feat: user personalization with Honcho-backed preferences

- WelcomeScreen: first-time name entry with cute onboarding
- identify WS message: sets user_id, loads saved prefs from Honcho
- set_preference WS message: saves scene/outfit/accessory to Honcho metadata
- Preferences auto-load on return visits via localStorage + Honcho peer meta
- Kira uses the user's name in greeting and prompts
- Backend: get/set preference methods in KiraMemory service
- Frontend: optimistic preference updates, synced to backend on change

backend/main.py

@@ -6,6 +6,7 @@ Real-time speech-to-speech pipeline:
 Honcho memory integration:
   Cross-session user context injected into LLM prompts,
   conversation exchanges stored for continuous learning.
+  User preferences (name, scene, outfit, accessory) persisted in peer metadata.
 """
 
 import json
@@ -65,10 +66,8 @@ def build_system_prompt(user_id: str) -> dict:
     """Build system prompt with Honcho memory context injected."""
     base = BASE_SYSTEM_PROMPT
 
-    # Append memory context if Honcho is available
     if kira_memory.enabled:
         try:
-            # Get user-specific context from Honcho
             kira_memory.ensure_peers(user_id)
             memory_suffix = kira_memory.build_system_prompt_suffix()
             if memory_suffix:
@@ -79,35 +78,89 @@ def build_system_prompt(user_id: str) -> dict:
     return {"role": "system", "content": base}
 
 
+def handle_identify(msg: dict, session_id: str) -> dict | None:
+    """Handle user identification. Returns user preferences or None."""
+    user_id = msg.get("user_id", "").strip()
+    if not user_id:
+        return {"type": "error", "message": "user_id is required"}
+
+    user_name = msg.get("name", "").strip()
+    if user_name:
+        kira_memory.set_user_preference(user_id, "name", user_name)
+
+    prefs = kira_memory.get_user_preferences(user_id)
+    logger.info(f"[{session_id}] Identified as {user_id} (name={user_name or prefs.get('name', '')})")
+
+    return {
+        "type": "identified",
+        "user_id": user_id,
+        "preferences": prefs,
+    }
+
+
+def handle_set_preference(msg: dict, session_id: str, user_id: str) -> dict | None:
+    """Handle preference update. Returns success status."""
+    if not user_id or user_id == "default-user":
+        return {"type": "error", "message": "Must identify first"}
+
+    key = msg.get("key", "").strip()
+    value = msg.get("value", "").strip()
+
+    if not key:
+        return {"type": "error", "message": "key is required"}
+
+    ok = kira_memory.set_user_preference(user_id, key, value)
+    return {
+        "type": "preference_saved",
+        "key": key,
+        "success": ok,
+    }
+
+
 @app.websocket("/api/ws")
 async def conversation_ws(websocket: WebSocket):
     await websocket.accept()
     session_id = str(uuid.uuid4())[:8]
     user_id = "default-user"
+    identified = False
     logger.info(f"[{session_id}] WebSocket connected")
 
-    # Audio buffer accumulates chunks from one utterance
     audio_buffer = bytearray()
     conversation_history: list[dict] = []
 
-    # Initialize Honcho for this session
-    if kira_memory.enabled:
     try:
-            kira_memory.ensure_peers(user_id)
-            kira_memory.ensure_session(session_id)
-            logger.info(f"[{session_id}] Honcho session ready")
-        except Exception as e:
-            logger.warning(f"[{session_id}] Honcho setup failed: {e}")
-
-    try:
-        first_message = True
-
         while True:
             raw = await websocket.receive_text()
             msg = json.loads(raw)
             msg_type = msg.get("type", "")
 
