fix(voice): unify overlay send flow

main
Peter Steinberger 2025-12-10 02:52:42 +01:00
parent cf2b659491
commit 657450c40c
4 changed files with 32 additions and 40 deletions

View File

@ -288,7 +288,9 @@ actor VoicePushToTalk {
VoiceWakeChimePlayer.play(chime, reason: "ptt.fallback_send") VoiceWakeChimePlayer.play(chime, reason: "ptt.fallback_send")
} }
Task.detached { Task.detached {
await VoiceWakeForwarder.forward(transcript: finalText, config: forward) await VoiceWakeForwarder.forward(
transcript: VoiceWakeForwarder.prefixedTranscript(finalText),
config: forward)
} }
} }
} }

View File

@ -21,7 +21,6 @@ final class VoiceSessionCoordinator: ObservableObject {
private let logger = Logger(subsystem: "com.steipete.clawdis", category: "voicewake.coordinator") private let logger = Logger(subsystem: "com.steipete.clawdis", category: "voicewake.coordinator")
private var session: Session? private var session: Session?
private var autoSendTask: Task<Void, Never>?
// MARK: - API // MARK: - API
@ -40,7 +39,7 @@ final class VoiceSessionCoordinator: ObservableObject {
let token = UUID() let token = UUID()
self.logger.info("coordinator start token=\(token.uuidString) source=\(source.rawValue) len=\(text.count)") self.logger.info("coordinator start token=\(token.uuidString) source=\(source.rawValue) len=\(text.count)")
let attributedText = attributed ?? VoiceWakeOverlayController.shared.makeAttributed(from: text) let attributedText = attributed ?? VoiceWakeOverlayController.shared.makeAttributed(from: text)
self.session = Session( let session = Session(
token: token, token: token,
source: source, source: source,
text: text, text: text,
@ -49,7 +48,9 @@ final class VoiceSessionCoordinator: ObservableObject {
forwardConfig: forwardEnabled ? AppStateStore.shared.voiceWakeForwardConfig : nil, forwardConfig: forwardEnabled ? AppStateStore.shared.voiceWakeForwardConfig : nil,
sendChime: .none, sendChime: .none,
autoSendDelay: nil) autoSendDelay: nil)
self.session = session
VoiceWakeOverlayController.shared.startSession( VoiceWakeOverlayController.shared.startSession(
token: token,
source: VoiceWakeOverlayController.Source(rawValue: source.rawValue) ?? .wakeWord, source: VoiceWakeOverlayController.Source(rawValue: source.rawValue) ?? .wakeWord,
transcript: text, transcript: text,
attributed: attributedText, attributed: attributedText,
@ -76,7 +77,6 @@ final class VoiceSessionCoordinator: ObservableObject {
self.logger self.logger
.info( .info(
"coordinator finalize token=\(token.uuidString) len=\(text.count) autoSendAfter=\(autoSendAfter ?? -1)") "coordinator finalize token=\(token.uuidString) len=\(text.count) autoSendAfter=\(autoSendAfter ?? -1)")
self.autoSendTask?.cancel(); self.autoSendTask = nil
self.session?.text = text self.session?.text = text
self.session?.isFinal = true self.session?.isFinal = true
self.session?.forwardConfig = forwardConfig self.session?.forwardConfig = forwardConfig
@ -108,7 +108,7 @@ final class VoiceSessionCoordinator: ObservableObject {
self.clearSession() self.clearSession()
return return
} }
VoiceWakeOverlayController.shared.sendNow(token: token, sendChime: session.sendChime) VoiceWakeOverlayController.shared.beginSendUI(token: token, sendChime: session.sendChime)
Task.detached { Task.detached {
_ = await VoiceWakeForwarder.forward( _ = await VoiceWakeForwarder.forward(
transcript: VoiceWakeForwarder.prefixedTranscript(text), transcript: VoiceWakeForwarder.prefixedTranscript(text),
@ -139,7 +139,5 @@ final class VoiceSessionCoordinator: ObservableObject {
private func clearSession() { private func clearSession() {
self.session = nil self.session = nil
self.autoSendTask?.cancel()
self.autoSendTask = nil
} }
} }

