283 lines
10 KiB
Swift
283 lines
10 KiB
Swift
import SwiftUI
|
|
import UIKit
|
|
|
|
/// 「长按 + 语音直达」面板:开口说想记什么 → 端侧转写(SpeechDictationService)
|
|
/// → LLM 意图分类(VoiceIntentService)→ 回调 RootView 打开对应新建入口。
|
|
///
|
|
/// 状态机:
|
|
/// ```
|
|
/// requesting(权限) → recording(实时字幕) → classifying → onResolve(intent) 关闭
|
|
/// │ 拒绝 → denied │ 没听到/没听懂 → failed(再说一次 / 打开菜单)
|
|
/// ```
|
|
/// 全程本机:转写 requiresOnDeviceRecognition,分类走端侧 LLM。
|
|
struct VoiceCommandSheet: View {
|
|
/// 识别成功:RootView 负责关闭本 sheet 并路由。
|
|
let onResolve: (VoiceIntent) -> Void
|
|
/// 兜底:打开普通新建菜单(RecordSheet)。
|
|
let onOpenMenu: () -> Void
|
|
@Environment(\.dismiss) private var dismiss
|
|
|
|
enum Phase: Equatable {
|
|
case requesting
|
|
case denied
|
|
case recording
|
|
case classifying
|
|
case failed(message: String)
|
|
}
|
|
|
|
@State private var phase: Phase = .requesting
|
|
@State private var transcript = ""
|
|
@State private var seconds = 0
|
|
/// @State 保证视图身份期内实例唯一(同 DiaryQuickSheet 的注释,防止重建后麦克风悬挂)。
|
|
@State private var dictation = SpeechDictationService()
|
|
@State private var ticker: Task<Void, Never>?
|
|
|
|
/// 录音超过 20s 自动结束:语音直达说的都是短句,长录是忘了点完成。
|
|
private let maxSeconds = 20
|
|
|
|
var body: some View {
|
|
VStack(spacing: 0) {
|
|
Capsule()
|
|
.fill(Tj.Palette.line)
|
|
.frame(width: 40, height: 4)
|
|
.padding(.top, 10)
|
|
.padding(.bottom, 16)
|
|
|
|
HStack {
|
|
VStack(alignment: .leading, spacing: 2) {
|
|
Text("说出想记的内容")
|
|
.font(.tjH2())
|
|
.foregroundStyle(Tj.Palette.text)
|
|
Text("比如:记一下血压 / 我头疼 / 拍个药盒")
|
|
.font(.tjScaled( 11))
|
|
.foregroundStyle(Tj.Palette.text3)
|
|
}
|
|
Spacer()
|
|
Text("全程本机")
|
|
.font(.tjScaled( 12))
|
|
.foregroundStyle(Tj.Palette.text3)
|
|
}
|
|
.padding(.horizontal, 20)
|
|
.padding(.bottom, 16)
|
|
|
|
content
|
|
.frame(maxWidth: .infinity, maxHeight: .infinity, alignment: .top)
|
|
.padding(.horizontal, 20)
|
|
|
|
buttons
|
|
.padding(.horizontal, 20)
|
|
.padding(.vertical, 14)
|
|
}
|
|
.background(
|
|
Tj.Palette.sand
|
|
.clipShape(RoundedRectangle(cornerRadius: Tj.Radius.xl, style: .continuous))
|
|
.ignoresSafeArea(edges: .bottom)
|
|
)
|
|
.presentationDetents([.fraction(0.5)])
|
|
.presentationDragIndicator(.hidden)
|
|
.presentationBackground(Tj.Palette.sand)
|
|
.presentationCornerRadius(Tj.Radius.xl)
|
|
.task { await begin() }
|
|
.onDisappear {
|
|
ticker?.cancel()
|
|
dictation.abort()
|
|
}
|
|
}
|
|
|
|
// MARK: - 分阶段内容
|
|
|
|
@ViewBuilder
|
|
private var content: some View {
|
|
switch phase {
|
|
case .requesting:
|
|
ProgressView().tint(Tj.Palette.ink)
|
|
.frame(maxWidth: .infinity)
|
|
.padding(.top, 30)
|
