tts support (#962)

* initial TTS setup

* minor improvements

* fixed incorrect download state of tts model

* fixed cell reuse bug which was causing action buttons to hide

* added voice picker

* added TTS as a tool

* improved visual updates for tts tool call

* fixed a bug in speaker button in message cell

* stop audio playback at appropriate places

Author: RaajeevChandran

Packages/OsaurusCore/AppDelegate.swift

@@ -90,6 +90,10 @@ public final class AppDelegate: NSObject, NSApplicationDelegate, NSPopoverDelega
         // Configure local notifications
         NotificationService.shared.configureOnLaunch()
 
+        // If PocketTTS models are already on disk, preload them so the first
+        // speaker tap plays immediately without routing to settings.
+        TTSService.shared.refreshModelState()
+
         // Set up observers for server state changes
         setupObservers()
 
@@ -394,6 +398,14 @@ public final class AppDelegate: NSObject, NSApplicationDelegate, NSPopoverDelega
             object: nil
         )
 
+        // Route "user tapped speaker but model isn't ready" to the TTS settings tab.
+        NotificationCenter.default.addObserver(
+            self,
+            selector: #selector(handleOpenTTSSettings(_:)),
+            name: .openTTSSettingsRequested,
+            object: nil
+        )
+
         // Listen for chat view closed to resume VAD
         NotificationCenter.default.addObserver(
             self,
@@ -487,6 +499,13 @@ public final class AppDelegate: NSObject, NSApplicationDelegate, NSPopoverDelega
         }
     }
 
+    @objc private func handleOpenTTSSettings(_ notification: Notification) {
+        Task { @MainActor in
+            ManagementStateManager.shared.voiceSubTabRequest = "TTS"
+            showManagementWindow(initialTab: .voice)
+        }
+    }
+
     public func application(_ application: NSApplication, open urls: [URL]) {
         for url in urls {
             handleDeepLink(url)

Packages/OsaurusCore/Folder/ChatExecutionContext.swift

@@ -24,4 +24,12 @@ public enum ChatExecutionContext {
 
     /// The agent ID whose context is active for the current execution.
     @TaskLocal public static var currentAgentId: UUID?
+
+    /// Assistant turn dispatching the current tool call. Used by `speak`
+    /// to bind TTS playback to the right message bubble
+    @TaskLocal public static var currentAssistantTurnId: UUID?
+
+    /// Specific tool invocation id. Used by `speak` so the inline card
+    /// can swap its check for a spinner while its audio plays
+    @TaskLocal public static var currentToolCallId: String?
 }

Packages/OsaurusCore/Managers/Chat/ChatWindowState.swift

@@ -136,6 +136,7 @@ final class ChatWindowState: ObservableObject {
     }
 
     func switchAgent(to newAgentId: UUID) {
+        TTSService.shared.stop()
         if !session.turns.isEmpty { session.save() }
         agentId = newAgentId
         removeEphemeralProviderIfNeeded()
@@ -150,6 +151,7 @@ final class ChatWindowState: ObservableObject {
     }
 
     func startNewChat() {
+        TTSService.shared.stop()
         if !session.turns.isEmpty { session.save() }
         flushCurrentSession()
         session.reset(for: agentId)
@@ -158,6 +160,7 @@ final class ChatWindowState: ObservableObject {
 
     func loadSession(_ sessionData: ChatSessionData) {
         guard sessionData.id != session.sessionId else { return }
+        TTSService.shared.stop()
         if !session.turns.isEmpty { session.save() }
         flushCurrentSession()
 

Packages/OsaurusCore/Managers/ManagementStateManager.swift

@@ -16,5 +16,9 @@ public final class ManagementStateManager: ObservableObject {
     /// Persists the last selected tab within the current app session.
     @Published public var selectedTab: ManagementTab = .settings
 
+    /// One-shot request to focus a specific sub-tab inside `VoiceView`.
+    /// VoiceView observes this and resets it to nil after applying.
+    @Published public var voiceSubTabRequest: String?
+
     private init() {}
 }