feat: flow session reliability, keyboard UX, and audio pre-roll fix

Renew voice sessions while the host app stays foreground, auto-start flow from the keyboard with a Start action, and prevent leading audio loss via pre-roll buffering and faster signal polling.
2026-06-20 00:57:16 +08:00
parent 38a36ed504
commit db5151ad9a
12 changed files with 196 additions and 77 deletions
@@ -66,6 +66,8 @@ public final class KeyboardViewController: UIInputViewController {
    private var flowSessionMonitorTask: Task<Void, Never>?
    private var flowSessionDarwinObserver: FlowSessionDarwinObserver?
    private var isAwaitingFlowResult = false
+    private var lastFlowAutoStartAttempt: TimeInterval = 0
+    private static let flowAutoStartCooldown: TimeInterval = 20

    // MARK: - Lifecycle

@@ -121,6 +123,7 @@ public final class KeyboardViewController: UIInputViewController {
        state.endRecording        = { [weak self] in self?.pressEnded() }
        state.tapMic              = { [weak self] in self?.toggleRecording() }
        state.openSettings        = { [weak self] in self?.openHostApp() }
+        state.startFlowSession    = { [weak self] in self?.beginFlowStart() }
        state.setMode             = { [weak self] m in self?.persistMode(m) }
        state.setLocale           = { [weak self] l in self?.persistLocale(l) }
        state.setEngineMode       = { [weak self] m in self?.persistEngineMode(m) }
@@ -199,6 +202,24 @@ public final class KeyboardViewController: UIInputViewController {
            }
        }
        wasFlowSessionActive = active
+
+        if !active {
+            maybeAutoStartFlowSession()
+        }
+    }
+
+    /// When the host session is down, proactively jump to the app to start it.
+    private func maybeAutoStartFlowSession() {
+        guard !FlowSessionBridge.isSessionActive() else { return }
+        guard !isPendingFlowStart, !isFlowRecording, !isAwaitingFlowResult else { return }
+        guard state.mode != .off else { return }
+        guard hasFullAccess, AppGroup.isAvailable else { return }
+        guard case .idle = state.phase else { return }
+
+        let now = Date().timeIntervalSince1970
+        guard now - lastFlowAutoStartAttempt >= Self.flowAutoStartCooldown else { return }
+        lastFlowAutoStartAttempt = now
+        beginFlowStart()
    }

    private func showFlowSessionExpiredHint() {
@@ -303,6 +324,7 @@ public final class KeyboardViewController: UIInputViewController {
    }

    private func beginFlowStart() {
+        guard !isPendingFlowStart else { return }
        isPendingFlowStart = true
        isFlowRecording = false
        flowStartDeadline = Date().timeIntervalSince1970 + FlowWatchdog.startTimeout
@@ -19,7 +19,13 @@ import SwiftUI
 import OSGKeyboardShared

 private enum KeyboardLayoutMetrics {
-    static let sideActionButtonSize: CGFloat = 44
+    static let sideActionButtonSize: CGFloat = 53
+    static let sideActionIconSize: CGFloat = 19
+    static let sideSpaceBarWidth: CGFloat = 19
+    static let micFlankMinSpacing: CGFloat = 36
+    static let sideActionStackSpacing: CGFloat = 16
+    /// Outer inset for delete / return·space from screen edges (8 pt → 24 pt, +200%).
+    static let sideActionHorizontalInset: CGFloat = Spacing.xs * 3
 }

 public struct KeyboardRootView: View {
@@ -36,7 +42,6 @@ public struct KeyboardRootView: View {
    /// it up.
    static let totalHeight: CGFloat = 280
    private static let topBarHeight: CGFloat = 38
-    private static let sideActionStackSpacing: CGFloat = 10

    private var palette: ThemePalette {
        colorScheme == .dark ? Palette.dark : Palette.light
@@ -52,14 +57,16 @@ public struct KeyboardRootView: View {
        }
        .padding(.top, 4)
        .padding(.bottom, 6)
-        // Let the system UI chrome show through by drawing no background
-        // of our own.
-        .background(Color.clear)
+        .background(keyboardBackground)
        .frame(height: Self.totalHeight)
        // Feed the resolved palette to all nested chips/buttons.
        .environment(\.themePalette, palette)
    }

