Private-AI

Running

App Files Files Community

seawolf2357 commited on 5 days ago

Commit

c90fc65

verified ·

1 Parent(s): a9741e7

Update app-backup4.py

Browse files

Files changed (1) hide show

app-backup4.py +329 -329

app-backup4.py CHANGED Viewed

@@ -23,43 +23,15 @@ import gradio as gr
 import io
 from scipy import signal
 import wave
 load_dotenv()
 SAMPLE_RATE = 24000
-# Supported languages for OpenAI Realtime API
-SUPPORTED_LANGUAGES = {
-    "ko": "한국어 (Korean)",
-    "en": "English",
-    "es": "Español (Spanish)",
-    "fr": "Français (French)",
-    "de": "Deutsch (German)",
-    "it": "Italiano (Italian)",
-    "pt": "Português (Portuguese)",
-    "ru": "Русский (Russian)",
-    "ja": "日本語 (Japanese)",
-    "zh": "中文 (Chinese)",
-    "ar": "العربية (Arabic)",
-    "hi": "हिन्दी (Hindi)",
-    "nl": "Nederlands (Dutch)",
-    "pl": "Polski (Polish)",
-    "tr": "Türkçe (Turkish)",
-    "vi": "Tiếng Việt (Vietnamese)",
-    "th": "ไทย (Thai)",
-    "id": "Bahasa Indonesia",
-    "sv": "Svenska (Swedish)",
-    "da": "Dansk (Danish)",
-    "no": "Norsk (Norwegian)",
-    "fi": "Suomi (Finnish)",
-    "he": "עברית (Hebrew)",
-    "uk": "Українська (Ukrainian)",
-    "cs": "Čeština (Czech)",
-    "el": "Ελληνικά (Greek)",
-    "ro": "Română (Romanian)",
-    "hu": "Magyar (Hungarian)",
-    "ms": "Bahasa Melayu (Malay)"
-}
 # HTML content embedded as a string
 HTML_CONTENT = """<!DOCTYPE html>
@@ -199,21 +171,37 @@ HTML_CONTENT = """<!DOCTYPE html>
         .toggle-switch.active .toggle-slider {
             transform: translateX(24px);
         }
-        /* Select dropdown */
-        select {
             background-color: var(--card-bg);
-            color: var(--text-color);
             border: 1px solid var(--border-color);
-            padding: 8px 12px;
             border-radius: 6px;
-            font-size: 14px;
             cursor: pointer;
-            min-width: 120px;
-            max-width: 200px;
         }
-        select:focus {
-            outline: none;
-            border-color: var(--primary-color);
         }
         /* Text inputs */
         .text-input-section {
@@ -519,7 +507,7 @@ HTML_CONTENT = """<!DOCTYPE html>
         <div class="main-content">
             <div class="sidebar">
                 <div class="settings-section">
-                    <h3 style="margin: 0 0 15px 0; color: var(--primary-color);">설정(텍스트 채팅에만 적용)</h3>
                     <div class="settings-grid">
                         <div class="setting-item">
                             <span class="setting-label">웹 검색</span>
@@ -527,41 +515,6 @@ HTML_CONTENT = """<!DOCTYPE html>
                                 <div class="toggle-slider"></div>
                             </div>
                         </div>
-                        <div class="setting-item">
-                            <span class="setting-label">다국어 번역 채팅</span>
-                            <select id="language-select">
-                                <option value="">비활성화</option>
-                                <option value="ko">한국어 (Korean)</option>
-                                <option value="en">English</option>
-                                <option value="es">Español (Spanish)</option>
-                                <option value="fr">Français (French)</option>
-                                <option value="de">Deutsch (German)</option>
-                                <option value="it">Italiano (Italian)</option>
-                                <option value="pt">Português (Portuguese)</option>
-                                <option value="ru">Русский (Russian)</option>
-                                <option value="ja">日本語 (Japanese)</option>
-                                <option value="zh">中文 (Chinese)</option>
-                                <option value="ar">العربية (Arabic)</option>
-                                <option value="hi">हिन्दी (Hindi)</option>
-                                <option value="nl">Nederlands (Dutch)</option>
-                                <option value="pl">Polski (Polish)</option>
-                                <option value="tr">Türkçe (Turkish)</option>
-                                <option value="vi">Tiếng Việt (Vietnamese)</option>
-                                <option value="th">ไทย (Thai)</option>
-                                <option value="id">Bahasa Indonesia</option>
-                                <option value="sv">Svenska (Swedish)</option>
-                                <option value="da">Dansk (Danish)</option>
-                                <option value="no">Norsk (Norwegian)</option>
-                                <option value="fi">Suomi (Finnish)</option>
-                                <option value="he">עברית (Hebrew)</option>
-                                <option value="uk">Українська (Ukrainian)</option>
-                                <option value="cs">Čeština (Czech)</option>
-                                <option value="el">Ελληνικά (Greek)</option>
-                                <option value="ro">Română (Romanian)</option>
-                                <option value="hu">Magyar (Hungarian)</option>
-                                <option value="ms">Bahasa Melayu (Malay)</option>
-                            </select>
-                        </div>
                     </div>
                     <div class="text-input-section">
                         <label for="system-prompt" class="setting-label">시스템 프롬프트:</label>
@@ -569,6 +522,11 @@ HTML_CONTENT = """<!DOCTYPE html>
                     </div>
