fix: minimax reasoning in macos

Signed-off-by: xunzhuo <xunzhuo@vllm-semantic-router.ai>

Author: Xunzhuo

apps/macos/Sources/AppModel.swift

@@ -160,13 +160,24 @@ struct ChatMessage: Identifiable, Equatable {
         case system
     }
 
+    enum InputModality: Equatable {
+        case text
+        case voice
+    }
+
     var id = UUID()
     var role: Role
     var text: String
     var date = Date()
     var attachments: [WakeAttachment] = []
     var toolEvents: [ToolUseEvent] = []
     var isStreaming = false
+    var inputModality: InputModality = .text
+    var voiceDuration: TimeInterval?
+
+    var isVoiceMessage: Bool {
+        role == .user && inputModality == .voice
+    }
 }
 
 struct WakeAttachment: Identifiable, Equatable {
@@ -189,8 +200,13 @@ struct WakeQueuedPrompt: Identifiable, Equatable {
     var text: String
     var attachments: [WakeAttachment] = []
     var date = Date()
+    var inputModality: ChatMessage.InputModality = .text
+    var voiceDuration: TimeInterval?
 
     var previewText: String {
+        if inputModality == .voice {
+            return "Voice message"
+        }
         let trimmed = text.trimmingCharacters(in: .whitespacesAndNewlines)
         if !trimmed.isEmpty {
             return trimmed
@@ -1590,7 +1606,6 @@ final class ElephantAppModel: ObservableObject {
                 if status.contains("completed") || status.contains("succeeded") || status == "success" {
                     onboardingFinalizationStatus = text(.learningReady)
                     onboardingFinalizationComplete = true
-                    onboardingStep = 17
                     scheduleOnboardingAutoCompletion()
                     return
                 }
@@ -1602,7 +1617,6 @@ final class ElephantAppModel: ObservableObject {
         }
         onboardingFinalizationStatus = text(.learningReady)
         onboardingFinalizationComplete = true
-        onboardingStep = 17
         scheduleOnboardingAutoCompletion()
     }
 
@@ -2787,12 +2801,20 @@ final class ElephantAppModel: ObservableObject {
     }
 
     func sendWakeMessage() async {
+        await enqueueWakeMessage(inputModality: .text, voiceDuration: nil)
+    }
+
+    func sendVoiceWakeMessage(duration: TimeInterval?) async {
+        await enqueueWakeMessage(inputModality: .voice, voiceDuration: duration)
+    }
+
+    private func enqueueWakeMessage(inputModality: ChatMessage.InputModality, voiceDuration: TimeInterval?) async {
         let text = wakeDraft.trimmingCharacters(in: .whitespacesAndNewlines)
         let attachments = wakeAttachments
         guard !text.isEmpty || !attachments.isEmpty else { return }
         wakeDraft = ""
         wakeAttachments = []
-        wakeQueue.append(WakeQueuedPrompt(text: text, attachments: attachments))
+        wakeQueue.append(WakeQueuedPrompt(text: text, attachments: attachments, inputModality: inputModality, voiceDuration: voiceDuration))
         focusComposer()
         await drainWakeQueueIfNeeded()
     }
@@ -2928,12 +2950,22 @@ final class ElephantAppModel: ObservableObject {
         }
         while !wakeQueue.isEmpty {
             let item = wakeQueue.removeFirst()
-            await runWakeMessage(item.text, attachments: item.attachments)
+            await runWakeMessage(
+                item.text,
+                attachments: item.attachments,
+                inputModality: item.inputModality,
+                voiceDuration: item.voiceDuration
+            )
         }
     }
 
-    private func runWakeMessage(_ text: String, attachments: [WakeAttachment]) async {
-        messages.append(ChatMessage(role: .user, text: text, attachments: attachments))
+    private func runWakeMessage(
+        _ text: String,
+        attachments: [WakeAttachment],
+        inputModality: ChatMessage.InputModality,
+        voiceDuration: TimeInterval?
+    ) async {
+        messages.append(ChatMessage(role: .user, text: text, attachments: attachments, inputModality: inputModality, voiceDuration: voiceDuration))
         chatScrollRevision += 1
 
         let prompt = Self.wakePrompt(text: text, attachments: attachments)