+    private var keyboardBackground: Color {
+        colorScheme == .dark ? palette.background : Color.clear
+    }
+
    // MARK: - Top bar

    private var topBar: some View {
@@ -99,7 +106,8 @@ public struct KeyboardRootView: View {
                phase: state.phase,
                transcript: state.lastTranscript,
                flowSessionActive: state.flowSessionActive,
-                openSettings: state.openSettings
+                openSettings: state.openSettings,
+                startFlowSession: state.startFlowSession
            )
            .frame(height: 22)

@@ -117,11 +125,13 @@ public struct KeyboardRootView: View {
    /// HStack vertical alignment keeps delete, mic centre, and the gap
    /// between return/space on one horizontal axis.
    private var micActionRow: some View {
-        HStack(alignment: .center, spacing: 20) {
+        HStack(alignment: .center, spacing: 0) {
            CircularToolbarButton(systemName: "delete.left", label: "delete") {
                state.deleteBackward()
            }

+            Spacer(minLength: KeyboardLayoutMetrics.micFlankMinSpacing)
+
            RecordButton(
                phase: buttonPhase,
                level: state.level,
@@ -130,7 +140,9 @@ public struct KeyboardRootView: View {
            )
            .frame(width: 132, height: 132)

-            VStack(spacing: Self.sideActionStackSpacing) {
+            Spacer(minLength: KeyboardLayoutMetrics.micFlankMinSpacing)
+
+            VStack(spacing: KeyboardLayoutMetrics.sideActionStackSpacing) {
                CircularToolbarButton(systemName: "return", label: "newline") {
                    state.insertNewline()
                }
@@ -139,6 +151,7 @@ public struct KeyboardRootView: View {
                }
            }
        }
+        .padding(.horizontal, KeyboardLayoutMetrics.sideActionHorizontalInset)
        .frame(maxWidth: .infinity)
    }

@@ -191,6 +204,7 @@ private struct TranscriptLine: View {
    let transcript: String
    let flowSessionActive: Bool
    let openSettings: () -> Void
+    let startFlowSession: () -> Void

    var body: some View {
        ZStack {
@@ -201,9 +215,18 @@ private struct TranscriptLine: View {
                        .font(TypeStyle.caption)
                        .foregroundStyle(palette.textTertiary)
                } else {
-                    ExtL10n.text("keyboard.flow.sessionInactive")
-                        .font(TypeStyle.caption)
-                        .foregroundStyle(palette.textTertiary)
+                    HStack(spacing: 6) {
+                        ExtL10n.text("keyboard.flow.sessionInactive")
+                            .font(TypeStyle.caption)
+                            .foregroundStyle(palette.textTertiary)
+                        Button(action: startFlowSession) {
+                            ExtL10n.text("keyboard.flow.start")
+                                .font(TypeStyle.caption)
+                                .foregroundStyle(palette.accent)
+                        }
+                        .buttonStyle(.plain)
+                        .accessibilityHint(ExtL10n.text("keyboard.flow.startA11y"))
+                    }
                }
            case .requestingPermissions:
                HStack(spacing: 6) {
@@ -267,6 +290,7 @@ private struct TranscriptLine: View {
 // MARK: - Circular toolbar button

 private struct CircularToolbarButton: View {
+    @Environment(\.colorScheme) private var colorScheme
    @Environment(\.themePalette) private var palette: ThemePalette