                 </div>
                 <div class="controls">
                     <button id="start-button">대화 시작</button>
                 </div>
@@ -594,8 +552,8 @@ HTML_CONTENT = """<!DOCTYPE html>
         let peerConnection;
         let webrtc_id;
         let webSearchEnabled = false;
-        let selectedLanguage = "";
         let systemPrompt = "You are a helpful assistant. Respond in a friendly and professional manner.";
         const audioOutput = document.getElementById('audio-output');
         const startButton = document.getElementById('start-button');
         const sendButton = document.getElementById('send-button');
@@ -603,15 +561,61 @@ HTML_CONTENT = """<!DOCTYPE html>
         const statusDot = document.getElementById('status-dot');
         const statusText = document.getElementById('status-text');
         const searchToggle = document.getElementById('search-toggle');
-        const languageSelect = document.getElementById('language-select');
         const systemPromptInput = document.getElementById('system-prompt');
         const textInput = document.getElementById('text-input');
         let audioLevel = 0;
         let animationFrame;
         let audioContext, analyser, audioSource;
         let dataChannel = null;
         let isVoiceActive = false;
         // Web search toggle functionality
         searchToggle.addEventListener('click', () => {
             webSearchEnabled = !webSearchEnabled;
@@ -619,12 +623,6 @@ HTML_CONTENT = """<!DOCTYPE html>
             console.log('Web search enabled:', webSearchEnabled);
         });
-        // Language selection
-        languageSelect.addEventListener('change', () => {
-            selectedLanguage = languageSelect.value;
-            console.log('Selected language:', selectedLanguage);
-        });
         // System prompt update
         systemPromptInput.addEventListener('input', () => {
             systemPrompt = systemPromptInput.value || "You are a helpful assistant. Respond in a friendly and professional manner.";
@@ -664,8 +662,8 @@ HTML_CONTENT = """<!DOCTYPE html>
                     body: JSON.stringify({
                         message: message,
                         web_search_enabled: webSearchEnabled,
-                        target_language: selectedLanguage,
-                        system_prompt: systemPrompt
                     })
                 });
@@ -680,8 +678,8 @@ HTML_CONTENT = """<!DOCTYPE html>
                 } else {
                     // Add assistant response
                     let content = data.response;
-                    if (selectedLanguage && data.language) {
-                        content += ` <span class="language-info">[${data.language}]</span>`;
                     }
                     addMessage('assistant', content);
                 }
@@ -849,8 +847,8 @@ HTML_CONTENT = """<!DOCTYPE html>
                 console.log('Sending offer with settings:', {
                     webrtc_id: webrtc_id,
                     web_search_enabled: webSearchEnabled,
-                    target_language: selectedLanguage,
-                    system_prompt: systemPrompt
                 });
                 const response = await fetch('/webrtc/offer', {
@@ -861,8 +859,8 @@ HTML_CONTENT = """<!DOCTYPE html>
                         type: peerConnection.localDescription.type,
                         webrtc_id: webrtc_id,
                         web_search_enabled: webSearchEnabled,
-                        target_language: selectedLanguage,
-                        system_prompt: systemPrompt
                     })
                 });
                 const serverResponse = await response.json();
@@ -879,8 +877,8 @@ HTML_CONTENT = """<!DOCTYPE html>
                     const eventJson = JSON.parse(event.data);
                     let content = eventJson.content;
-                    if (selectedLanguage && eventJson.language) {
-                        content += ` <span class="language-info">[${eventJson.language}]</span>`;
                     }
                     addMessage("assistant", content);
                 });
@@ -897,7 +895,7 @@ HTML_CONTENT = """<!DOCTYPE html>
                 stop();
             }
         }
-        function addMessage(role, content) {
             const messageDiv = document.createElement('div');
             messageDiv.classList.add('message', role);
@@ -908,6 +906,19 @@ HTML_CONTENT = """<!DOCTYPE html>
             }
             chatMessages.appendChild(messageDiv);
             chatMessages.scrollTop = chatMessages.scrollHeight;
         }
         function stop() {
@@ -1000,6 +1011,8 @@ HTML_CONTENT = """<!DOCTYPE html>
         // Initialize on page load
         window.addEventListener('DOMContentLoaded', () => {
             sendButton.style.display = 'block';
         });
     </script>
 </body>
@@ -1048,6 +1061,131 @@ class BraveSearchClient:
                 return []
 # Initialize search client globally
 brave_api_key = os.getenv("BSEARCH_API")
 search_client = BraveSearchClient(brave_api_key) if brave_api_key else None
@@ -1059,52 +1197,17 @@ connection_settings = {}
 # Initialize OpenAI client for text chat
 client = openai.AsyncOpenAI()
-def get_translation_instructions(target_language: str) -> str:
-    """Get instructions for translation based on target language"""
-    if not target_language:
-        return ""
-    language_name = SUPPORTED_LANGUAGES.get(target_language, target_language)
-    return (
-        f"\n\nIMPORTANT: You must respond in {language_name} ({target_language}). "
-        f"Translate all your responses to {language_name}."
