VoiceWake: capture utterance and add prefix
This commit is contained in:
@@ -178,15 +178,24 @@ final class AppState: ObservableObject {
|
||||
Task { await VoiceWakeRuntime.shared.refresh(state: self) }
|
||||
}
|
||||
|
||||
func triggerVoiceEars(ttl: TimeInterval = 5) {
|
||||
func triggerVoiceEars(ttl: TimeInterval? = 5) {
|
||||
self.earBoostTask?.cancel()
|
||||
self.earBoostActive = true
|
||||
|
||||
guard let ttl else { return }
|
||||
|
||||
self.earBoostTask = Task { [weak self] in
|
||||
try? await Task.sleep(nanoseconds: UInt64(ttl * 1_000_000_000))
|
||||
await MainActor.run { [weak self] in self?.earBoostActive = false }
|
||||
}
|
||||
}
|
||||
|
||||
func stopVoiceEars() {
|
||||
self.earBoostTask?.cancel()
|
||||
self.earBoostTask = nil
|
||||
self.earBoostActive = false
|
||||
}
|
||||
|
||||
func setVoiceWakeEnabled(_ enabled: Bool) async {
|
||||
guard voiceWakeSupported else {
|
||||
self.swabbleEnabled = false
|
||||
|
||||
@@ -28,6 +28,19 @@ enum VoiceWakeForwarder {
|
||||
private static let cliSearchCandidates = ["clawdis-mac"] + cliHelperSearchPaths.map { "\($0)/clawdis-mac" }
|
||||
private static let cliCache = CLICache()
|
||||
|
||||
static func prefixedTranscript(_ transcript: String, machineName: String? = nil) -> String {
|
||||
let resolvedMachine = machineName
|
||||
.flatMap { name -> String? in
|
||||
let trimmed = name.trimmingCharacters(in: .whitespacesAndNewlines)
|
||||
return trimmed.isEmpty ? nil : trimmed
|
||||
}
|
||||
?? Host.current().localizedName
|
||||
?? ProcessInfo.processInfo.hostName
|
||||
|
||||
let safeMachine = resolvedMachine.isEmpty ? "this Mac" : resolvedMachine
|
||||
return "User talked via voice recognition on \(safeMachine) - repeat prompt first + remember some words might be incorrectly transcribed.\n\n\(transcript)"
|
||||
}
|
||||
|
||||
static func clearCliCache() {
|
||||
self.cliCache.set(nil)
|
||||
}
|
||||
|
||||
@@ -14,9 +14,18 @@ actor VoiceWakeRuntime {
|
||||
private var recognitionRequest: SFSpeechAudioBufferRecognitionRequest?
|
||||
private var recognitionTask: SFSpeechRecognitionTask?
|
||||
private var lastHeard: Date?
|
||||
private var captureStartedAt: Date?
|
||||
private var captureTask: Task<Void, Never>?
|
||||
private var capturedTranscript: String = ""
|
||||
private var isCapturing: Bool = false
|
||||
private var cooldownUntil: Date?
|
||||
private var currentConfig: RuntimeConfig?
|
||||
|
||||
// Tunables
|
||||
private let silenceWindow: TimeInterval = 1.0
|
||||
private let captureHardStop: TimeInterval = 8.0
|
||||
private let debounceAfterSend: TimeInterval = 0.35
|
||||
|
||||
struct RuntimeConfig: Equatable {
|
||||
let triggers: [String]
|
||||
let micID: String?
|
||||
@@ -95,6 +104,11 @@ actor VoiceWakeRuntime {
|
||||
}
|
||||
|
||||
private func stop() {
|
||||
self.captureTask?.cancel()
|
||||
self.captureTask = nil
|
||||
self.isCapturing = false
|
||||
self.capturedTranscript = ""
|
||||
self.captureStartedAt = nil
|
||||
self.recognitionTask?.cancel()
|
||||
self.recognitionTask = nil
|
||||
self.recognitionRequest?.endAudio()
|
||||
@@ -120,21 +134,22 @@ actor VoiceWakeRuntime {
|
||||
}
|
||||
|
||||
guard let transcript else { return }
|
||||
if !transcript.isEmpty { self.lastHeard = Date() }
|
||||
|
||||
let now = Date()
|
||||
if !transcript.isEmpty {
|
||||
self.lastHeard = now
|
||||
if self.isCapturing {
|
||||
self.capturedTranscript = transcript
|
||||
}
|
||||
}
|
||||
|
||||
if self.isCapturing { return }
|
||||
|
||||
if Self.matches(text: transcript, triggers: config.triggers) {
|
||||
let now = Date()
|
||||
if let cooldown = cooldownUntil, now < cooldown {
|
||||
return
|
||||
}
|
||||
self.cooldownUntil = now.addingTimeInterval(2.5)
|
||||
await MainActor.run { AppStateStore.shared.triggerVoiceEars() }
|
||||
let forwardConfig = await MainActor.run { AppStateStore.shared.voiceWakeForwardConfig }
|
||||
if forwardConfig.enabled {
|
||||
Task.detached {
|
||||
await VoiceWakeForwarder.forward(transcript: transcript, config: forwardConfig)
|
||||
}
|
||||
}
|
||||
await self.beginCapture(transcript: transcript, config: config)
|
||||
}
|
||||
}
|
||||
|
||||
@@ -149,6 +164,77 @@ actor VoiceWakeRuntime {
|
||||
return false
|
||||
}
|
||||
|
||||
private func beginCapture(transcript: String, config: RuntimeConfig) async {
|
||||
self.isCapturing = true
|
||||
self.capturedTranscript = transcript
|
||||
self.captureStartedAt = Date()
|
||||
self.cooldownUntil = nil