-            # Build system prompt fresh each turn to get latest Honcho context
+            # ── Identity & Preferences ──
+
+            if msg_type == "identify":
+                response = handle_identify(msg, session_id)
+                if response:
+                    await websocket.send_json(response)
+                    if response["type"] == "identified":
+                        user_id = response["user_id"]
+                        identified = True
+                        # Set up Honcho for this user
+                        if kira_memory.enabled:
+                            try:
+                                kira_memory.ensure_peers(user_id)
+                                kira_memory.ensure_session(session_id)
+                                logger.info(f"[{session_id}] Honcho session ready for {user_id}")
+                            except Exception as e:
+                                logger.warning(f"[{session_id}] Honcho setup failed: {e}")
+                continue
+
+            if msg_type == "set_preference":
+                response = handle_set_preference(msg, session_id, user_id)
+                if response:
+                    await websocket.send_json(response)
+                continue
+
+            # ── Conversation ──
+
             system_prompt = build_system_prompt(user_id)
 
             if msg_type == "audio_chunk":
@@ -121,7 +174,6 @@ async def conversation_ws(websocket: WebSocket):
 
                 logger.info(f"[{session_id}] Transcribing {len(audio_buffer)} bytes...")
 
-                # 1. Speech-to-text
                 transcript = await transcribe_audio(bytes(audio_buffer))
                 audio_buffer.clear()
 
@@ -129,45 +181,27 @@ async def conversation_ws(websocket: WebSocket):
                     await websocket.send_json({"type": "error", "message": "Could not transcribe audio"})
                     continue
 
-                # Echo transcript
+                await websocket.send_json({"type": "transcript", "text": transcript})
-                await websocket.send_json({
-                    "type": "transcript",
-                    "text": transcript,
-                })
 
-                # 2. LLM call
                 logger.info(f"[{session_id}] User: {transcript}")
-                user_msg = {"role": "user", "content": transcript}
+                conversation_history.append({"role": "user", "content": transcript})
-                conversation_history.append(user_msg)
 
                 messages = [system_prompt] + conversation_history[-10:]
                 kira_text = await get_kira_response(messages)
 
-                assistant_msg = {"role": "assistant", "content": kira_text}
+                conversation_history.append({"role": "assistant", "content": kira_text})
-                conversation_history.append(assistant_msg)
                 logger.info(f"[{session_id}] Kira: {kira_text}")
 
-                # 3. Store in Honcho
+                if kira_memory.enabled and identified:
-                if kira_memory.enabled:
                     try:
                         kira_memory.store_messages(transcript, kira_text)
                     except Exception as e:
                         logger.warning(f"[{session_id}] Failed to store messages: {e}")
 
-                # 4. TTS
+                await websocket.send_json({"type": "speaking_start", "text": kira_text})
-                await websocket.send_json({
-                    "type": "speaking_start",
-                    "text": kira_text,
-                })
-
                 audio_bytes = await synthesize_speech(kira_text)
                 audio_b64 = base64.b64encode(audio_bytes).decode("utf-8")
-                await websocket.send_json({
+                await websocket.send_json({"type": "audio", "data": audio_b64, "text": kira_text})
-                    "type": "audio",
-                    "data": audio_b64,
-                    "text": kira_text,
-                })
-
                 await websocket.send_json({"type": "speaking_end"})
 
             elif msg_type == "ping":
@@ -179,32 +213,24 @@ async def conversation_ws(websocket: WebSocket):
                     continue
 
                 logger.info(f"[{session_id}] User (text): {user_text}")
-                user_msg = {"role": "user", "content": user_text}
+                conversation_history.append({"role": "user", "content": user_text})
-                conversation_history.append(user_msg)
 
                 messages = [system_prompt] + conversation_history[-10:]
                 kira_text = await get_kira_response(messages)
 
-                assistant_msg = {"role": "assistant", "content": kira_text}
+                conversation_history.append({"role": "assistant", "content": kira_text})
-                conversation_history.append(assistant_msg)
                 logger.info(f"[{session_id}] Kira: {kira_text}")
 
-                # Store in Honcho
+                if kira_memory.enabled and identified:
-                if kira_memory.enabled:
                     try:
                         kira_memory.store_messages(user_text, kira_text)
                     except Exception as e:
                         logger.warning(f"[{session_id}] Failed to store messages: {e}")
 