-    )
 def update_chatbot(chatbot: list[dict], response: ResponseAudioTranscriptDoneEvent):
     chatbot.append({"role": "assistant", "content": response.transcript})
     return chatbot
-async def process_text_chat(message: str, web_search_enabled: bool, target_language: str,
-                          system_prompt: str) -> Dict[str, str]:
     """Process text chat using GPT-4o-mini model"""
     try:
-        # If target language is set, override system prompt completely
-        if target_language:
-            language_name = SUPPORTED_LANGUAGES.get(target_language, target_language)
-            # Create system prompt in target language
-            if target_language == "en":
-                base_instructions = f"You are a helpful assistant. You speak ONLY English. Never use Korean or any other language. {system_prompt}"
-                user_prefix = "Please respond in English: "
-            elif target_language == "ja":
-                base_instructions = f"あなたは親切なアシスタントです。日本語のみを話します。韓国語や他の言語は絶対に使用しません。{system_prompt}"
-                user_prefix = "日本語で答えてください: "
-            elif target_language == "zh":
-                base_instructions = f"你是一个乐于助人的助手。你只说中文。绝不使用韩语或其他语言。{system_prompt}"
-                user_prefix = "请用中文回答: "
-            elif target_language == "es":
-                base_instructions = f"Eres un asistente útil. Solo hablas español. Nunca uses coreano u otros idiomas. {system_prompt}"
-                user_prefix = "Por favor responde en español: "
-            else:
-                base_instructions = f"You are a helpful assistant that speaks ONLY {language_name}. {system_prompt}"
-                user_prefix = f"Please respond in {language_name}: "
-        else:
-            base_instructions = system_prompt or "You are a helpful assistant."
-            user_prefix = ""
         messages = [
-            {"role": "system", "content": base_instructions}
         ]
         # Handle web search if enabled
@@ -1124,19 +1227,12 @@ async def process_text_chat(message: str, web_search_enabled: bool, target_langu
                     for i, result in enumerate(search_results[:5], 1):
                         search_context += f"{i}. {result['title']}\n{result['description']}\n\n"
-                    # Add search context in target language if set
-                    if target_language:
-                        search_instruction = f"Use this search information but respond in {SUPPORTED_LANGUAGES.get(target_language, target_language)} only: "
-                    else:
-                        search_instruction = "다음 웹 검색 결과를 참고하여 답변하세요: "
                     messages.append({
                         "role": "system",
-                        "content": search_instruction + "\n\n" + search_context
                     })
-        # Add user message with language prefix
-        messages.append({"role": "user", "content": user_prefix + message})
         # Call GPT-4o-mini
         response = await client.chat.completions.create(
@@ -1148,27 +1244,22 @@ async def process_text_chat(message: str, web_search_enabled: bool, target_langu
         response_text = response.choices[0].message.content
-        # Final check - remove any Korean if target language is not Korean
-        if target_language and target_language != "ko":
-            import re
-            if re.search(r'[가-힣]', response_text):
-                print(f"[TEXT CHAT] WARNING: Korean detected in response for {target_language}")
-                # Try again with stronger prompt
-                messages[-1] = {"role": "user", "content": f"ONLY {SUPPORTED_LANGUAGES.get(target_language, target_language)}, NO KOREAN: {message}"}
-                retry_response = await client.chat.completions.create(
-                    model="gpt-4.1-mini",
-                    messages=messages,
-                    temperature=0.3,
-                    max_tokens=2000
-                )
-                response_text = retry_response.choices[0].message.content
-        print(f"[TEXT CHAT] Target language: {target_language}")
-        print(f"[TEXT CHAT] Response preview: {response_text[:100]}...")
         return {
             "response": response_text,
-            "language": SUPPORTED_LANGUAGES.get(target_language, "") if target_language else ""
         }
     except Exception as e:
@@ -1177,8 +1268,8 @@ async def process_text_chat(message: str, web_search_enabled: bool, target_langu
 class OpenAIHandler(AsyncStreamHandler):
-    def __init__(self, web_search_enabled: bool = False, target_language: str = "",
-                 system_prompt: str = "", webrtc_id: str = None) -> None:
         super().__init__(
             expected_layout="mono",
             output_sample_rate=SAMPLE_RATE,
@@ -1193,11 +1284,10 @@ class OpenAIHandler(AsyncStreamHandler):
         self.current_call_id = None
         self.webrtc_id = webrtc_id
         self.web_search_enabled = web_search_enabled
-        self.target_language = target_language
         self.system_prompt = system_prompt
-        print(f"[INIT] Handler created with web_search={web_search_enabled}, "
-              f"target_language={target_language}")
     def copy(self):
         # Get the most recent settings
@@ -1215,9 +1305,9 @@ class OpenAIHandler(AsyncStreamHandler):
                 return OpenAIHandler(
                     web_search_enabled=settings.get('web_search_enabled', False),
-                    target_language=settings.get('target_language', ''),
                     system_prompt=settings.get('system_prompt', ''),
-                    webrtc_id=recent_id
                 )
         print(f"[COPY] No settings found, creating default handler")
@@ -1256,18 +1346,6 @@ class OpenAIHandler(AsyncStreamHandler):
             )
             await self.connection.response.create()
-    def get_translation_instructions(self):
-        """Get instructions for translation based on target language"""
-        if not self.target_language:
-            return ""
-        language_name = SUPPORTED_LANGUAGES.get(self.target_language, self.target_language)
-        return (
-            f"\n\nIMPORTANT: You must respond in {language_name} ({self.target_language}). "
-            f"Translate all your responses to {language_name}. "
-            f"This includes both spoken and written responses."
-        )
     async def start_up(self):
         """Connect to realtime API"""
         # First check if we have the most recent settings
@@ -1275,108 +1353,20 @@ class OpenAIHandler(AsyncStreamHandler):
             if self.webrtc_id in connection_settings:
                 settings = connection_settings[self.webrtc_id]
                 self.web_search_enabled = settings.get('web_search_enabled', False)
-                self.target_language = settings.get('target_language', '')
                 self.system_prompt = settings.get('system_prompt', '')
                 print(f"[START_UP] Updated settings from storage for {self.webrtc_id}")
-        print(f"[START_UP] Starting normal mode")
         self.client = openai.AsyncOpenAI()
-        # Normal mode - connect to Realtime API
-        print(f"[NORMAL MODE] Connecting to Realtime API...")