|
|
|
case .denied:
|
|
VStack(spacing: 10) {
|
|
Image(systemName: "mic.slash")
|
|
.font(.tjScaled( 30))
|
|
.foregroundStyle(Tj.Palette.text3)
|
|
Text("需要麦克风与语音识别权限")
|
|
.font(.tjScaled( 14, weight: .semibold))
|
|
.foregroundStyle(Tj.Palette.text)
|
|
Text("语音和文字都只在本机处理,不会上传。")
|
|
.font(.tjScaled( 12))
|
|
.foregroundStyle(Tj.Palette.text3)
|
|
Button("前往设置") {
|
|
if let url = URL(string: UIApplication.openSettingsURLString) {
|
|
UIApplication.shared.open(url)
|
|
}
|
|
}
|
|
.font(.tjScaled( 13, weight: .semibold))
|
|
.foregroundStyle(Tj.Palette.ink)
|
|
}
|
|
.frame(maxWidth: .infinity)
|
|
.padding(.top, 16)
|
|
|
|
case .recording:
|
|
VStack(spacing: 14) {
|
|
HStack(spacing: 8) {
|
|
Circle()
|
|
.fill(Tj.Palette.brick)
|
|
.frame(width: 8, height: 8)
|
|
Text("正在听 · \(seconds)s")
|
|
.font(.tjScaled( 12, weight: .semibold))
|
|
.foregroundStyle(Tj.Palette.brick)
|
|
}
|
|
transcriptBox(placeholder: String(appLoc: "请开口说话…"))
|
|
}
|
|
|
|
case .classifying:
|
|
VStack(spacing: 14) {
|
|
HStack(spacing: 8) {
|
|
ProgressView().tint(Tj.Palette.ink)
|
|
Text("正在理解…")
|
|
.font(.tjScaled( 12, weight: .semibold))
|
|
.foregroundStyle(Tj.Palette.text2)
|
|
}
|
|
transcriptBox(placeholder: "")
|
|
}
|
|
|
|
case .failed(let message):
|
|
VStack(spacing: 10) {
|
|
Image(systemName: "questionmark.bubble")
|
|
.font(.tjScaled( 28))
|
|
.foregroundStyle(Tj.Palette.text3)
|
|
Text(message)
|
|
.font(.tjScaled( 13))
|
|
.foregroundStyle(Tj.Palette.text2)
|
|
.multilineTextAlignment(.center)
|
|
if !transcript.isEmpty {
|
|
Text("“\(transcript)”")
|
|
.font(.tjScaled( 12))
|
|
.foregroundStyle(Tj.Palette.text3)
|
|
.lineLimit(2)
|
|
}
|
|
}
|
|
.frame(maxWidth: .infinity)
|
|
.padding(.top, 12)
|
|
}
|
|
}
|
|
|
|
private func transcriptBox(placeholder: String) -> some View {
|
|
ScrollView(showsIndicators: false) {
|
|
Text(transcript.isEmpty ? placeholder : transcript)
|
|
.font(.tjScaled( 15))
|
|
.foregroundStyle(transcript.isEmpty ? Tj.Palette.text3 : Tj.Palette.text)
|
|
.frame(maxWidth: .infinity, alignment: .leading)
|
|
}
|
|
.frame(minHeight: 64, maxHeight: 110)
|
|
.padding(.horizontal, 14)
|
|
.padding(.vertical, 12)
|
|
.background(
|
|
RoundedRectangle(cornerRadius: Tj.Radius.sm, style: .continuous)
|
|
.fill(Tj.Palette.paper)
|
|
)
|
|
.overlay(
|
|
RoundedRectangle(cornerRadius: Tj.Radius.sm, style: .continuous)
|
|
.strokeBorder(Tj.Palette.line, lineWidth: 1)
|
|
)
|
|
}
|
|
|
|
// MARK: - 底部按钮
|
|
|
|
@ViewBuilder
|
|
private var buttons: some View {
|
|
switch phase {
|
|
case .recording:
|
|
HStack(spacing: 12) {
|
|
Button("取消") { dismiss() }
|
|