-                # TTS
                 await websocket.send_json({"type": "speaking_start", "text": kira_text})
                 audio_bytes = await synthesize_speech(kira_text)
                 audio_b64 = base64.b64encode(audio_bytes).decode("utf-8")
-                await websocket.send_json({
+                await websocket.send_json({"type": "audio", "data": audio_b64, "text": kira_text})
-                    "type": "audio",
-                    "data": audio_b64,
-                    "text": kira_text,
-                })
                 await websocket.send_json({"type": "speaking_end"})
 
     except WebSocketDisconnect:

backend/services/memory.py

@@ -178,6 +178,54 @@ class KiraMemory:
         except Exception as e:
             logger.warning(f"Failed to store Kira message: {e}")
 
+    # ─── User preferences (stored in Honcho peer metadata) ───
+
+    DEFAULT_PREFERENCES: dict[str, str] = {
+        "name": "",
+        "scene": "cozy-room",
+        "outfit": "cozy-hoodie",
+        "accessory": "",
+    }
+
+    def get_user_preferences(self, user_id: str) -> dict[str, str]:
+        """Load user preferences from Honcho peer metadata."""
+        if not self.enabled:
+            return dict(self.DEFAULT_PREFERENCES)
+
+        try:
+            peer = self._honcho.peer(user_id)
+            meta = peer.metadata or {}
+            prefs = dict(self.DEFAULT_PREFERENCES)
+            for key in prefs:
+                val = meta.get(key)
+                if val is not None:
+                    prefs[key] = str(val)
+            logger.info(f"Loaded preferences for {user_id}: {prefs}")
+            return prefs
+        except Exception as e:
+            logger.warning(f"Failed to load preferences: {e}")
+            return dict(self.DEFAULT_PREFERENCES)
+
+    def set_user_preference(
+        self, user_id: str, key: str, value: str
+    ) -> bool:
+        """Save a single user preference to Honcho peer metadata."""
+        if key not in self.DEFAULT_PREFERENCES:
+            logger.warning(f"Unknown preference key: {key}")
+            return False
+
+        try:
+            peer = self._honcho.peer(user_id)
+            # Merge with existing metadata
+            meta = dict(peer.metadata or {})
+            meta[key] = value
+            peer.set_metadata(meta)
+            logger.info(f"Saved preference {user_id}.{key} = {value}")
+            return True
+        except Exception as e:
+            logger.warning(f"Failed to save preference: {e}")
+            return False
+
 
 # Singleton instance for the app
 kira_memory = KiraMemory()

frontend/src/App.tsx

@@ -1,4 +1,4 @@
-import { useState } from 'react';
+import { useState, useEffect } from 'react';
 import Clock from './components/Clock';
 import BackgroundScene from './components/BackgroundScene';
 import MusicPlayer from './components/MusicPlayer';
@@ -10,17 +10,54 @@ import PetZone from './components/PetZone';
 import Wardrobe from './components/Wardrobe';
 import Toolbar from './components/Toolbar';
 import Particles from './components/Particles';
+import WelcomeScreen from './components/WelcomeScreen';
 import { SCENES, type Scene } from './components/scenes';
 import { useConversation } from './hooks/useConversation';
 
 export default function App() {
+  const {
+    messages,
+    isConnected,
+    isKiraSpeaking,
+    isRecording,
+    identified,
+    preferences,
+    loadingPrefs,
+    identify,
+    setPreference,
+    sendText,
+    startRecording,
+    stopRecording,
+  } = useConversation();
+
   const [currentSceneId, setCurrentSceneId] = useState('cozy-room');
   const [currentOutfit, setCurrentOutfit] = useState('cozy-hoodie');
   const [currentAcc, setCurrentAcc] = useState<string | null>(null);
   const [textInput, setTextInput] = useState('');
 