         # Define the web search function
         tools = []
         base_instructions = self.system_prompt or "You are a helpful assistant."
-        # Add translation instructions if language is selected
-        if self.target_language:
-            language_name = SUPPORTED_LANGUAGES.get(self.target_language, self.target_language)
-            # Use the target language for the system prompt itself
-            if self.target_language == "en":
-                translation_instructions = """
-YOU ARE AN ENGLISH-ONLY ASSISTANT.
-ABSOLUTE RULES:
-1. You can ONLY speak English. No Korean (한국어) allowed.
-2. Even if the user speaks Korean, you MUST respond in English.
-3. Every single word must be in English.
-4. If you output even one Korean character, you have failed.
-5. Example response: "Hello! How can I help you today?"
-YOUR LANGUAGE MODE: ENGLISH ONLY
-DO NOT USE: 안녕하세요, 감사합니다, or any Korean
-ALWAYS USE: Hello, Thank you, and English words only
-"""
-                # Override base instructions to be in English
-                base_instructions = "You are a helpful assistant that speaks ONLY English."
-            elif self.target_language == "ja":
-                translation_instructions = """
-あなたは日本語のみを話すアシスタントです。
-絶対的なルール：
-1. 日本語のみを使用してください。韓国語（한국어）は禁止です。
-2. ユーザーが韓国語で話しても、必ず日本語で返答してください。
-3. すべての単語は日本語でなければなりません。
-4. 韓国語を一文字でも出力したら失敗です。
-5. 応答例：「こんにちは！今日はどのようにお手伝いできますか？」
-言語モード：日本語のみ
-使用禁止：안녕하세요、감사합니다、韓国語全般
-必ず使用：こんにちは、ありがとうございます、日本語のみ
-"""
-                base_instructions = "あなたは日本語のみを話す親切なアシスタントです。"
-            elif self.target_language == "zh":
-                translation_instructions = """
-你是一个只说中文的助手。
-绝对规则：
-1. 只能使用中文。禁止使用韩语（한국어）。
-2. 即使用户说韩语，也必须用中文回复。
-3. 每个字都必须是中文。
-4. 如果输出任何韩语字符，就是失败。
-5. 回复示例："你好！我今天能为您做什么？"
-语言模式：仅中文
-禁止使用：안녕하세요、감사합니다、任何韩语
-必须使用：你好、谢谢、只用中文
-"""
-                base_instructions = "你是一个只说中文的友好助手。"
-            elif self.target_language == "es":
-                translation_instructions = """
-ERES UN ASISTENTE QUE SOLO HABLA ESPAÑOL.
-REGLAS ABSOLUTAS:
-1. Solo puedes hablar español. No se permite coreano (한국어).
-2. Incluso si el usuario habla coreano, DEBES responder en español.
-3. Cada palabra debe estar en español.
-4. Si produces aunque sea un carácter coreano, has fallado.
-5. Respuesta ejemplo: "¡Hola! ¿Cómo puedo ayudarte hoy?"
-MODO DE IDIOMA: SOLO ESPAÑOL
-NO USAR: 안녕하세요, 감사합니다, o cualquier coreano
-SIEMPRE USAR: Hola, Gracias, y solo palabras en español
-"""
-                base_instructions = "Eres un asistente útil que habla SOLO español."
-            else:
-                translation_instructions = f"""
-YOU MUST ONLY SPEAK {language_name.upper()}.
-RULES:
-1. Output only in {language_name}
-2. Never use Korean
-3. Always respond in {language_name}
-"""
-                base_instructions = f"You are a helpful assistant that speaks ONLY {language_name}."
-        else:
-            translation_instructions = ""
         if self.web_search_enabled and self.search_client:
             tools = [{
                 "type": "function",
@@ -1411,23 +1401,9 @@ RULES:
                 "than to guess or use outdated information."
             )
-            # Combine all instructions
-            if translation_instructions:
-                # Translation instructions already include base_instructions
-                instructions = translation_instructions + search_instructions
-            else:
-                instructions = base_instructions + search_instructions
         else:
-            # No web search
-            if translation_instructions:
-                instructions = translation_instructions
-            else:
-                instructions = base_instructions
-        print(f"[NORMAL MODE] Base instructions: {base_instructions[:100]}...")
-        print(f"[NORMAL MODE] Translation instructions: {translation_instructions[:200] if translation_instructions else 'None'}...")
-        print(f"[NORMAL MODE] Combined instructions length: {len(instructions)}")
-        print(f"[NORMAL MODE] Target language: {self.target_language}")
         async with self.client.beta.realtime.connect(
             model="gpt-4o-mini-realtime-preview-2024-12-17"
@@ -1441,39 +1417,12 @@ RULES:
                 "temperature": 0.7,
                 "max_response_output_tokens": 4096,
                 "modalities": ["text", "audio"],
-                "voice": "alloy"  # Default voice
             }
-            # Use appropriate voice for the language
-            if self.target_language:
-                # Force language through multiple mechanisms
-                # 1. Use voice that's known to work well with the language
-                voice_map = {
-                    "en": "nova",      # Nova has clearer English
-                    "es": "nova",      # Nova works for Spanish
-                    "fr": "shimmer",   # Shimmer for French
-                    "de": "echo",      # Echo for German
-                    "ja": "alloy",     # Alloy can do Japanese
-                    "zh": "alloy",     # Alloy can do Chinese
-                    "ko": "nova",      # Nova for Korean
-                }
-                session_update["voice"] = voice_map.get(self.target_language, "nova")
-                # 2. Add language to modalities (experimental)
-                session_update["modalities"] = ["text", "audio"]
-                # 3. Set output format
-                session_update["output_audio_format"] = "pcm16"
-                # 4. Add language hint to the system (if supported by API)
-                if self.target_language in ["en", "es", "fr", "de", "ja", "zh"]:
-                    session_update["language"] = self.target_language  # Try setting language directly
-                print(f"[TRANSLATION MODE] Session update: {json.dumps(session_update, indent=2)}")
             await conn.session.update(session=session_update)
             self.connection = conn
-            print(f"Connected with tools: {len(tools)} functions, voice: {session_update.get('voice', 'default')}")
             async for event in self.connection:
                 # Debug logging for function calls
@@ -1482,11 +1431,22 @@ RULES:
                 if event.type == "response.audio_transcript.done":
                     print(f"[RESPONSE] Transcript: {event.transcript[:100]}...")
