macos: polish voice overlay and remote command handling

apps/macos/Sources/Clawdis/ControlChannel.swift

@@ -170,27 +170,21 @@ final class ControlChannel: ObservableObject {
         self.connection = conn
 
         try await withCheckedThrowingContinuation { (cont: CheckedContinuation<Void, Error>) in
-            var resumed = false
+            conn.stateUpdateHandler = { [weak self, weak conn] state in
-            let resume: (Result<Void, Error>) -> Void = { result in
+                guard let self else { return }
-                guard !resumed else { return }
-                resumed = true
-                switch result {
-                case .success: cont.resume(returning: ())
-                case let .failure(err): cont.resume(throwing: err)
-                }
-            }
-
-            conn.stateUpdateHandler = { state in
                 switch state {
                 case .ready:
                     Task { @MainActor in self.state = .connected }
-                    resume(.success(()))
+                    conn?.stateUpdateHandler = nil
+                    cont.resume(returning: ())
                 case let .failed(err):
                     Task { @MainActor in self.state = .degraded(err.localizedDescription) }
-                    resume(.failure(err))
+                    conn?.stateUpdateHandler = nil
+                    cont.resume(throwing: err)
                 case let .waiting(err):
                     Task { @MainActor in self.state = .degraded(err.localizedDescription) }
-                    resume(.failure(err))
+                    conn?.stateUpdateHandler = nil
+                    cont.resume(throwing: err)
                 default:
                     break
                 }

apps/macos/Sources/Clawdis/Utilities.swift

@@ -350,21 +350,36 @@ enum CommandResolver {
         // Run the real clawdis CLI on the remote host; do not fall back to clawdis-mac.
         let exportedPath = "/opt/homebrew/bin:/usr/local/bin:/usr/bin:/bin:/usr/sbin:/sbin:/Users/steipete/Library/pnpm:$PATH"
         let quotedArgs = ([subcommand] + extraArgs).map(self.shellQuote).joined(separator: " ")
-        let userPRJ = settings.projectRoot
+        let userPRJ = settings.projectRoot.trimmingCharacters(in: .whitespacesAndNewlines)
-        let prjInit = userPRJ.isEmpty ? "" : "PRJ=\(self.shellQuote(userPRJ));"
+
-        let scriptBody = """
+        let projectSection: String
-        PATH=\(exportedPath);
+        if userPRJ.isEmpty {
-        \(prjInit)
+            projectSection = """
         DEFAULT_PRJ="$HOME/Projects/clawdis"
-        if [ -z "${PRJ:-}" ] && [ -d "$DEFAULT_PRJ" ]; then PRJ="$DEFAULT_PRJ"; fi
+        if [ -d "$DEFAULT_PRJ" ]; then
-        if [ -n "${PRJ:-}" ]; then
+          PRJ="$DEFAULT_PRJ"
           cd "$PRJ" || { echo "Project root not found: $PRJ"; exit 127; }
         fi
+        """
+        } else {
+            projectSection = """
+        PRJ=\(self.shellQuote(userPRJ))
+        cd \(self.shellQuote(userPRJ)) || { echo "Project root not found: \(userPRJ)"; exit 127; }
+        """
+        }
+
+        let scriptBody = """
+        PATH=\(exportedPath);
+        CLI="";
+        \(projectSection)
         if command -v clawdis >/dev/null 2>&1; then
+          CLI="$(command -v clawdis)"
           clawdis \(quotedArgs);
         elif [ -n "${PRJ:-}" ] && [ -f "$PRJ/bin/clawdis.js" ] && command -v node >/dev/null 2>&1; then
+          CLI="node $PRJ/bin/clawdis.js"
           node "$PRJ/bin/clawdis.js" \(quotedArgs);
         elif command -v pnpm >/dev/null 2>&1; then
+          CLI="pnpm --silent clawdis"
           pnpm --silent clawdis \(quotedArgs);
         else
           echo "clawdis CLI missing on remote host"; exit 127;

apps/macos/Sources/Clawdis/VoiceWakeOverlay.swift

@@ -284,27 +284,6 @@ final class VoiceWakeOverlayController: ObservableObject {
             ])
     }
 }
-
-private struct CloseHoverButton: View {
-    var onClose: () -> Void
-
-    var body: some View {
-        Button(action: self.onClose) {
-            Image(systemName: "xmark")
-                .font(.system(size: 12, weight: .bold))
-                .foregroundColor(Color.white.opacity(0.85))
-                .frame(width: 22, height: 22)
-                .background(Color.black.opacity(0.42))
-                .clipShape(Circle())
-                .shadow(color: Color.black.opacity(0.4), radius: 8, y: 2)
-        }
-        .buttonStyle(.plain)
-        .focusable(false)
-        .contentShape(Circle())
-        .padding(6)
-    }
-}
-
 private struct VoiceWakeOverlayView: View {
     @ObservedObject var controller: VoiceWakeOverlayController
     @FocusState private var textFocused: Bool
@@ -334,7 +313,7 @@ private struct VoiceWakeOverlayView: View {
                             self.controller.sendNow()
                         })
                         .focused(self.$textFocused)
-                        .frame(minHeight: 32, maxHeight: .infinity)
+                        .frame(maxWidth: .infinity, minHeight: 32, maxHeight: .infinity, alignment: .topLeading)
                         .id("editing")
                 } else {
                     VibrantLabelView(
@@ -343,7 +322,7 @@ private struct VoiceWakeOverlayView: View {
                             self.controller.userBeganEditing()
                             self.textFocused = true
                         })
-                        .frame(minHeight: 32, maxHeight: .infinity)
+                        .frame(maxWidth: .infinity, minHeight: 32, maxHeight: .infinity, alignment: .topLeading)
                         .id("display")
                 }
 