View File

@ -46,6 +46,7 @@ final class VoiceWakeOverlayController: ObservableObject {
@discardableResult @discardableResult
func startSession( func startSession(
token: UUID = UUID(),
source: Source, source: Source,
transcript: String, transcript: String,
attributed: NSAttributedString? = nil, attributed: NSAttributedString? = nil,
@ -56,7 +57,6 @@ final class VoiceWakeOverlayController: ObservableObject {
self.logger.log(level: .info, "overlay drop session_start while sending") self.logger.log(level: .info, "overlay drop session_start while sending")
return self.activeToken ?? UUID() return self.activeToken ?? UUID()
} }
let token = UUID()
let message = """ let message = """
overlay session_start source=\(source.rawValue) \ overlay session_start source=\(source.rawValue) \
len=\(transcript.count) len=\(transcript.count)
@ -133,9 +133,9 @@ final class VoiceWakeOverlayController: ObservableObject {
if let delay { if let delay {
if delay <= 0 { if delay <= 0 {
self.logger.log(level: .info, "overlay autoSend immediate token=\(token.uuidString)") self.logger.log(level: .info, "overlay autoSend immediate token=\(token.uuidString)")
self.sendNow(token: token, sendChime: sendChime) VoiceSessionCoordinator.shared.sendNow(token: token, reason: "autoSendImmediate")
} else { } else {
self.scheduleAutoSend(token: token, after: delay, sendChime: sendChime) self.scheduleAutoSend(token: token, after: delay)
} }
} }
} }
@ -164,50 +164,41 @@ final class VoiceWakeOverlayController: ObservableObject {
self.updateWindowFrame(animate: true) self.updateWindowFrame(animate: true)
} }
func sendNow(token: UUID? = nil, sendChime: VoiceWakeChime = .none) { /// UI-only path: show sending state and dismiss; actual forwarding is handled by the coordinator.
guard self.guardToken(token, context: "send") else { return } func beginSendUI(token: UUID, sendChime: VoiceWakeChime = .none) {
guard self.guardToken(token, context: "beginSendUI") else { return }
self.autoSendTask?.cancel(); self.autoSendToken = nil
let message = """ let message = """
overlay sendNow called token=\(self.activeToken?.uuidString ?? "nil") \ overlay beginSendUI token=\(token.uuidString) \
isSending=\(self.model.isSending) \ isSending=\(self.model.isSending) \
forwardEnabled=\(self.model.forwardEnabled) \ forwardEnabled=\(self.model.forwardEnabled) \
textLen=\(self.model.text.count) textLen=\(self.model.text.count)
""" """
self.logger.log(level: .info, "\(message)") self.logger.log(level: .info, "\(message)")
self.autoSendTask?.cancel(); self.autoSendToken = nil
if self.model.isSending { return } if self.model.isSending { return }
self.model.isEditing = false self.model.isEditing = false
guard let forwardConfig, forwardConfig.enabled else {
self.logger.log(level: .info, "overlay sendNow disabled -> dismiss")
self.dismiss(reason: .explicit)
return
}
let text = self.model.text.trimmingCharacters(in: .whitespacesAndNewlines)
guard !text.isEmpty else {
self.logger.log(level: .info, "overlay sendNow empty -> dismiss")
self.dismiss(reason: .empty)
return
}
if sendChime != .none { if sendChime != .none {
let message = "overlay sendNow playing sendChime=\(String(describing: sendChime))" let message = "overlay beginSendUI playing sendChime=\(String(describing: sendChime))"
self.logger.log(level: .info, "\(message)") self.logger.log(level: .info, "\(message)")
VoiceWakeChimePlayer.play(sendChime, reason: "overlay.send") VoiceWakeChimePlayer.play(sendChime, reason: "overlay.send")
} }
self.model.isSending = true self.model.isSending = true
let payload = VoiceWakeForwarder.prefixedTranscript(text)
self.logger.log(level: .info, "overlay sendNow forwarding len=\(payload.count, privacy: .public)")
Task.detached {
await VoiceWakeForwarder.forward(transcript: payload, config: forwardConfig)
}
DispatchQueue.main.asyncAfter(deadline: .now() + 0.28) { DispatchQueue.main.asyncAfter(deadline: .now() + 0.28) {