-                    print(f"[RESPONSE] Expected language: {self.target_language}")
                     output_data = {
                         "event": event,
-                        "language": SUPPORTED_LANGUAGES.get(self.target_language, "") if self.target_language else ""
                     }
                     await self.output_queue.put(AdditionalOutputs(output_data))
@@ -1547,9 +1507,8 @@ RULES:
                             self.current_call_id = None
     async def receive(self, frame: tuple[int, np.ndarray]) -> None:
-        # Normal mode - use Realtime API
         if not self.connection:
-            print(f"[RECEIVE] No connection in normal mode, skipping")
             return
         try:
             _, array = frame
@@ -1560,7 +1519,6 @@ RULES:
             print(f"Error in receive: {e}")
     async def emit(self) -> tuple[int, np.ndarray] | AdditionalOutputs | None:
-        # Normal mode
         item = await wait_for_item(self.output_queue)
         # Check if it's a dict with text message
@@ -1573,11 +1531,10 @@ RULES:
     async def shutdown(self) -> None:
         print(f"[SHUTDOWN] Called")
-        # Normal mode - close Realtime API connection
         if self.connection:
             await self.connection.close()
             self.connection = None
-            print("[NORMAL MODE] Connection closed")
 # Create initial handler instance
@@ -1604,6 +1561,12 @@ app = FastAPI()
 # Mount stream
 stream.mount(app)
 # Intercept offer to capture settings
 @app.post("/webrtc/offer", include_in_schema=False)
 async def custom_offer(request: Request):
@@ -1612,19 +1575,19 @@ async def custom_offer(request: Request):
     webrtc_id = body.get("webrtc_id")
     web_search_enabled = body.get("web_search_enabled", False)
-    target_language = body.get("target_language", "")
     system_prompt = body.get("system_prompt", "")
     print(f"[OFFER] Received offer with webrtc_id: {webrtc_id}")
     print(f"[OFFER] web_search_enabled: {web_search_enabled}")
-    print(f"[OFFER] target_language: {target_language}")
     # Store settings with timestamp
     if webrtc_id:
         connection_settings[webrtc_id] = {
             'web_search_enabled': web_search_enabled,
-            'target_language': target_language,
             'system_prompt': system_prompt,
             'timestamp': asyncio.get_event_loop().time()
         }
@@ -1651,6 +1614,43 @@ async def custom_offer(request: Request):
     return response
 @app.post("/chat/text")
 async def chat_text(request: Request):
     """Handle text chat messages using GPT-4o-mini"""
@@ -1658,14 +1658,14 @@ async def chat_text(request: Request):
         body = await request.json()
         message = body.get("message", "")
         web_search_enabled = body.get("web_search_enabled", False)
-        target_language = body.get("target_language", "")
         system_prompt = body.get("system_prompt", "")
         if not message:
             return {"error": "메시지가 비어있습니다."}
         # Process text chat
-        result = await process_text_chat(message, web_search_enabled, target_language, system_prompt)
         return result
@@ -1708,7 +1708,7 @@ async def outputs(webrtc_id: str):
                         data = {
                             "role": "assistant",
                             "content": event_data['event'].transcript,
-                            "language": event_data.get('language', '')
                         }
                         yield f"event: output\ndata: {json.dumps(data)}\n\n"

 import io
 from scipy import signal
 import wave
+import aiosqlite
+from langdetect import detect, LangDetectException
+from datetime import datetime
+import uuid
 load_dotenv()
 SAMPLE_RATE = 24000
+DB_PATH = "chat_history.db"
 # HTML content embedded as a string
 HTML_CONTENT = """<!DOCTYPE html>
         .toggle-switch.active .toggle-slider {
             transform: translateX(24px);
         }
+        /* History section */
+        .history-section {
             background-color: var(--card-bg);
+            border-radius: 12px;
+            padding: 20px;
             border: 1px solid var(--border-color);
+            margin-top: 20px;
+            max-height: 300px;
+            overflow-y: auto;
+        }
+        .history-item {
+            padding: 10px;
+            margin-bottom: 10px;
+            background-color: var(--dark-bg);
             border-radius: 6px;
             cursor: pointer;
+            transition: background-color 0.2s;
         }
+        .history-item:hover {
+            background-color: var(--hover-color);
+        }
+        .history-date {
+            font-size: 12px;
+            color: #888;
+        }
+        .history-preview {
+            font-size: 14px;
+            margin-top: 5px;
+            overflow: hidden;
+            text-overflow: ellipsis;
+            white-space: nowrap;
         }
         /* Text inputs */
         .text-input-section {
         <div class="main-content">
             <div class="sidebar">
                 <div class="settings-section">
+                    <h3 style="margin: 0 0 15px 0; color: var(--primary-color);">설정</h3>
                     <div class="settings-grid">
                         <div class="setting-item">
                             <span class="setting-label">웹 검색</span>
                                 <div class="toggle-slider"></div>
                             </div>
                         </div>
                     </div>
                     <div class="text-input-section">
                         <label for="system-prompt" class="setting-label">시스템 프롬프트:</label>
                     </div>
                 </div>
+                <div class="history-section">
+                    <h3 style="margin: 0 0 15px 0; color: var(--primary-color);">대화 기록</h3>
+                    <div id="history-list"></div>
+                </div>
                 <div class="controls">
                     <button id="start-button">대화 시작</button>
                 </div>
         let peerConnection;
         let webrtc_id;
         let webSearchEnabled = false;
         let systemPrompt = "You are a helpful assistant. Respond in a friendly and professional manner.";
+        let currentSessionId = null;
         const audioOutput = document.getElementById('audio-output');
         const startButton = document.getElementById('start-button');
         const sendButton = document.getElementById('send-button');
         const statusDot = document.getElementById('status-dot');
         const statusText = document.getElementById('status-text');
         const searchToggle = document.getElementById('search-toggle');
         const systemPromptInput = document.getElementById('system-prompt');
         const textInput = document.getElementById('text-input');
+        const historyList = document.getElementById('history-list');
         let audioLevel = 0;