    let systemName: String?
@@ -294,20 +318,26 @@ private struct CircularToolbarButton: View {
                if spaceStyle {
                    Capsule()
                        .fill(palette.textPrimary)
-                        .frame(width: 16, height: 3)
+                        .frame(width: KeyboardLayoutMetrics.sideSpaceBarWidth, height: 3)
                } else if let systemName {
                    Image(systemName: systemName)
-                        .font(.system(size: 16, weight: .medium))
+                        .font(.system(size: KeyboardLayoutMetrics.sideActionIconSize, weight: .medium))
                        .foregroundStyle(palette.textPrimary)
                }
            }
            .frame(width: KeyboardLayoutMetrics.sideActionButtonSize, height: KeyboardLayoutMetrics.sideActionButtonSize)
-            .background(palette.surfaceElevated, in: Circle())
-            .overlay(Circle().stroke(palette.divider, lineWidth: 0.5))
+            .background(sideButtonFill, in: Circle())
+            .overlay(Circle().stroke(palette.dividerStrong, lineWidth: 0.5))
        }
        .buttonStyle(.plain)
        .accessibilityLabel(Text(label))
    }
+
+    private var sideButtonFill: Color {
+        colorScheme == .dark
+            ? Color(red: 0.20, green: 0.20, blue: 0.22)
+            : palette.surfaceElevated
+    }
 }

 // MARK: - Status badge
@@ -27,12 +27,12 @@
 "common.newline" = "Return";

 /* Privacy footnote (onboarding welcome page) */
-"privacy.audio.title" = "Audio stays on device";
-"privacy.audio.body" = "Transcribed locally with Apple’s speech engine.";
-"privacy.network.title" = "Only the polished text is sent";
-"privacy.network.body" = "Sent to your chosen LLM to add structure & punctuation.";
+"privacy.audio.title" = "On-device transcription";
+"privacy.audio.body" = "Powered by the on-device iOS speech engine.";
+"privacy.network.title" = "Automatic text polish";
+"privacy.network.body" = "AI formats and polishes your text automatically.";
 "privacy.universal.title" = "Works everywhere";
-"privacy.universal.body" = "WeChat, Notes, Mail, ChatGPT, Claude, Cursor — anywhere a keyboard appears.";
+"privacy.universal.body" = "Chat, email, vibe coding — just speak.";

 /* Home */
 "home.status.ready" = "Ready";
@@ -127,7 +127,9 @@
 "keyboard.tapToTalkA11y" = "Tap to talk";

 /* Flow session (keyboard) */
-"keyboard.flow.sessionInactive" = "Open OSGKeyboard to start voice session";
+"keyboard.flow.sessionInactive" = "Voice session off";
+"keyboard.flow.start" = "Start";
+"keyboard.flow.startA11y" = "Start voice session";
 "keyboard.flow.sessionExpired" = "Voice session ended. Open OSGKeyboard to restart.";
 "keyboard.flow.startingSession" = "Starting voice session…";
 "keyboard.flow.transcribing" = "Transcribing…";
@@ -27,12 +27,12 @@
 "common.newline" = "换行";

 /* Privacy footnote (onboarding welcome page) */
-"privacy.audio.title" = "音频不出本机";
-"privacy.audio.body" = "由 Apple 端侧引擎转录。";
-"privacy.network.title" = "仅发送润色后文字";
-"privacy.network.body" = "仅向所选 LLM 发送润色后文字,用于整理结构和标点。";
+"privacy.audio.title" = "支持本地转写";
+"privacy.audio.body" = "支持 iOS 本地引擎转录";
+"privacy.network.title" = "文字自动润色";
+"privacy.network.body" = "通过 AI 自动对文字进行排版润色";
 "privacy.universal.title" = "处处可用";
-"privacy.universal.body" = "微信、备忘录、邮件、ChatGPT、Claude、Cursor — 任何键盘出现的地方。";
+"privacy.universal.body" = "聊天、邮件、Vibe Coding 动嘴就行";

 /* Home */
 "home.status.ready" = "就绪";
@@ -127,7 +127,9 @@
 "keyboard.tapToTalkA11y" = "点按说话";

 /* Flow session (keyboard) */
-"keyboard.flow.sessionInactive" = "请打开 OSGKeyboard 启动语音会话";
+"keyboard.flow.sessionInactive" = "语音会话未启动";
+"keyboard.flow.start" = "启动";
+"keyboard.flow.startA11y" = "启动语音会话";
 "keyboard.flow.sessionExpired" = "语音会话已结束，请打开 OSGKeyboard 重新启动";
 "keyboard.flow.startingSession" = "正在启动语音会话…";
 "keyboard.flow.transcribing" = "识别中…";