+  // Apply saved preferences once they load
+  useEffect(() => {
+    if (preferences.scene) setCurrentSceneId(preferences.scene);
+    if (preferences.outfit) setCurrentOutfit(preferences.outfit);
+    if (preferences.accessory) setCurrentAcc(preferences.accessory);
+  }, [preferences.scene, preferences.outfit, preferences.accessory]);
+
   const currentScene: Scene = SCENES.find((s) => s.id === currentSceneId) ?? SCENES[0];
-  const { messages, isConnected, isKiraSpeaking, isRecording, sendText, startRecording, stopRecording } = useConversation();
+
+  const handleSceneChange = (id: string) => {
+    setCurrentSceneId(id);
+    setPreference('scene', id);
+  };
+
+  const handleOutfitChange = (id: string) => {
+    setCurrentOutfit(id);
+    setPreference('outfit', id);
+  };
+
+  const handleAccessoryChange = (id: string | null) => {
+    setCurrentAcc(id);
+    setPreference('accessory', id || '');
+  };
 
   const handleTalkToggle = () => {
     if (isRecording) stopRecording();
@@ -33,6 +70,42 @@ export default function App() {
     setTextInput('');
   };
 
+  const handleWelcome = (name: string) => {
+    identify(name);
+  };
+
+  // Show WelcomeScreen for first-time users
+  if (!identified && !loadingPrefs) {
+    const savedId = localStorage.getItem('kira-user-id');
+    if (!savedId) {
+      return <WelcomeScreen onComplete={handleWelcome} />;
+    }
+    // Has saved ID but not identified yet — show welcome with their name
+    return (
+      <div className="min-h-screen flex items-center justify-center bg-gradient-to-br from-kira-bg via-kira-glow to-kira-pink/20 p-4">
+        <div className="glass-card max-w-sm w-full p-6 text-center space-y-4">
+          <div className="w-16 h-16 rounded-full bg-gradient-to-br from-kira-pink via-kira-lav to-kira-mint p-1 mx-auto animate-pulse-glow">
+            <div className="w-full h-full rounded-full bg-white flex items-center justify-center">
+              <span className="text-2xl">🌸</span>
+            </div>
+          </div>
+          <p className="text-kira-plum/60 text-sm">coming back? say your name to pick up where you left off</p>
+          <WelcomeScreen onComplete={handleWelcome} />
+        </div>
+        <style>{`
+          @keyframes pulse-glow {
+            0%, 100% { box-shadow: 0 0 12px rgba(255,182,193,0.3); }
+            50% { box-shadow: 0 0 28px rgba(216,180,254,0.5); }
+          }
+          .animate-pulse-glow { animation: pulse-glow 3s ease-in-out infinite; }
+        `}</style>
+      </div>
+    );
+  }
+
+  // Main app with user's name in the greeting
+  const userName = preferences.name || 'there';
+
   return (
     <div
       className="min-h-screen relative transition-all duration-1000"
@@ -43,10 +116,10 @@ export default function App() {
       <div className="relative z-20 h-screen flex flex-col">
         {/* Top toolbar */}
         <div className="px-4 pt-4">
-          <Toolbar currentScene={currentSceneId} onSceneChange={setCurrentSceneId} />
+          <Toolbar currentScene={currentSceneId} onSceneChange={handleSceneChange} />
         </div>
 
-        {/* Main grid — scrollable center */}
+        {/* Main grid */}
         <div className="flex-1 overflow-y-auto px-4 py-4 scrollbar-thin">
           <div className="grid grid-cols-1 md:grid-cols-2 lg:grid-cols-3 xl:grid-cols-4 gap-4 max-w-7xl mx-auto">
 
@@ -70,26 +143,26 @@ export default function App() {
 
             {/* Column 3: Chat + Text Input */}
             <div className="space-y-4">
-              <ChatBubble messages={messages} isKiraSpeaking={isKiraSpeaking} />
+              <ChatBubble messages={messages} isKiraSpeaking={isKiraSpeaking} userName={userName} />
 