apps/macos/Sources/DesignSystem.swift

@@ -357,6 +357,7 @@ struct ProviderLogoView: View {
         if normalized.contains("openclaw") { return "openclaw" }
         if normalized.contains("opencode") { return "opencode" }
         if normalized.contains("kimi") || normalized.contains("moonshot") { return "kimi" }
+        if normalized.contains("minimax") { return "minimax" }
         if normalized == "pi" || normalized.contains("inflection") { return "pi" }
         if normalized.contains("codex") || normalized.contains("openai") { return "codex" }
         return normalized.isEmpty ? "agent" : normalized
@@ -392,6 +393,7 @@ struct ProviderLogoView: View {
         if normalizedProvider.contains("openclaw") { return "OpenClaw" }
         if normalizedProvider.contains("opencode") { return "OpenCode" }
         if normalizedProvider.contains("kimi") || normalizedProvider.contains("moonshot") { return "Kimi" }
+        if normalizedProvider.contains("minimax") { return "MiniMax" }
         if normalizedProvider.contains("codex") || normalizedProvider.contains("openai") { return "Codex" }
         return providerID.isEmpty ? "Local agent" : providerID
     }
@@ -405,6 +407,7 @@ struct ProviderLogoView: View {
         if normalizedProvider.contains("openclaw") { return "hand.raised" }
         if normalizedProvider.contains("opencode") { return "chevron.left.forwardslash.chevron.right" }
         if normalizedProvider.contains("kimi") || normalizedProvider.contains("moonshot") { return "moon.stars" }
+        if normalizedProvider.contains("minimax") { return "sparkles" }
         if normalizedProvider.contains("codex") || normalizedProvider.contains("openai") { return "terminal" }
         return "terminal"
     }
@@ -417,6 +420,7 @@ struct ProviderLogoView: View {
         if normalizedProvider.contains("hermes") { return ElephantTheme.orange }
         if normalizedProvider.contains("openclaw") { return ElephantTheme.ember }
         if normalizedProvider.contains("kimi") || normalizedProvider.contains("moonshot") { return ElephantTheme.green }
+        if normalizedProvider.contains("minimax") { return ElephantTheme.green }
         return ElephantTheme.accent
     }
 }

apps/macos/Sources/Resources/ProviderIcons/minimax.png

@@ -6,6 +6,9 @@ import Speech
 final class SpeechInputController: NSObject, ObservableObject {
     @Published private(set) var isRecording = false
     @Published private(set) var statusText = ""
+    @Published private(set) var recordingStartedAt: Date?
+    @Published private(set) var capturedDuration: TimeInterval = 0
+    @Published private(set) var recognizedText = ""
 
     private let audioEngine = AVAudioEngine()
     private let recognizer = SFSpeechRecognizer(locale: Locale.current)
@@ -26,6 +29,9 @@ final class SpeechInputController: NSObject, ObservableObject {
         guard !isRecording else { return }
         baseText = text.trimmingCharacters(in: .whitespacesAndNewlines)
         self.onText = onText
+        recognizedText = ""
+        capturedDuration = 0
+        recordingStartedAt = nil
         statusText = "Requesting microphone access..."
 
         requestMicrophoneAccess { [weak self] allowed in
@@ -50,16 +56,28 @@ final class SpeechInputController: NSObject, ObservableObject {
 
     func stop() {
         guard isRecording || audioEngine.isRunning else { return }
+        updateCapturedDuration()
         audioEngine.stop()
         audioEngine.inputNode.removeTap(onBus: 0)
         recognitionRequest?.endAudio()
         recognitionTask?.cancel()
         recognitionRequest = nil
         recognitionTask = nil
         isRecording = false
+        recordingStartedAt = nil
         statusText = "Voice input stopped."
     }
 
+    func resetCapture() {
+        if isRecording || audioEngine.isRunning {
+            stop()
+        }
+        recognizedText = ""
+        capturedDuration = 0
+        recordingStartedAt = nil
+        statusText = ""
+    }
+
     private func requestMicrophoneAccess(_ completion: @escaping (Bool) -> Void) {
         switch AVCaptureDevice.authorizationStatus(for: .audio) {
         case .authorized:
@@ -105,12 +123,15 @@ final class SpeechInputController: NSObject, ObservableObject {
         }
 
         isRecording = true
+        recordingStartedAt = Date()
+        capturedDuration = 0
         statusText = "Listening..."
         recognitionTask = recognizer.recognitionTask(with: recognitionRequest) { [weak self] result, error in
             Task { @MainActor [weak self] in
                 guard let self else { return }
                 if let result {
                     let spoken = result.bestTranscription.formattedString.trimmingCharacters(in: .whitespacesAndNewlines)
+                    self.recognizedText = spoken
                     let combined = [self.baseText, spoken].filter { !$0.isEmpty }.joined(separator: " ")
                     self.onText?(combined)
                     self.statusText = result.isFinal ? "Voice input captured." : "Listening..."
@@ -124,6 +145,7 @@ final class SpeechInputController: NSObject, ObservableObject {
     }
 
     private func finishRecognition() {
+        updateCapturedDuration()
         if audioEngine.isRunning {
             audioEngine.stop()
             audioEngine.inputNode.removeTap(onBus: 0)
@@ -132,5 +154,12 @@ final class SpeechInputController: NSObject, ObservableObject {
         recognitionRequest = nil
         recognitionTask = nil
         isRecording = false
+        recordingStartedAt = nil
+    }
+
+    private func updateCapturedDuration() {
+        if let recordingStartedAt {
+            capturedDuration = max(capturedDuration, Date().timeIntervalSince(recordingStartedAt))
+        }
     }
 }