: Meeting())
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,78 @@
|
||||
import SwiftUI
|
||||
|
||||
struct MeetingListView: View {
|
||||
@StateObject private var viewModel = MeetingListViewModel()
|
||||
@State private var navigationPath = NavigationPath()
|
||||
|
||||
var body: some View {
|
||||
NavigationStack(path: $navigationPath) {
|
||||
List {
|
||||
if viewModel.meetings.isEmpty && !viewModel.isLoading {
|
||||
ContentUnavailableView(
|
||||
"No Meetings Yet",
|
||||
systemImage: "mic.slash",
|
||||
description: Text("Start a new meeting to begin transcribing")
|
||||
)
|
||||
} else {
|
||||
ForEach(viewModel.meetings) { meeting in
|
||||
NavigationLink(value: meeting) {
|
||||
MeetingRowView(meeting: meeting)
|
||||
}
|
||||
}
|
||||
.onDelete { indexSet in
|
||||
for index in indexSet {
|
||||
viewModel.deleteMeeting(viewModel.meetings[index])
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
.navigationTitle("Meetings")
|
||||
.toolbar {
|
||||
ToolbarItem(placement: .primaryAction) {
|
||||
Button {
|
||||
let newMeeting = viewModel.createNewMeeting()
|
||||
navigationPath.append(newMeeting)
|
||||
} label: {
|
||||
Label("New Meeting", systemImage: "plus")
|
||||
}
|
||||
}
|
||||
}
|
||||
.navigationDestination(for: Meeting.self) { meeting in
|
||||
MeetingDetailView(meeting: meeting)
|
||||
}
|
||||
}
|
||||
.overlay {
|
||||
if viewModel.isLoading {
|
||||
ProgressView("Loading meetings...")
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
struct MeetingRowView: View {
|
||||
let meeting: Meeting
|
||||
|
||||
var body: some View {
|
||||
VStack(alignment: .leading, spacing: 4) {
|
||||
Text(meeting.date, style: .date)
|
||||
.font(.headline)
|
||||
|
||||
if !meeting.transcript.isEmpty {
|
||||
Text(meeting.transcript)
|
||||
.font(.caption)
|
||||
.foregroundColor(.secondary)
|
||||
.lineLimit(2)
|
||||
} else {
|
||||
Text("No transcript")
|
||||
.font(.caption)
|
||||
.foregroundColor(.secondary)
|
||||
.italic()
|
||||
}
|
||||
}
|
||||
.padding(.vertical, 4)
|
||||
}
|
||||
}
|
||||
|
||||
#Preview {
|
||||
MeetingListView()
|
||||
}
|
||||
@@ -0,0 +1,96 @@
|
||||
import SwiftUI
|
||||
|
||||
struct SettingsView: View {
|
||||
@StateObject private var viewModel = SettingsViewModel()
|
||||
@Environment(\.dismiss) private var dismiss
|
||||
|
||||
var body: some View {
|
||||
NavigationStack {
|
||||
Form {
|
||||
Section("API Keys") {
|
||||
SecureField("Deepgram API Key", text: $viewModel.settings.deepgramKey)
|
||||
.textFieldStyle(.roundedBorder)
|
||||
|
||||
SecureField("OpenAI API Key", text: $viewModel.settings.openAIKey)
|
||||
.textFieldStyle(.roundedBorder)
|
||||
}
|
||||
|
||||
Section("Personal Information") {
|
||||
VStack(alignment: .leading) {
|
||||
Text("About Yourself")
|
||||
.font(.caption)
|
||||
.foregroundColor(.secondary)
|
||||
TextEditor(text: $viewModel.settings.userBlurb)
|
||||
.frame(minHeight: 80)
|
||||
.scrollContentBackground(.hidden)
|
||||
.background(Color.gray.opacity(0.05))
|
||||
.cornerRadius(8)
|
||||
}
|
||||
}
|
||||
|
||||
Section("AI Generation") {
|
||||
VStack(alignment: .leading) {
|
||||
HStack {
|
||||
Text("System Prompt")
|
||||
.font(.caption)
|
||||
.foregroundColor(.secondary)
|
||||
Spacer()
|
||||
Button("Reset to Default") {
|
||||
viewModel.resetToDefaults()
|
||||
}
|
||||
.font(.caption)
|
||||
}
|
||||
TextEditor(text: $viewModel.settings.systemPrompt)
|
||||
.frame(minHeight: 120)
|
||||
.scrollContentBackground(.hidden)
|
||||
.background(Color.gray.opacity(0.05))
|
||||
.cornerRadius(8)
|
||||
}
|
||||
}
|
||||
}
|
||||
.navigationTitle("Settings")
|
||||
.toolbar {
|
||||
ToolbarItem(placement: .cancellationAction) {
|
||||
Button("Cancel") {
|
||||
dismiss()
|
||||
}
|
||||
}
|
||||
|
||||
ToolbarItem(placement: .confirmationAction) {
|
||||
Button("Save") {
|
||||
viewModel.saveSettings()
|
||||
}
|
||||
.disabled(viewModel.isSaving)
|
||||
}
|
||||
}
|
||||
.overlay {
|
||||
if viewModel.saveSuccessful {
|
||||
VStack {
|
||||
Spacer()
|
||||
HStack {
|
||||
Image(systemName: "checkmark.circle.fill")
|
||||
.foregroundColor(.green)
|
||||
Text("Settings saved successfully")
|
||||
}
|
||||
.padding()
|
||||
.background(Color.green.opacity(0.1))
|
||||
.cornerRadius(10)
|
||||
.padding()
|
||||
}
|
||||
.transition(.move(edge: .bottom).combined(with: .opacity))
|
||||
}
|
||||
}
|
||||
.alert("Error", isPresented: .constant(viewModel.errorMessage != nil)) {
|
||||
Button("OK") {
|
||||
viewModel.errorMessage = nil
|
||||
}
|
||||
} message: {
|
||||
Text(viewModel.errorMessage ?? "")
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
#Preview {
|
||||
SettingsView()
|
||||
}
|
||||
@@ -4,17 +4,17 @@
|
||||
<dict>
|
||||
<key>com.apple.security.app-sandbox</key>
|
||||
<true/>
|
||||
<key>com.apple.security.application-groups</key>
|
||||
<array>
|
||||
<string>group.owen.notetaker</string>
|
||||
</array>
|
||||
<key>com.apple.security.device.audio-input</key>
|
||||
<true/>
|
||||
<key>com.apple.security.device.screen-capture</key>
|
||||
<true/>
|
||||
<key>com.apple.security.files.user-selected.read-only</key>
|
||||
<true/>
|
||||
<key>com.apple.security.network.client</key>
|
||||
<true/>
|
||||
<key>com.apple.security.application-groups</key>
|
||||
<array>
|
||||
<string>group.owen.notetaker</string>
|
||||
</array>
|
||||
<key>com.apple.security.device.screen-recording</key>
|
||||
<true/>
|
||||
</dict>
|
||||
</plist>
|
||||
|
||||
@@ -12,6 +12,8 @@ struct notetakerApp: App {
|
||||
var body: some Scene {
|
||||
WindowGroup {
|
||||
ContentView()
|
||||
.frame(minWidth: 800, minHeight: 600)
|
||||
}
|
||||
.windowResizability(.contentSize)
|
||||
}
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user