From 89b1c12d5859d7d64196a39b513d42c3f6d0a970 Mon Sep 17 00:00:00 2001
From: Carsten Abele <carsten@ravensburg.dev>
Date: Thu, 9 Apr 2026 13:02:21 +0200
Subject: [PATCH] feat: auto-stop on silence, fix transcription hangs

- Auto-stop recording after 20 seconds of continuous silence
  (RMS-based silence detection in AudioCapture)
- Add WebSocket ping every 10s to keep connection alive
- Auto-reconnect if no events received for 30s (stale connection)
- Add diagnostic logging for all event types to help debug issues

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---
 MyVoxtral/MyVoxtral/Audio/AudioCapture.swift  | 39 +++++++++++++-
 .../Models/TranscriptionManager.swift         | 52 ++++++++++++++++---
 .../Network/VoxtralWebSocketClient.swift      |  8 +++
 3 files changed, 90 insertions(+), 9 deletions(-)

diff --git a/MyVoxtral/MyVoxtral/Audio/AudioCapture.swift b/MyVoxtral/MyVoxtral/Audio/AudioCapture.swift
index f72b3f4..23b949c 100644
--- a/MyVoxtral/MyVoxtral/Audio/AudioCapture.swift
+++ b/MyVoxtral/MyVoxtral/Audio/AudioCapture.swift
@@ -6,9 +6,17 @@ final class AudioCapture {
     private let chunkDurationMs: Double = 480
 
     var onChunk: ((Data) -> Void)?
+    var onSilenceTimeout: (() -> Void)?
+
+    /// RMS threshold below which audio is considered silence (16-bit PCM range)
+    var silenceThreshold: Int16 = 200
+    /// Seconds of continuous silence before triggering timeout
+    var silenceTimeoutSeconds: Double = 20
 
     private var buffer = Data()
     private let bytesPerChunk: Int
+    private var consecutiveSilentChunks: Int = 0
+    private var chunksPerSecond: Double { 1000.0 / chunkDurationMs }
 
     init() {
         // 16kHz * 2 bytes (16-bit) * 1 channel * 0.48s = 15360 bytes
@@ -58,12 +66,25 @@ final class AudioCapture {
             self.buffer.append(data)
 
             while self.buffer.count >= self.bytesPerChunk {
-                let chunk = self.buffer.prefix(self.bytesPerChunk)
+                let chunk = Data(self.buffer.prefix(self.bytesPerChunk))
                 self.buffer = Data(self.buffer.dropFirst(self.bytesPerChunk))
-                self.onChunk?(Data(chunk))
+
+                if self.isSilent(chunk) {
+                    self.consecutiveSilentChunks += 1
+                    let silentSeconds = Double(self.consecutiveSilentChunks) / self.chunksPerSecond
+                    if silentSeconds >= self.silenceTimeoutSeconds {
+                        self.onSilenceTimeout?()
+                        return
+                    }
+                } else {
+                    self.consecutiveSilentChunks = 0
+                }
+
+                self.onChunk?(chunk)
             }
         }
 
+        consecutiveSilentChunks = 0
         engine.prepare()
         try engine.start()
     }
@@ -78,6 +99,20 @@ final class AudioCapture {
             buffer = Data()
         }
     }
+
+    private func isSilent(_ chunk: Data) -> Bool {
+        chunk.withUnsafeBytes { rawBuffer in
+            guard let samples = rawBuffer.baseAddress?.assumingMemoryBound(to: Int16.self) else { return true }
+            let count = chunk.count / 2
+            var sumSquares: Int64 = 0
+            for i in 0..<count {
+                let sample = Int64(samples[i])
+                sumSquares += sample * sample
+            }
+            let rms = Int16(sqrt(Double(sumSquares) / Double(max(count, 1))))
+            return rms < silenceThreshold
+        }
+    }
 }
 
 enum AudioCaptureError: Error, LocalizedError {
diff --git a/MyVoxtral/MyVoxtral/Models/TranscriptionManager.swift b/MyVoxtral/MyVoxtral/Models/TranscriptionManager.swift
index 76f9312..27d350d 100644
--- a/MyVoxtral/MyVoxtral/Models/TranscriptionManager.swift
+++ b/MyVoxtral/MyVoxtral/Models/TranscriptionManager.swift
@@ -15,6 +15,8 @@ final class TranscriptionManager: ObservableObject {
     private let wsClient = VoxtralWebSocketClient()
     private let settings = AppSettings.shared
     private var hasRetried = false
+    private var lastEventTime = Date()
+    private var staleCheckTimer: Timer?
 
