Fix shared input session lifecycle and expand industry lexicons

Package.swift

@@ -48,6 +48,7 @@ let package = Package(
                 .copy("Resources/SettingsIntegrationsIllustration.png"),
                 .copy("Resources/SettingsAboutIllustration.png"),
                 .copy("Resources/IndustryLexicons.json"),
+                .copy("Resources/THUOCL-LICENSE.txt"),
                 .copy("Resources/Sounds"),
                 .copy("Resources/AppIcon.icns"),
                 .copy("Resources/AppIconLight.icns"),

Sources/App/AppDelegate+Integrations.swift

@@ -117,10 +117,9 @@ extension AppDelegate {
     func makeIntegrationSessionCoordinator(service: OpenTypeService) -> InputSessionCoordinator {
         InputSessionCoordinator(
             service: service,
+            engineProvider: speechEngineProvider,
             textProcessor: textProcessor,
-            isUserWorkflowBusy: { [weak self] in
-                self?.appState.isBusy ?? false
-            }
+            ownership: inputOwnership
         )
     }
 }

Sources/App/InputSessionOwnership.swift

@@ -0,0 +1,35 @@
+import Foundation
+
+/// All input entries share this reservation, including preparation and cancellation drain.
+@MainActor
+final class InputSessionOwnership {
+    private var current: UUID?
+    private var cancelled = false
+
+    var isBusy: Bool { current != nil }
+
+    func acquire() throws -> UUID {
+        guard current == nil else { throw IntegrationError.busy }
+        let id = UUID()
+        current = id
+        cancelled = false
+        return id
+    }
+
+    func check(_ id: UUID) throws {
+        try Task.checkCancellation()
+        guard current == id, !cancelled else { throw CancellationError() }
+    }
+
+    func cancel(_ id: UUID) {
+        guard current == id else { return }
+        cancelled = true
+    }
+
+    /// Called only after the owner has stopped touching execution resources.
+    func release(_ id: UUID) {
+        guard current == id else { return }
+        current = nil
+        cancelled = false
+    }
+}

Sources/App/OpenTypeApp.swift

@@ -24,6 +24,8 @@ final class AppDelegate: NSObject, NSApplicationDelegate, ObservableObject {
     private var settingsWindowDelegate: SettingsWindowDelegate?
     private var onboardingWindow: NSWindow?
     private let popoverOutsideClickMonitor = PopoverOutsideClickMonitor()
+    let inputOwnership = InputSessionOwnership()
+    let speechEngineProvider = SpeechEngineProvider()
     let textProcessor: TextProcessor
     var cancellables = Set<AnyCancellable>()
     var iconTimer: Timer?
@@ -102,7 +104,7 @@ final class AppDelegate: NSObject, NSApplicationDelegate, ObservableObject {
     }
 
     private func setupPipeline() {
-        pipeline = VoicePipeline(appState: appState, textProcessor: textProcessor)
+        pipeline = VoicePipeline(appState: appState, textProcessor: textProcessor, ownership: inputOwnership, engineProvider: speechEngineProvider)
         Task { await pipeline?.warmUp() }
     }
 
@@ -150,26 +152,23 @@ final class AppDelegate: NSObject, NSApplicationDelegate, ObservableObject {
         action: HotkeyAction,
         remoteSessionToken: UInt64? = nil
     ) -> Task<Void, Never>? {
-        if integrationSessionCoordinator.isBusy {
-            pipeline?.showBusyHint()
-            return nil
-        }
         savePreviousApp()
         if popover.isShown { closePopover() }
         let mode: VoiceInputMode = action == .translation
             ? .translation(AppSettings.shared.translationTargetLanguage)
             : .dictation
+        let targetApp = previousApp
         return Task {
             await pipeline?.start(
                 mode: mode,
-                targetApp: previousApp,
+                targetApp: targetApp,
                 remoteSessionToken: remoteSessionToken
             )
         }
     }
 
     func stopRecording() {
-        Task { await pipeline?.stop(targetApp: previousApp) }
+        Task { await pipeline?.stop() }
     }
 