         let animationFrame;
         let audioContext, analyser, audioSource;
         let dataChannel = null;
         let isVoiceActive = false;
+        // Start new session
+        async function startNewSession() {
+            const response = await fetch('/session/new', { method: 'POST' });
+            const data = await response.json();
+            currentSessionId = data.session_id;
+            console.log('New session started:', currentSessionId);
+            loadHistory();
+        }
+        // Load conversation history
+        async function loadHistory() {
+            try {
+                const response = await fetch('/history/recent');
+                const conversations = await response.json();
+                historyList.innerHTML = '';
+                conversations.forEach(conv => {
+                    const item = document.createElement('div');
+                    item.className = 'history-item';
+                    item.innerHTML = `
+                        <div class="history-date">${new Date(conv.created_at).toLocaleString()}</div>
+                        <div class="history-preview">${conv.summary || '대화 시작'}</div>
+                    `;
+                    item.onclick = () => loadConversation(conv.id);
+                    historyList.appendChild(item);
+                });
+            } catch (error) {
+                console.error('Failed to load history:', error);
+            }
+        }
+        // Load specific conversation
+        async function loadConversation(sessionId) {
+            try {
+                const response = await fetch(`/history/${sessionId}`);
+                const messages = await response.json();
+                chatMessages.innerHTML = '';
+                messages.forEach(msg => {
+                    addMessage(msg.role, msg.content, false);
+                });
+            } catch (error) {
+                console.error('Failed to load conversation:', error);
+            }
+        }
         // Web search toggle functionality
         searchToggle.addEventListener('click', () => {
             webSearchEnabled = !webSearchEnabled;
             console.log('Web search enabled:', webSearchEnabled);
         });
         // System prompt update
         systemPromptInput.addEventListener('input', () => {
             systemPrompt = systemPromptInput.value || "You are a helpful assistant. Respond in a friendly and professional manner.";
                     body: JSON.stringify({
                         message: message,
                         web_search_enabled: webSearchEnabled,
+                        system_prompt: systemPrompt,
+                        session_id: currentSessionId
                     })
                 });
                 } else {
                     // Add assistant response
                     let content = data.response;
+                    if (data.detected_language) {
+                        content += ` <span class="language-info">[${data.detected_language}]</span>`;
                     }
                     addMessage('assistant', content);
                 }
                 console.log('Sending offer with settings:', {
                     webrtc_id: webrtc_id,
                     web_search_enabled: webSearchEnabled,
+                    system_prompt: systemPrompt,
+                    session_id: currentSessionId
                 });
                 const response = await fetch('/webrtc/offer', {
                         type: peerConnection.localDescription.type,
                         webrtc_id: webrtc_id,
                         web_search_enabled: webSearchEnabled,
+                        system_prompt: systemPrompt,
+                        session_id: currentSessionId
                     })
                 });
                 const serverResponse = await response.json();
                     const eventJson = JSON.parse(event.data);
                     let content = eventJson.content;
+                    if (eventJson.detected_language) {
+                        content += ` <span class="language-info">[${eventJson.detected_language}]</span>`;
                     }
                     addMessage("assistant", content);
                 });
                 stop();
             }
         }
+        function addMessage(role, content, save = true) {
             const messageDiv = document.createElement('div');
             messageDiv.classList.add('message', role);
             }
             chatMessages.appendChild(messageDiv);
             chatMessages.scrollTop = chatMessages.scrollHeight;
+            // Save message to database if save flag is true
+            if (save && currentSessionId) {
+                fetch('/message/save', {
+                    method: 'POST',
+                    headers: { 'Content-Type': 'application/json' },
+                    body: JSON.stringify({
+                        session_id: currentSessionId,
+                        role: role,
+                        content: content
+                    })
+                }).catch(error => console.error('Failed to save message:', error));
+            }
         }
         function stop() {
         // Initialize on page load
         window.addEventListener('DOMContentLoaded', () => {
             sendButton.style.display = 'block';
+            startNewSession();
+            loadHistory();
         });
     </script>
 </body>
                 return []
+# Database helper class
+class ChatDatabase:
+    """Database manager for chat history"""
+    @staticmethod
+    async def init():
+        """Initialize database tables"""
+        async with aiosqlite.connect(DB_PATH) as db:
+            await db.execute("""
+                CREATE TABLE IF NOT EXISTS conversations (
+                    id TEXT PRIMARY KEY,
+                    created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
+                    updated_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
+                    summary TEXT
+                )
+            """)
+            await db.execute("""
+                CREATE TABLE IF NOT EXISTS messages (
+                    id INTEGER PRIMARY KEY AUTOINCREMENT,
+                    session_id TEXT NOT NULL,
+                    role TEXT NOT NULL,
+                    content TEXT NOT NULL,
+                    detected_language TEXT,
+                    timestamp TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
+                    FOREIGN KEY (session_id) REFERENCES conversations(id)
+                )
+            """)
+            await db.commit()
+    @staticmethod
+    async def create_session(session_id: str):
+        """Create a new conversation session"""
+        async with aiosqlite.connect(DB_PATH) as db:
+            await db.execute(
+                "INSERT INTO conversations (id) VALUES (?)",