     var isRecording: Bool { state == .recording }
 
@@ -40,6 +42,8 @@ final class TranscriptionManager: ObservableObject {
         }
 
         wsClient.connect(apiKey: settings.apiKey, delayMs: settings.streamingDelayMs)
+        lastEventTime = Date()
+        startStaleCheck()
 
         audioCapture.onChunk = { [weak self] chunk in
             Task { @MainActor in
@@ -47,6 +51,13 @@ final class TranscriptionManager: ObservableObject {
             }
         }
 
+        audioCapture.onSilenceTimeout = { [weak self] in
+            Task { @MainActor in
+                print("[MyVoxtral] Silence timeout — stopping recording")
+                self?.stop()
+            }
+        }
+
         do {
             try audioCapture.start()
             state = .recording
@@ -56,6 +67,8 @@ final class TranscriptionManager: ObservableObject {
     }
 
     func stop() {
+        staleCheckTimer?.invalidate()
+        staleCheckTimer = nil
         audioCapture.stop()
         wsClient.flush()
         wsClient.disconnect()
@@ -66,34 +79,59 @@ final class TranscriptionManager: ObservableObject {
         }
     }
 
+    private func startStaleCheck() {
+        staleCheckTimer?.invalidate()
+        staleCheckTimer = Timer.scheduledTimer(withTimeInterval: 10, repeats: true) { [weak self] _ in
+            Task { @MainActor in
+                guard let self, self.isRecording else { return }
+                let elapsed = Date().timeIntervalSince(self.lastEventTime)
+                if elapsed > 30 {
+                    print("[MyVoxtral] Connection stale (\(Int(elapsed))s since last event) — reconnecting")
+                    self.wsClient.disconnect()
+                    self.wsClient.connect(apiKey: self.settings.apiKey, delayMs: self.settings.streamingDelayMs)
+                    self.lastEventTime = Date()
+                } else {
+                    self.wsClient.ping()
+                }
+            }
+        }
+    }
+
     private func handleEvent(_ event: VoxtralEvent) {
+        lastEventTime = Date()
+
         switch event {
         case .sessionCreated:
-            break
+            print("[MyVoxtral] Session created")
         case .textDelta(let text):
             currentText += text
             if settings.outputMode == .cursorInjection {
                 CursorInjector.typeText(text)
             }
-        case .segment:
-            break
-        case .language:
-            break
+        case .segment(let text, let start, let end):
+            print("[MyVoxtral] Segment [\(String(format: "%.1f", start))-\(String(format: "%.1f", end))]: \(text.prefix(60))")
+        case .language(let lang):
+            print("[MyVoxtral] Language detected: \(lang)")
         case .done(let text):
+            print("[MyVoxtral] Done, final text length: \(text.count)")
             if currentText.isEmpty {
                 currentText = text
             }
         case .error(let message):
+            print("[MyVoxtral] Error: \(message)")
             if !hasRetried && state == .recording {
                 hasRetried = true
                 wsClient.disconnect()
                 wsClient.connect(apiKey: settings.apiKey, delayMs: settings.streamingDelayMs)
+                lastEventTime = Date()
             } else {
                 state = .error(message)
+                staleCheckTimer?.invalidate()
+                staleCheckTimer = nil
                 audioCapture.stop()
             }
-        case .unknown:
-            break
+        case .unknown(let type):
+            print("[MyVoxtral] Unknown event: \(type)")
         }
     }
 }
diff --git a/MyVoxtral/MyVoxtral/Network/VoxtralWebSocketClient.swift b/MyVoxtral/MyVoxtral/Network/VoxtralWebSocketClient.swift
index ec4d1b8..9548f4c 100644
--- a/MyVoxtral/MyVoxtral/Network/VoxtralWebSocketClient.swift
+++ b/MyVoxtral/MyVoxtral/Network/VoxtralWebSocketClient.swift
@@ -47,6 +47,14 @@ final class VoxtralWebSocketClient {
         sendJSON(AudioFlushMessage())
     }
 
+    func ping() {
+        webSocketTask?.sendPing { error in
+            if let error {
+                print("[MyVoxtral] Ping failed: \(error.localizedDescription)")
+            }
+        }
+    }
+
     func disconnect() {
         intentionalDisconnect = true
         sendJSON(AudioEndMessage())