fix(mac): separate passive state sync from active voice dictation and force replace

mac/src/RelayClient.swift

@@ -17,8 +17,13 @@ public final class RelayClient: NSObject, URLSessionWebSocketDelegate {
     private var monitorTimerSource: DispatchSourceTimer?
     private let monitorQueue = DispatchQueue(label: "com.soniox.macsync.monitor", qos: .userInteractive)
     
+    // Callback: (text, cursor, isFullReplace)
     public var onRemoteUpdateReceived: ((String, Int?, Bool) -> Void)?
     
+    // Dedup guard: prevent rapid duplicate keystrokes within 200ms
+    private var lastInjectedText: String = ""
+    private var lastInjectedTime: Double = 0
+    
     override private init() {
         super.init()
         let config = URLSessionConfiguration.default
@@ -97,7 +102,7 @@ public final class RelayClient: NSObject, URLSessionWebSocketDelegate {
             return
         }
         
-        let action = json["action"] as? String ?? json["type"] as? String
+        let action = json["action"] as? String ?? json["type"] as? String ?? ""
         let source = json["source"] as? String ?? ""
         
         // Ignore echo messages originated by Mac itself
@@ -106,12 +111,42 @@ public final class RelayClient: NSObject, URLSessionWebSocketDelegate {
         let updateText = json["text"] as? String ?? json["insert_text"] as? String ?? ""
         let cursor = json["cursor"] as? Int ?? json["cursor_pos"] as? Int
         
-        if action == "paste" || action == "set_text" || action == "update_input" || action == "phone_input_edit" || action == "sync_state" || action == "insert_speech" {
-            let isFullReplace = (action == "update_input" || action == "set_text" || action == "force_replace")
-            print("RelayClient: ⚡ Received remote update: '\(updateText.prefix(30))...' (replace: \(isFullReplace), cursor: \(cursor ?? -1))")
+        let now = Date().timeIntervalSince1970
+        
+        // 1. Live Voice Dictation from Phone Mic: Paste ONLY new speech at active cursor (Cmd+V)
+        if action == "insert_speech" || action == "speech_insert" {
+            guard !updateText.isEmpty else { return }
+            
+            // Dedup exact duplicate within 250ms
+            if updateText == lastInjectedText && (now - lastInjectedTime) < 0.25 {
+                return
+            }
+            lastInjectedText = updateText
+            lastInjectedTime = now
+            
+            print("RelayClient: 🎙️ Voice Dictation Paste: '\(updateText.prefix(30))...' (Cmd+V at cursor)")
+            DispatchQueue.main.async {
+                self.onRemoteUpdateReceived?(updateText, cursor, false)
+            }
+        }
+        
+        // 2. Explicit "Send to Mac" Button on Phone: Full Replace (Cmd+A + Cmd+V)
+        else if action == "update_input" || action == "force_replace" || action == "paste" {
+            guard !updateText.isEmpty else { return }
+            
+            lastInjectedText = updateText
+            lastInjectedTime = now
+            
+            print("RelayClient: 🚀 Force Replace All: '\(updateText.prefix(30))...' (Cmd+A + Cmd+V)")
             DispatchQueue.main.async {
-                self.onRemoteUpdateReceived?(updateText, cursor, isFullReplace)
+                self.onRemoteUpdateReceived?(updateText, cursor, true)
+            }
         }
+        
+        // 3. Passive State Synchronization (typing on phone keyboard):
+        // Updates room state cache only — DOES NOT trigger keystrokes on Mac!
+        else if action == "sync_state" || action == "phone_input_edit" {
+            // Passive sync only
         }
     }