From 9a38c329aa4ac466067631bd4a2187035f401081 Mon Sep 17 00:00:00 2001
From: Ali Alavi <aligopalpour13@gmail.com>
Date: Wed, 26 Aug 2026 11:03:18 +0000
Subject: [PATCH] fix(sync): single-instance mac enforcement to stop duplicate
 pastes, unified speech sync pipeline from android

---
 .../java/com/soniox/remotemic/MainActivity.kt     |  9 ++++++++-
 mac/src/AppDelegate.swift                         | 10 ++++++++++
 server/relay_server.py                            | 15 +--------------
 3 files changed, 19 insertions(+), 15 deletions(-)

diff --git a/android/app/src/main/java/com/soniox/remotemic/MainActivity.kt b/android/app/src/main/java/com/soniox/remotemic/MainActivity.kt
index 3b1dd52..c2d9f3e 100644
--- a/android/app/src/main/java/com/soniox/remotemic/MainActivity.kt
+++ b/android/app/src/main/java/com/soniox/remotemic/MainActivity.kt
@@ -283,6 +283,7 @@ class MainActivity : AppCompatActivity() {
         isApplyingRemoteUpdate = true
         lastLocalText = mergedText
         lastLocalUserEditTime = System.currentTimeMillis()
+        currentRevision++
         binding.etTranscript.setText(mergedText)
         binding.etTranscript.setTextColor(Color.WHITE)
         binding.etTranscript.setSelection(newCursor)
@@ -290,7 +291,13 @@ class MainActivity : AppCompatActivity() {
         binding.tvCharCount.text = "$wordCount کلمه"
         isApplyingRemoteUpdate = false
         
-        AppLogger.log("Main", "تزریق گفتار در نشانگر: '$trimmedSpeech' (موقعیت جدید: $newCursor)")
+        // Guarantees immediate 1:1 synchronization with Mac input box
+        streamDictationClient?.sendPhoneEdit(mergedText, newCursor)
+        lifecycleScope.launch {
+            sendDirectPaste(mergedText, newCursor)
+        }
+        
+        AppLogger.log("Main", "تزریق گفتار در نشانگر و همگام‌سازی با مک: '$trimmedSpeech' (موقعیت جدید: $newCursor)")
     }
 
     private fun setupUI() {
diff --git a/mac/src/AppDelegate.swift b/mac/src/AppDelegate.swift
index 25b0fa3..8db1a1f 100644
--- a/mac/src/AppDelegate.swift
+++ b/mac/src/AppDelegate.swift
@@ -16,6 +16,16 @@ public final class AppDelegate: NSObject, NSApplicationDelegate {
     private var remoteListener: NWListener?
     
     public func applicationDidFinishLaunching(_ notification: Notification) {
+        // 1. Enforce strict single-instance: Terminate any previous/duplicate running instances of SonioxVoice
+        let currentPid = ProcessInfo.processInfo.processIdentifier
+        let apps = NSRunningApplication.runningApplications(withBundleIdentifier: Bundle.main.bundleIdentifier ?? "com.soniox.voice")
+        for app in apps {
+            if app.processIdentifier != currentPid {
+                print("AppDelegate: 🛑 Terminating duplicate instance PID \(app.processIdentifier)")
+                app.terminate()
+            }
+        }
+        
         // Completely silent operation by default
         UserDefaults.standard.set(false, forKey: "SonioxPlaySounds")
         
diff --git a/server/relay_server.py b/server/relay_server.py
index 4a8c095..ff28dd1 100644
--- a/server/relay_server.py
+++ b/server/relay_server.py
@@ -298,20 +298,7 @@ async def handle_phone_stream_ws(request):
                     clean_final = sanitize_text(raw_final, allow_multiline=False, preserve_trailing_space=True)
                     logger.info("⚡ Session %s final text: '%s'", sid, clean_final)
 
-                    # 1. DIRECT FAST-PATH TO MAC (Sub-10ms injection without waiting for phone roundtrip!)
-                    if clean_final:
-                        mac_payload = {
-                            "type": "insert_speech",
-                            "action": "insert_speech",
-                            "source": "server_direct",
-                            "text": clean_final,
-                            "session_id": sid,
-                            "timestamp": time.time()
-                        }
-                        await broadcast_state(mac_payload)
-                        logger.info("🚀 Fast-Path: Directly dispatched speech '%s' to Mac", clean_final)
-
-                    # 2. Return final speech text to phone (for passive UI mirroring)
+                    # Return final speech text to phone (phone updates UI and broadcasts full state to Mac)
                     if is_ws_open(ws):
                         await ws.send_str(json.dumps({
                             "type": "final",