     private func applyPendingReplacement() async {

Sources/App/VoiceInputSettings.swift

@@ -0,0 +1,35 @@
+import Foundation
+
+/// Immutable choices for one utterance; settings changes apply to the next session.
+struct VoiceInputSettings {
+    let processing: TextProcessingOptions
+    let speech: SpeechEngineProvider.Selection
+    let dictionary: PersonalDictionarySnapshot
+    let outputMode: OutputMode
+    let enableInstantInsert: Bool
+    let enableMemory: Bool
+    let memoryWindowMinutes: Int
+    let useScreenContext: Bool
+    let streamingEnabled: Bool
+    let microphoneID: String?
+    let audioActivityThresholds: AudioActivityThresholds
+
+    var inputLanguage: InputLanguage { processing.inputLanguage }
+    var llmModel: String { processing.llmModel }
+    var espressoModelPath: String { processing.espressoModelPath }
+
+    @MainActor
+    init(settings: AppSettings, inputLanguage: InputLanguage? = nil) {
+        processing = TextProcessingOptions(settings: settings, inputLanguage: inputLanguage)
+        speech = SpeechEngineProvider.Selection(settings: settings, inputLanguage: inputLanguage)
+        dictionary = PersonalDictionary.shared.snapshot(settings: settings)
+        outputMode = settings.outputMode
+        enableInstantInsert = settings.enableInstantInsert
+        enableMemory = settings.enableMemory
+        memoryWindowMinutes = settings.memoryWindowMinutes
+        useScreenContext = settings.useScreenContext
+        streamingEnabled = settings.enableStreamingRecognitionBeta
+        microphoneID = settings.microphoneID
+        audioActivityThresholds = settings.audioActivityThresholds
+    }
+}

Sources/App/VoicePipeline+EditCommandHelpers.swift

@@ -4,7 +4,7 @@ import Foundation
 @MainActor
 extension VoicePipeline {
     func replacementInputContext(
-        settings: AppSettings,
+        settings: VoiceInputSettings,
         targetApp: NSRunningApplication?
     ) -> InputContext {
         InputContext.capture(
@@ -18,7 +18,7 @@ extension VoicePipeline {
 
     func finalizedReplacementText(
         _ text: String,
-        settings: AppSettings
+        settings: VoiceInputSettings
     ) -> String {
         textProcessor.cleanCommandGeneratedOutput(
             text,
@@ -29,7 +29,7 @@ extension VoicePipeline {
     func rewriteSelectedText(
         raw: String,
         intent: SelectionRewriteIntent,
-        settings: AppSettings,
+        settings: VoiceInputSettings,
         targetApp: NSRunningApplication?
     ) async {
         cancelScreenContextCapture()
@@ -48,11 +48,12 @@ extension VoicePipeline {
             inputLanguage: settings.inputLanguage,
             source: .menuBar
         )
-        var options = TextProcessingOptions(settings: settings)
+        var options = settings.processing
         options.llmModel = settings.llmModel
         let memoryContext = VoicePipelinePolicy.memoryContext(
             for: .command,
-            settings: settings,
+            enableMemory: settings.enableMemory,
+            memoryWindowMinutes: settings.memoryWindowMinutes,
             currentContext: context
         )
 
@@ -76,6 +77,7 @@ extension VoicePipeline {
         appState.statusMessage = L("pipeline.replacing")
 
         let result = await textInserter.replaceSelectedText(text: rewrittenText, targetApp: targetApp)
+        guard !Task.isCancelled else { return }
         appState.phase = .done
         appState.statusMessage = L("status.done")
         hideOverlayAfterDelay()

Sources/App/VoicePipeline+EditCommandResolution.swift

@@ -5,7 +5,7 @@ import Foundation
 extension VoicePipeline {
     func resolvedSpokenEditCommand(
         raw: String,
-        settings: AppSettings,
+        settings: VoiceInputSettings,
         targetApp: NSRunningApplication?
     ) async -> SpokenEditCommand? {
         guard VoicePipelinePolicy.shouldResolveEditCommandWithLLMFirst(outputMode: settings.outputMode) else {
@@ -47,10 +47,10 @@ extension VoicePipeline {
 
     private func resolveSpokenEditCommandWithLLM(
         raw: String,
-        settings: AppSettings,
+        settings: VoiceInputSettings,
         targetApp: NSRunningApplication?
     ) async -> SpokenEditCommandLLMResolution? {
-        var options = TextProcessingOptions(settings: settings)
+        var options = settings.processing
         options.llmModel = settings.llmModel
         return await textProcessor.resolveSpokenEditCommandResolution(
             text: raw,