+                (session_id,)
+            )
+            await db.commit()
+    @staticmethod
+    async def save_message(session_id: str, role: str, content: str):
+        """Save a message to the database"""
+        # Detect language
+        detected_language = None
+        try:
+            if content and len(content) > 10:  # Only detect for substantial content
+                detected_language = detect(content)
+        except LangDetectException:
+            pass
+        async with aiosqlite.connect(DB_PATH) as db:
+            await db.execute(
+                """INSERT INTO messages (session_id, role, content, detected_language)
+                   VALUES (?, ?, ?, ?)""",
+                (session_id, role, content, detected_language)
+            )
+            # Update conversation's updated_at timestamp
+            await db.execute(
+                "UPDATE conversations SET updated_at = CURRENT_TIMESTAMP WHERE id = ?",
+                (session_id,)
+            )
+            # Update conversation summary (use first user message as summary)
+            if role == "user":
+                cursor = await db.execute(
+                    "SELECT summary FROM conversations WHERE id = ?",
+                    (session_id,)
+                )
+                row = await cursor.fetchone()
+                if row and not row[0]:  # If no summary exists
+                    summary = content[:100] + "..." if len(content) > 100 else content
+                    await db.execute(
+                        "UPDATE conversations SET summary = ? WHERE id = ?",
+                        (summary, session_id)
+                    )
+            await db.commit()
+    @staticmethod
+    async def get_recent_conversations(limit: int = 10):
+        """Get recent conversations"""
+        async with aiosqlite.connect(DB_PATH) as db:
+            cursor = await db.execute(
+                """SELECT id, created_at, summary
+                   FROM conversations
+                   ORDER BY updated_at DESC
+                   LIMIT ?""",
+                (limit,)
+            )
+            rows = await cursor.fetchall()
+            return [
+                {
+                    "id": row[0],
+                    "created_at": row[1],
+                    "summary": row[2] or "새 대화"
+                }
+                for row in rows
+            ]
+    @staticmethod
+    async def get_conversation_messages(session_id: str):
+        """Get all messages for a conversation"""
+        async with aiosqlite.connect(DB_PATH) as db:
+            cursor = await db.execute(
+                """SELECT role, content, detected_language, timestamp
+                   FROM messages
+                   WHERE session_id = ?
+                   ORDER BY timestamp ASC""",
+                (session_id,)
+            )
+            rows = await cursor.fetchall()
+            return [
+                {
+                    "role": row[0],
+                    "content": row[1],
+                    "detected_language": row[2],
+                    "timestamp": row[3]
+                }
+                for row in rows
+            ]
 # Initialize search client globally
 brave_api_key = os.getenv("BSEARCH_API")
 search_client = BraveSearchClient(brave_api_key) if brave_api_key else None
 # Initialize OpenAI client for text chat
 client = openai.AsyncOpenAI()
 def update_chatbot(chatbot: list[dict], response: ResponseAudioTranscriptDoneEvent):
     chatbot.append({"role": "assistant", "content": response.transcript})
     return chatbot
+async def process_text_chat(message: str, web_search_enabled: bool, system_prompt: str, session_id: str) -> Dict[str, str]:
     """Process text chat using GPT-4o-mini model"""
     try:
         messages = [
+            {"role": "system", "content": system_prompt or "You are a helpful assistant."}
         ]
         # Handle web search if enabled
                     for i, result in enumerate(search_results[:5], 1):
                         search_context += f"{i}. {result['title']}\n{result['description']}\n\n"
                     messages.append({
                         "role": "system",
+                        "content": "다음 웹 검색 결과를 참고하여 답변하세요:\n\n" + search_context
                     })
+        messages.append({"role": "user", "content": message})
         # Call GPT-4o-mini
         response = await client.chat.completions.create(
         response_text = response.choices[0].message.content
+        # Detect language
+        detected_language = None
+        try:
+            if response_text and len(response_text) > 10:
+                detected_language = detect(response_text)
+        except:
+            pass
+        # Save messages to database
+        if session_id:
+            await ChatDatabase.save_message(session_id, "user", message)
+            await ChatDatabase.save_message(session_id, "assistant", response_text)
         return {
             "response": response_text,
+            "detected_language": detected_language
         }
     except Exception as e:
 class OpenAIHandler(AsyncStreamHandler):
+    def __init__(self, web_search_enabled: bool = False, system_prompt: str = "",
+                 webrtc_id: str = None, session_id: str = None) -> None:
         super().__init__(
             expected_layout="mono",
             output_sample_rate=SAMPLE_RATE,
         self.current_call_id = None
         self.webrtc_id = webrtc_id
         self.web_search_enabled = web_search_enabled
         self.system_prompt = system_prompt
+        self.session_id = session_id
+        print(f"[INIT] Handler created with web_search={web_search_enabled}, session_id={session_id}")
     def copy(self):
         # Get the most recent settings
                 return OpenAIHandler(
                     web_search_enabled=settings.get('web_search_enabled', False),
                     system_prompt=settings.get('system_prompt', ''),
+                    webrtc_id=recent_id,
+                    session_id=settings.get('session_id')
                 )
         print(f"[COPY] No settings found, creating default handler")
             )
             await self.connection.response.create()
     async def start_up(self):
         """Connect to realtime API"""
         # First check if we have the most recent settings
             if self.webrtc_id in connection_settings:
                 settings = connection_settings[self.webrtc_id]
                 self.web_search_enabled = settings.get('web_search_enabled', False)
                 self.system_prompt = settings.get('system_prompt', '')
+                self.session_id = settings.get('session_id')
                 print(f"[START_UP] Updated settings from storage for {self.webrtc_id}")
         self.client = openai.AsyncOpenAI()
+        # Connect to Realtime API
+        print(f"[REALTIME API] Connecting...")