               {/* Text input fallback */}
               <div className="glass-card p-3">
+                {/* Subtle greeting */}
+                <div className="text-[10px] text-kira-plum/30 mb-2">
+                  hey {userName} ✨
+                </div>
                 <div className="flex gap-2">
                   <input
                     value={textInput}
                     onChange={(e) => setTextInput(e.target.value)}
                     onKeyDown={(e) => e.key === 'Enter' && handleTextSend()}
-                    placeholder="type a message..."
+                    placeholder={`what's up, ${userName}?`}
                     className="flex-1 bg-white/60 rounded-xl px-3 py-2 text-sm text-kira-plum placeholder-kira-plum/30 border border-kira-pink/20 focus:outline-none focus:border-kira-pink/50"
                   />
-                  <button
+                  <button onClick={handleTextSend} className="btn-kira px-3 text-sm">
-                    onClick={handleTextSend}
-                    className="btn-kira px-3 text-sm"
-                  >
                     Send
                   </button>
                 </div>
-                {/* Connection indicator */}
                 <div className="flex items-center gap-2 mt-2 text-[10px] text-kira-plum/30">
                   <span className={`w-1.5 h-1.5 rounded-full ${isConnected ? 'bg-kira-mint' : 'bg-red-300'}`} />
                   {isConnected ? 'connected' : 'connecting...'}
@@ -100,7 +173,7 @@ export default function App() {
             {/* Column 4: Cats + Wardrobe */}
             <div className="space-y-4">
               <PetZone />
-              <Wardrobe onOutfitChange={setCurrentOutfit} onAccessoryChange={setCurrentAcc} />
+              <Wardrobe onOutfitChange={handleOutfitChange} onAccessoryChange={handleAccessoryChange} />
             </div>
 
           </div>
@@ -111,10 +184,10 @@ export default function App() {
           <div className="glass-card px-4 py-2 flex items-center justify-between text-xs text-kira-plum/40">
             <div className="flex items-center gap-3">
               <span className={`w-2 h-2 rounded-full ${isRecording ? 'bg-red-400 animate-pulse' : isKiraSpeaking ? 'bg-kira-pink animate-pulse' : 'bg-kira-mint'} inline-block`} />
-              <span>{isRecording ? 'listening...' : isKiraSpeaking ? 'kira speaking' : 'kira is here'}</span>
+              <span>{isRecording ? 'listening...' : isKiraSpeaking ? 'kira speaking' : `kira's here for you, ${userName}`}</span>
             </div>
             <div className="flex items-center gap-3">
-              <span className="hidden sm:inline">{isConnected ? 'body double mode' : 'offline'}</span>
+              <span className="hidden sm:inline">{identified ? `hi ${userName}` : 'body double mode'}</span>
               <span>🌸</span>
             </div>
           </div>

frontend/src/components/ChatBubble.tsx

@@ -4,11 +4,13 @@ interface Message {
   id: string;
   role: 'user' | 'kira';
   text: string;
+  timestamp: number;
 }
 
 interface Props {
   messages: Message[];
   isKiraSpeaking: boolean;
+  userName?: string;
 }
 