|
||||
|
||||
await MainActor.run { AppStateStore.shared.triggerVoiceEars(ttl: nil) }
|
||||
|
||||
self.captureTask?.cancel()
|
||||
self.captureTask = Task { [weak self] in
|
||||
guard let self else { return }
|
||||
await self.monitorCapture(config: config)
|
||||
}
|
||||
}
|
||||
|
||||
private func monitorCapture(config: RuntimeConfig) async {
|
||||
let start = self.captureStartedAt ?? Date()
|
||||
let hardStop = start.addingTimeInterval(self.captureHardStop)
|
||||
|
||||
while self.isCapturing {
|
||||
let now = Date()
|
||||
if now >= hardStop {
|
||||
await self.finalizeCapture(config: config)
|
||||
return
|
||||
}
|
||||
|
||||
if let last = self.lastHeard, now.timeIntervalSince(last) >= self.silenceWindow {
|
||||
await self.finalizeCapture(config: config)
|
||||
return
|
||||
}
|
||||
|
||||
try? await Task.sleep(nanoseconds: 200_000_000)
|
||||
}
|
||||
}
|
||||
|
||||
private func finalizeCapture(config: RuntimeConfig) async {
|
||||
guard self.isCapturing else { return }
|
||||
self.isCapturing = false
|
||||
self.captureTask?.cancel()
|
||||
self.captureTask = nil
|
||||
|
||||
let finalTranscript = self.capturedTranscript.trimmingCharacters(in: .whitespacesAndNewlines)
|
||||
self.capturedTranscript = ""
|
||||
self.captureStartedAt = nil
|
||||
self.lastHeard = nil
|
||||
|
||||
await MainActor.run { AppStateStore.shared.stopVoiceEars() }
|
||||
|
||||
if !finalTranscript.isEmpty {
|
||||
await self.send(transcript: finalTranscript, config: config)
|
||||
}
|
||||
|
||||
self.cooldownUntil = Date().addingTimeInterval(self.debounceAfterSend)
|
||||
|
||||
// Restart the recognizer so we listen for the next trigger with a clean buffer.
|
||||
let current = self.currentConfig
|
||||
self.stop()
|
||||
if let current { await self.start(with: current) }
|
||||
}
|
||||
|
||||
private func send(transcript: String, config: RuntimeConfig) async {
|
||||
let forwardConfig = await MainActor.run { AppStateStore.shared.voiceWakeForwardConfig }
|
||||
guard forwardConfig.enabled else { return }
|
||||
|
||||
let payload = VoiceWakeForwarder.prefixedTranscript(transcript)
|
||||
|
||||
Task.detached {
|
||||
await VoiceWakeForwarder.forward(transcript: payload, config: forwardConfig)
|
||||
}
|
||||
}
|
||||
|
||||
#if DEBUG
|
||||
static func _testMatches(text: String, triggers: [String]) -> Bool {
|
||||
self.matches(text: text, triggers: triggers)
|
||||
|
||||
@@ -23,6 +23,7 @@ final class VoiceWakeTester {
|
||||
private var holdingAfterDetect = false
|
||||
private var detectedText: String?
|
||||
private let logger = Logger(subsystem: "com.steipete.clawdis", category: "voicewake")
|
||||
private let silenceWindow: TimeInterval = 1.0
|
||||
|
||||
init(locale: Locale = .current) {
|
||||
self.recognizer = SFSpeechRecognizer(locale: locale)
|
||||
@@ -132,10 +133,11 @@ final class VoiceWakeTester {
|
||||
self.holdingAfterDetect = true
|
||||
self.detectedText = text
|
||||
self.logger.info("voice wake detected; forwarding (len=\(text.count))")
|
||||
await MainActor.run { AppStateStore.shared.triggerVoiceEars() }
|
||||
await MainActor.run { AppStateStore.shared.triggerVoiceEars(ttl: nil) }
|
||||
let config = await MainActor.run { AppStateStore.shared.voiceWakeForwardConfig }
|
||||
Task.detached {
|
||||
await VoiceWakeForwarder.forward(transcript: text, config: config)
|
||||
let payload = VoiceWakeForwarder.prefixedTranscript(text)
|
||||
await VoiceWakeForwarder.forward(transcript: payload, config: config)
|
||||
}
|
||||
Task { @MainActor in onUpdate(.detected(text)) }
|
||||
self.holdUntilSilence(onUpdate: onUpdate)
|
||||
@@ -162,8 +164,7 @@ final class VoiceWakeTester {
|
||||
Task { [weak self] in
|
||||
guard let self else { return }
|
||||
let detectedAt = Date()
|
||||
let hardStop = detectedAt.addingTimeInterval(3) // cap overall listen after trigger
|
||||
let silenceWindow: TimeInterval = 0.8
|
||||
let hardStop = detectedAt.addingTimeInterval(6) // cap overall listen after trigger
|
||||
|
||||
while !self.isStopping {
|
||||
let now = Date()
|
||||
@@ -175,6 +176,7 @@ final class VoiceWakeTester {
|
||||
}
|
||||
if !self.isStopping {
|
||||
self.stop()
|
||||
await MainActor.run { AppStateStore.shared.stopVoiceEars() }
|
||||
if let detectedText {
|
||||
self.logger.info("voice wake hold finished; len=\(detectedText.count)")
|
||||
Task { @MainActor in onUpdate(.detected(detectedText)) }
|
||||
|
||||
Reference in New Issue
Block a user