Sources/App/VoicePipeline+EditCommands.swift

@@ -5,7 +5,7 @@ import Foundation
 extension VoicePipeline {
     func handleSpokenEditCommandIfNeeded(
         raw: String,
-        settings: AppSettings,
+        settings: VoiceInputSettings,
         targetApp: NSRunningApplication?
     ) async -> Bool {
         let expectedEspressoModelPath = settings.espressoModelPath
@@ -17,6 +17,7 @@ extension VoicePipeline {
             return false
         }
 
+        guard !Task.isCancelled else { return true }
         switch command {
         case .replaceLast(let replacementRaw):
             await replaceLastInsertion(
@@ -54,7 +55,7 @@ extension VoicePipeline {
 
         guard !Task.isCancelled else { return true }
         if let espressoOutcome = await consumeEspressoOutcome(
-            settings: settings,
+            settings: appState.settings,
             expectedEspressoModelPath: expectedEspressoModelPath
         ) {
             if case .error = appState.phase, espressoOutcome == .fallback {
@@ -74,7 +75,7 @@ extension VoicePipeline {
     private func replaceLastInsertion(
         raw: String,
         replacementRaw: String,
-        settings: AppSettings,
+        settings: VoiceInputSettings,
         targetApp: NSRunningApplication?
     ) async {
         cancelScreenContextCapture()
@@ -101,6 +102,7 @@ extension VoicePipeline {
             previouslyInserted: appState.lastInsertedText,
             targetApp: targetApp
         )
+        guard !Task.isCancelled else { return }
 
         appState.phase = .done
         appState.statusMessage = L("status.done")
@@ -130,7 +132,7 @@ extension VoicePipeline {
     private func replaceSelectedText(
         raw: String,
         replacementRaw: String,
-        settings: AppSettings,
+        settings: VoiceInputSettings,
         targetApp: NSRunningApplication?
     ) async {
         cancelScreenContextCapture()
@@ -148,6 +150,7 @@ extension VoicePipeline {
 
         Log.sensitive("[VoicePipeline] voice edit replace selection \(replacementText.count) chars")
         let result = await textInserter.replaceSelectedText(text: replacementText, targetApp: targetApp)
+        guard !Task.isCancelled else { return }
 
         appState.phase = .done
         appState.statusMessage = L("status.done")
@@ -183,6 +186,7 @@ extension VoicePipeline {
 
         Log.info("[VoicePipeline] voice edit delete selection")
         let result = await textInserter.deleteSelectedText(targetApp: targetApp)
+        guard !Task.isCancelled else { return }
 
         if case .probablyFailed(let reason) = result {
             Log.info("[VoicePipeline] voice edit delete selection probably failed: \(reason)")
@@ -213,6 +217,7 @@ extension VoicePipeline {
             previouslyInserted: appState.lastInsertedText,
             targetApp: targetApp
         )
+        guard !Task.isCancelled else { return }
 
         if case .probablyFailed(let reason) = result {
             Log.info("[VoicePipeline] voice edit undo probably failed: \(reason)")

Sources/App/VoicePipeline+ModelLifecycle.swift

@@ -5,7 +5,6 @@ extension VoicePipeline {
     func unloadLLM() {
         formattingPreloadGeneration += 1
         processingTask?.cancel()
-        processingTask = nil
         replacementTask?.cancel()
         replacementTask = nil
         appState.clearPendingReplacement()
@@ -37,7 +36,8 @@ extension VoicePipeline {
         let precedingTask = formattingModelLifecycleTask
         let task: Task<EspressoGenerationOutcome?, Never> = Task { @MainActor [weak self] in
             _ = await precedingTask?.value
-            guard let self else { return nil }
+            guard let self, let lease = try? self.ownership.acquire() else { return nil }
+            defer { self.ownership.release(lease) }
             return await self.preloadFormattingModelNow(
                 showFailureInStatus: showFailureInStatus
             )