 export default function ChatBubble({ messages, isKiraSpeaking }: Props) {

frontend/src/components/WelcomeScreen.tsx

@@ -0,0 +1,75 @@
+import { useState } from 'react';
+
+interface Props {
+  onComplete: (name: string) => void;
+}
+
+export default function WelcomeScreen({ onComplete }: Props) {
+  const [name, setName] = useState('');
+
+  const handleSubmit = (e: React.FormEvent) => {
+    e.preventDefault();
+    const trimmed = name.trim();
+    if (trimmed) onComplete(trimmed);
+  };
+
+  return (
+    <div className="min-h-screen flex items-center justify-center bg-gradient-to-br from-kira-bg via-kira-glow to-kira-pink/20 p-4">
+      <div className="glass-card max-w-md w-full p-8 text-center space-y-6">
+        {/* Avatar preview */}
+        <div className="flex justify-center">
+          <div className="w-20 h-20 rounded-full bg-gradient-to-br from-kira-pink via-kira-lav to-kira-mint p-1 animate-pulse-glow">
+            <div className="w-full h-full rounded-full bg-white flex items-center justify-center">
+              <span className="text-3xl">🌸</span>
+            </div>
+          </div>
+        </div>
+
+        <div className="space-y-2">
+          <h1 className="text-2xl font-extrabold text-kira-plum">Welcome to Kira!</h1>
+          <p className="text-sm text-kira-violet/60 leading-relaxed">
+            I'm your focus bestie. I'm here to body-double with you — 
+            keep you company, cheer you on, and help you get things done. 
+            Lo-fi beats, timers, cats, the whole deal.
+          </p>
+        </div>
+
+        <form onSubmit={handleSubmit} className="space-y-4">
+          <div>
+            <label className="block text-sm font-semibold text-kira-plum/70 mb-1.5 text-left">
+              What should I call you?
+            </label>
+            <input
+              value={name}
+              onChange={(e) => setName(e.target.value)}
+              placeholder="Your name..."
+              className="w-full bg-white/60 rounded-xl px-4 py-3 text-sm text-kira-plum placeholder-kira-plum/30 border border-kira-pink/20 focus:outline-none focus:border-kira-pink/50 transition-all"
+              autoFocus
+              maxLength={30}
+            />
+          </div>
+
+          <button
+            type="submit"
+            disabled={!name.trim()}
+            className="btn-kira w-full py-3 text-base disabled:opacity-40 disabled:cursor-not-allowed transition-all"
+          >
+            Let's Go! ✨
+          </button>
+        </form>
+
+        <p className="text-[10px] text-kira-plum/30">
+          everything stays between us — your name, your preferences, your focus vibes
+        </p>
+      </div>
+
+      <style>{`
+        @keyframes pulse-glow {
+          0%, 100% { box-shadow: 0 0 12px rgba(255,182,193,0.3); }
+          50% { box-shadow: 0 0 28px rgba(216,180,254,0.5); }
+        }
+        .animate-pulse-glow { animation: pulse-glow 3s ease-in-out infinite; }
+      `}</style>
+    </div>
+  );
+}

frontend/src/hooks/useConversation.ts

@@ -1,5 +1,12 @@
 import { useState, useCallback, useRef, useEffect } from 'react';
 
+export interface UserPreferences {
+  name: string;
+  scene: string;
+  outfit: string;
+  accessory: string;
+}
+
 interface Message {
   id: string;
   role: 'user' | 'kira';
@@ -8,25 +15,54 @@ interface Message {
 }
 
 const WS_URL = `${location.protocol === 'https:' ? 'wss:' : 'ws:'}//${location.host}/api/ws`;
+const USER_ID_KEY = 'kira-user-id';
+
+function loadUserId(): string {
+  return localStorage.getItem(USER_ID_KEY) || '';
+}
+
+function saveUserId(id: string) {
+  localStorage.setItem(USER_ID_KEY, id);
+}
 
 export function useConversation() {
   const [messages, setMessages] = useState<Message[]>([]);
   const [isConnected, setIsConnected] = useState(false);
   const [isKiraSpeaking, setIsKiraSpeaking] = useState(false);
+  const [isRecording, setIsRecording] = useState(false);
+  const [identified, setIdentified] = useState(false);
+  const [preferences, setPreferences] = useState<UserPreferences>({
+    name: '',
+    scene: 'cozy-room',
+    outfit: 'cozy-hoodie',
+    accessory: '',
+  });
+  const [loadingPrefs, setLoadingPrefs] = useState(true);
+
   const wsRef = useRef<WebSocket | null>(null);
   const audioRef = useRef<HTMLAudioElement | null>(null);
   const recorderRef = useRef<MediaRecorder | null>(null);
   const streamRef = useRef<MediaStream | null>(null);
-  const [isRecording, setIsRecording] = useState(false);
 
   // Connect WebSocket
   const connect = useCallback(() => {
     if (wsRef.current?.readyState === WebSocket.OPEN) return;
+    setLoadingPrefs(true);
 
     const ws = new WebSocket(WS_URL);
     wsRef.current = ws;
 