@@ -450,7 +429,7 @@ private struct TranscriptTextView: NSViewRepresentable {
         let scroll = NSScrollView()
         scroll.drawsBackground = false
         scroll.borderType = .noBorder
-        scroll.hasVerticalScroller = self.isOverflowing
+        scroll.hasVerticalScroller = true
         scroll.autohidesScrollers = true
         scroll.scrollerStyle = .overlay
         scroll.hasHorizontalScroller = false
@@ -502,8 +481,7 @@ private struct VibrantLabelView: NSViewRepresentable {
     var onTap: () -> Void
 
     func makeNSView(context: Context) -> NSView {
-        let display = self.attributed.strippingForegroundColor()
+        let label = NSTextField(labelWithAttributedString: self.attributed)
-        let label = NSTextField(labelWithAttributedString: display)
         label.isEditable = false
         label.isBordered = false
         label.drawsBackground = false
@@ -514,7 +492,9 @@ private struct VibrantLabelView: NSViewRepresentable {
         label.cell?.isScrollable = false
         label.setContentHuggingPriority(.defaultLow, for: .horizontal)
         label.setContentCompressionResistancePriority(.defaultLow, for: .horizontal)
-        label.textColor = .textColor
+        label.setContentHuggingPriority(.required, for: .vertical)
+        label.setContentCompressionResistancePriority(.required, for: .vertical)
+        label.textColor = .labelColor
 
         let container = ClickCatcher(onTap: onTap)
         container.addSubview(label)
@@ -532,7 +512,7 @@ private struct VibrantLabelView: NSViewRepresentable {
     func updateNSView(_ nsView: NSView, context: Context) {
         guard let container = nsView as? ClickCatcher,
               let label = container.subviews.first as? NSTextField else { return }
-        label.attributedStringValue = self.attributed.strippingForegroundColor()
+        label.attributedStringValue = self.attributed
     }
 
 }
@@ -588,15 +568,6 @@ private struct CloseButtonOverlay: View {
         .allowsHitTesting(isVisible)
     }
 }
-}
-
-private extension NSAttributedString {
-    func strippingForegroundColor() -> NSAttributedString {
-        let mutable = NSMutableAttributedString(attributedString: self)
-        mutable.removeAttribute(.foregroundColor, range: NSRange(location: 0, length: mutable.length))
-        return mutable
-    }
-}
 
 private final class TranscriptNSTextView: NSTextView {
     var onSend: (() -> Void)?

apps/macos/Sources/Clawdis/VoiceWakeRuntime.swift

@@ -17,6 +17,7 @@ actor VoiceWakeRuntime {
     private var recognitionRequest: SFSpeechAudioBufferRecognitionRequest?
     private var recognitionTask: SFSpeechRecognitionTask?
     private var lastHeard: Date?
+    private var noiseFloorRMS: Double = 1e-4
     private var captureStartedAt: Date?
     private var captureTask: Task<Void, Never>?
     private var capturedTranscript: String = ""
@@ -36,6 +37,19 @@ actor VoiceWakeRuntime {
     // Maximum capture duration from trigger until we force-send, to avoid runaway sessions.
     private let captureHardStop: TimeInterval = 120.0
     private let debounceAfterSend: TimeInterval = 0.35
+    // Voice activity detection parameters (RMS-based).
+    private let minSpeechRMS: Double = 1e-3
+    private let speechBoostFactor: Double = 6.0 // how far above noise floor we require to mark speech
+
+    /// Stops the active Speech pipeline without clearing the stored config, so we can restart cleanly.
+    private func haltRecognitionPipeline() {
+        self.recognitionTask?.cancel()
+        self.recognitionTask = nil
+        self.recognitionRequest?.endAudio()
+        self.recognitionRequest = nil
+        self.audioEngine.inputNode.removeTap(onBus: 0)
+        self.audioEngine.stop()
+    }
 
     struct RuntimeConfig: Equatable {
         let triggers: [String]
@@ -94,8 +108,11 @@ actor VoiceWakeRuntime {
             let input = self.audioEngine.inputNode
             let format = input.outputFormat(forBus: 0)
             input.removeTap(onBus: 0)
-            input.installTap(onBus: 0, bufferSize: 2048, format: format) { [weak request] buffer, _ in
+            input.installTap(onBus: 0, bufferSize: 2048, format: format) { [weak self, weak request] buffer, _ in
                 request?.append(buffer)
+                if let rms = Self.rmsLevel(buffer: buffer) {
+                    Task { await self?.noteAudioLevel(rms: rms) }
+                }
             }
 
             self.audioEngine.prepare()
@@ -266,6 +283,8 @@ actor VoiceWakeRuntime {
         self.captureTask = nil
 
         let finalTranscript = self.capturedTranscript.trimmingCharacters(in: .whitespacesAndNewlines)
+        // Stop further recognition events so we don't retrigger immediately with buffered audio.
+        self.haltRecognitionPipeline()
         self.capturedTranscript = ""
         self.captureStartedAt = nil
         self.lastHeard = nil