self.logger.log( self.logger.log(
level: .info, level: .info,
"overlay sendNow dismiss ticking token=\(self.activeToken?.uuidString ?? "nil")") "overlay beginSendUI dismiss ticking token=\(self.activeToken?.uuidString ?? "nil")")
self.dismiss(token: token, reason: .explicit, outcome: .sent) self.dismiss(token: token, reason: .explicit, outcome: .sent)
} }
} }
func requestSend(token: UUID? = nil, reason: String = "overlay_request") {
guard self.guardToken(token, context: "requestSend") else { return }
guard let active = token ?? self.activeToken else { return }
VoiceSessionCoordinator.shared.sendNow(token: active, reason: reason)
}
func dismiss(token: UUID? = nil, reason: DismissReason = .explicit, outcome: SendOutcome = .empty) { func dismiss(token: UUID? = nil, reason: DismissReason = .explicit, outcome: SendOutcome = .empty) {
guard self.guardToken(token, context: "dismiss") else { return } guard self.guardToken(token, context: "dismiss") else { return }
let message = """ let message = """
@ -408,17 +399,16 @@ final class VoiceWakeOverlayController: ObservableObject {
} }
} }
private func scheduleAutoSend(token: UUID, after delay: TimeInterval, sendChime: VoiceWakeChime) { private func scheduleAutoSend(token: UUID, after delay: TimeInterval) {
self.logger.log( self.logger.log(
level: .info, level: .info,
""" """
overlay scheduleAutoSend token=\(token.uuidString) \ overlay scheduleAutoSend token=\(token.uuidString) \
after=\(delay) \ after=\(delay)
sendChime=\(String(describing: sendChime))
""") """)
self.autoSendTask?.cancel() self.autoSendTask?.cancel()
self.autoSendToken = token self.autoSendToken = token
self.autoSendTask = Task<Void, Never> { [weak self, sendChime, token] in self.autoSendTask = Task<Void, Never> { [weak self, token] in
let nanos = UInt64(max(0, delay) * 1_000_000_000) let nanos = UInt64(max(0, delay) * 1_000_000_000)
try? await Task.sleep(nanoseconds: nanos) try? await Task.sleep(nanoseconds: nanos)
guard !Task.isCancelled else { return } guard !Task.isCancelled else { return }
@ -428,7 +418,7 @@ final class VoiceWakeOverlayController: ObservableObject {
self.logger.log( self.logger.log(
level: .info, level: .info,
"overlay autoSend firing token=\(token.uuidString, privacy: .public)") "overlay autoSend firing token=\(token.uuidString, privacy: .public)")
self.sendNow(token: token, sendChime: sendChime) VoiceSessionCoordinator.shared.sendNow(token: token, reason: "autoSendDelay")
self.autoSendTask = nil self.autoSendTask = nil
} }
} }
@ -471,7 +461,7 @@ private struct VoiceWakeOverlayView: View {
self.controller.endEditing() self.controller.endEditing()
}, },
onSend: { onSend: {
self.controller.sendNow() self.controller.requestSend()
}) })
.focused(self.$textFocused) .focused(self.$textFocused)
.frame(maxWidth: .infinity, minHeight: 32, maxHeight: .infinity, alignment: .topLeading) .frame(maxWidth: .infinity, minHeight: 32, maxHeight: .infinity, alignment: .topLeading)
@ -488,7 +478,7 @@ private struct VoiceWakeOverlayView: View {
} }
Button { Button {
self.controller.sendNow() self.controller.requestSend()
} label: { } label: {
let sending = self.controller.model.isSending let sending = self.controller.model.isSending
let level = self.controller.model.level let level = self.controller.model.level

View File

@ -351,7 +351,9 @@ actor VoiceWakeRuntime {
await MainActor.run { VoiceWakeChimePlayer.play(sendChime, reason: "voicewake.send") } await MainActor.run { VoiceWakeChimePlayer.play(sendChime, reason: "voicewake.send") }
} }
Task.detached { Task.detached {
await VoiceWakeForwarder.forward(transcript: finalTranscript, config: forwardConfig) await VoiceWakeForwarder.forward(
transcript: VoiceWakeForwarder.prefixedTranscript(finalTranscript),
config: forwardConfig)
} }
} }
self.overlayToken = nil self.overlayToken = nil