.buttonStyle(TjGhostButton(height: 44, fontSize: 15, horizontalPadding: 18))
|
|
Button("说完了") { finishRecording() }
|
|
.buttonStyle(TjPrimaryButton(height: 44, fontSize: 15, horizontalPadding: 18))
|
|
}
|
|
case .failed:
|
|
HStack(spacing: 12) {
|
|
Button("打开新建菜单") { onOpenMenu() }
|
|
.buttonStyle(TjGhostButton(height: 44, fontSize: 14, horizontalPadding: 14))
|
|
Button("再说一次") { Task { await begin() } }
|
|
.buttonStyle(TjPrimaryButton(height: 44, fontSize: 14, horizontalPadding: 18))
|
|
}
|
|
case .denied:
|
|
Button("取消") { dismiss() }
|
|
.buttonStyle(TjGhostButton(height: 44, fontSize: 15, horizontalPadding: 18))
|
|
case .requesting, .classifying:
|
|
Button("取消") { dismiss() }
|
|
.buttonStyle(TjGhostButton(height: 44, fontSize: 15, horizontalPadding: 18))
|
|
}
|
|
}
|
|
|
|
// MARK: - 流程
|
|
|
|
private func begin() async {
|
|
ticker?.cancel()
|
|
transcript = ""
|
|
seconds = 0
|
|
guard SpeechDictationService.isAvailable else {
|
|
phase = .failed(message: String(appLoc: "本机不支持端侧语音识别,试试下面的新建菜单"))
|
|
return
|
|
}
|
|
phase = .requesting
|
|
guard await dictation.requestAuthorization() else {
|
|
phase = .denied
|
|
return
|
|
}
|
|
do {
|
|
try dictation.start { transcript = $0 }
|
|
phase = .recording
|
|
startTicker()
|
|
} catch {
|
|
phase = .failed(message: error.localizedDescription)
|
|
}
|
|
}
|
|
|
|
private func startTicker() {
|
|
ticker = Task { @MainActor in
|
|
while !Task.isCancelled {
|
|
try? await Task.sleep(nanoseconds: 1_000_000_000)
|
|
guard phase == .recording else { return }
|
|
seconds += 1
|
|
if seconds >= maxSeconds {
|
|
finishRecording()
|
|
return
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
private func finishRecording() {
|
|
guard phase == .recording else { return }
|
|
ticker?.cancel()
|
|
// 已经在屏幕上的实时字幕:stop() 偶发因最终结果竞争/取消返回空,
|
|
// 这时别把用户已经看到的内容丢掉,回退到实时字幕。
|
|
let live = transcript
|
|
phase = .classifying
|
|
Task {
|
|
let finalText = await dictation.stop()
|
|
let text = finalText.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty
|
|
? live
|
|
: finalText
|
|
transcript = text
|
|
let trimmed = text.trimmingCharacters(in: .whitespacesAndNewlines)
|
|
guard !trimmed.isEmpty else {
|
|
phase = .failed(message: String(appLoc: "没听到内容,再试一次?"))
|
|
return
|
|
}
|
|
if let intent = await VoiceIntentService.classify(trimmed) {
|
|
onResolve(intent)
|
|
} else {
|
|
phase = .failed(message: String(appLoc: "没听懂想记什么,再说一次,或直接选菜单"))
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
#Preview {
|
|
Text("bg")
|
|
.sheet(isPresented: .constant(true)) {
|
|
VoiceCommandSheet(onResolve: { print($0) }, onOpenMenu: {})
|
|
}
|
|
}
|