-    ws.onopen = () => setIsConnected(true);
+    ws.onopen = () => {
+      setIsConnected(true);
+      // Auto-identify if returning user
+      const savedId = loadUserId();
+      if (savedId) {
+        ws.send(JSON.stringify({ type: 'identify', user_id: savedId }));
+      } else {
+        setLoadingPrefs(false);
+      }
+    };
+
     ws.onclose = () => {
       setIsConnected(false);
       setTimeout(connect, 3000);
@@ -51,6 +87,25 @@ export function useConversation() {
   // Handle incoming WS messages
   const handleMessage = useCallback((msg: any) => {
     switch (msg.type) {
+      case 'identified': {
+        setIdentified(true);
+        setLoadingPrefs(false);
+        if (msg.user_id) saveUserId(msg.user_id);
+        if (msg.preferences) {
+          setPreferences({
+            name: msg.preferences.name || '',
+            scene: msg.preferences.scene || 'cozy-room',
+            outfit: msg.preferences.outfit || 'cozy-hoodie',
+            accessory: msg.preferences.accessory || '',
+          });
+        }
+        break;
+      }
+
+      case 'preference_saved':
+        // Already optimistically updated locally
+        break;
+
       case 'transcript':
         addMessage('user', msg.text);
         break;
@@ -91,15 +146,52 @@ export function useConversation() {
     ]);
   }, []);
 
-  // Send text directly (no microphone)
+  // ── Identity ──
-  const sendText = useCallback((text: string) => {
+
-    if (!text.trim()) return;
+  const identify = useCallback((name: string) => {
+    const userId = `kira-${name.toLowerCase().replace(/[^a-z0-9]/g, '-')}`;
+    saveUserId(userId);
+    setPreferences((p) => ({ ...p, name }));
+
     if (wsRef.current?.readyState === WebSocket.OPEN) {
-      wsRef.current.send(JSON.stringify({ type: 'conversation_text', text: text.trim() }));
+      wsRef.current.send(JSON.stringify({
+        type: 'identify',
+        user_id: userId,
+        name,
+      }));
     }
   }, []);
 
-  // Push-to-talk: start recording
+  // ── Preferences ──
+
+  const setPreference = useCallback((key: string, value: string) => {
+    // Optimistic update
+    setPreferences((p) => ({ ...p, [key]: value }));
+
+    // Sync to backend
+    if (wsRef.current?.readyState === WebSocket.OPEN && identified) {
+      wsRef.current.send(JSON.stringify({
+        type: 'set_preference',
+        key,
+        value,
+      }));
+    }
+  }, [identified]);
+
+  // ── Text ──
+
+  const sendText = useCallback((text: string) => {
+    if (!text.trim()) return;
+    if (wsRef.current?.readyState === WebSocket.OPEN) {
+      wsRef.current.send(JSON.stringify({
+        type: 'conversation_text',
+        text: text.trim(),
+      }));
+    }
+  }, []);
+
+  // ── Audio ──
+
   const startRecording = useCallback(async () => {
     try {
       const stream = await navigator.mediaDevices.getUserMedia({ audio: true });
@@ -117,16 +209,12 @@ export function useConversation() {
       };
 
       recorder.onstop = () => {
-        // Send all audio chunks as one blob
         const blob = new Blob(chunks, { type: 'audio/webm' });
         const reader = new FileReader();
         reader.onload = () => {
           const base64 = (reader.result as string).split(',')[1];
           if (wsRef.current?.readyState === WebSocket.OPEN) {
-            wsRef.current.send(JSON.stringify({
+            wsRef.current.send(JSON.stringify({ type: 'audio_chunk', data: base64 }));
-              type: 'audio_chunk',
-              data: base64,
-            }));
             wsRef.current.send(JSON.stringify({ type: 'transcribe' }));
           }
         };
@@ -144,7 +232,6 @@ export function useConversation() {
     }
   }, []);
 
-  // Push-to-talk: stop recording
   const stopRecording = useCallback(() => {
     recorderRef.current?.stop();
   }, []);
@@ -163,6 +250,11 @@ export function useConversation() {
     isConnected,
     isKiraSpeaking,
     isRecording,
+    identified,
+    preferences,
+    loadingPrefs,
+    identify,
+    setPreference,
     sendText,
     startRecording,
     stopRecording,