         # Define the web search function
         tools = []
         base_instructions = self.system_prompt or "You are a helpful assistant."
         if self.web_search_enabled and self.search_client:
             tools = [{
                 "type": "function",
                 "than to guess or use outdated information."
             )
+            instructions = base_instructions + search_instructions
         else:
+            instructions = base_instructions
         async with self.client.beta.realtime.connect(
             model="gpt-4o-mini-realtime-preview-2024-12-17"
                 "temperature": 0.7,
                 "max_response_output_tokens": 4096,
                 "modalities": ["text", "audio"],
+                "voice": "alloy"
             }
             await conn.session.update(session=session_update)
             self.connection = conn
+            print(f"Connected with tools: {len(tools)} functions")
             async for event in self.connection:
                 # Debug logging for function calls
                 if event.type == "response.audio_transcript.done":
                     print(f"[RESPONSE] Transcript: {event.transcript[:100]}...")
+                    # Detect language
+                    detected_language = None
+                    try:
+                        if event.transcript and len(event.transcript) > 10:
+                            detected_language = detect(event.transcript)
+                    except:
+                        pass
+                    # Save to database
+                    if self.session_id:
+                        await ChatDatabase.save_message(self.session_id, "assistant", event.transcript)
                     output_data = {
                         "event": event,
+                        "detected_language": detected_language
                     }
                     await self.output_queue.put(AdditionalOutputs(output_data))
                             self.current_call_id = None
     async def receive(self, frame: tuple[int, np.ndarray]) -> None:
         if not self.connection:
+            print(f"[RECEIVE] No connection, skipping")
             return
         try:
             _, array = frame
             print(f"Error in receive: {e}")
     async def emit(self) -> tuple[int, np.ndarray] | AdditionalOutputs | None:
         item = await wait_for_item(self.output_queue)
         # Check if it's a dict with text message
     async def shutdown(self) -> None:
         print(f"[SHUTDOWN] Called")
         if self.connection:
             await self.connection.close()
             self.connection = None
+            print("[REALTIME API] Connection closed")
 # Create initial handler instance
 # Mount stream
 stream.mount(app)
+# Initialize database on startup
+@app.on_event("startup")
+async def startup_event():
+    await ChatDatabase.init()
+    print("Database initialized")
 # Intercept offer to capture settings
 @app.post("/webrtc/offer", include_in_schema=False)
 async def custom_offer(request: Request):
     webrtc_id = body.get("webrtc_id")
     web_search_enabled = body.get("web_search_enabled", False)
     system_prompt = body.get("system_prompt", "")
+    session_id = body.get("session_id")
     print(f"[OFFER] Received offer with webrtc_id: {webrtc_id}")
     print(f"[OFFER] web_search_enabled: {web_search_enabled}")
+    print(f"[OFFER] session_id: {session_id}")
     # Store settings with timestamp
     if webrtc_id:
         connection_settings[webrtc_id] = {
             'web_search_enabled': web_search_enabled,
             'system_prompt': system_prompt,
+            'session_id': session_id,
             'timestamp': asyncio.get_event_loop().time()
         }
     return response
+@app.post("/session/new")
+async def create_new_session():
+    """Create a new chat session"""
+    session_id = str(uuid.uuid4())
+    await ChatDatabase.create_session(session_id)
+    return {"session_id": session_id}
+@app.post("/message/save")
+async def save_message(request: Request):
+    """Save a message to the database"""
+    body = await request.json()
+    session_id = body.get("session_id")
+    role = body.get("role")
+    content = body.get("content")
+    if not all([session_id, role, content]):
+        return {"error": "Missing required fields"}
+    await ChatDatabase.save_message(session_id, role, content)
+    return {"status": "ok"}
+@app.get("/history/recent")
+async def get_recent_history():
+    """Get recent conversation history"""
+    conversations = await ChatDatabase.get_recent_conversations()
+    return conversations
+@app.get("/history/{session_id}")
+async def get_conversation(session_id: str):
+    """Get messages for a specific conversation"""
+    messages = await ChatDatabase.get_conversation_messages(session_id)
+    return messages
 @app.post("/chat/text")
 async def chat_text(request: Request):
     """Handle text chat messages using GPT-4o-mini"""
         body = await request.json()
         message = body.get("message", "")
         web_search_enabled = body.get("web_search_enabled", False)
         system_prompt = body.get("system_prompt", "")
+        session_id = body.get("session_id")
         if not message:
             return {"error": "메시지가 비어있습니다."}
         # Process text chat
+        result = await process_text_chat(message, web_search_enabled, system_prompt, session_id)
         return result
                         data = {
                             "role": "assistant",
                             "content": event_data['event'].transcript,
+                            "detected_language": event_data.get('detected_language')
                         }
                         yield f"event: output\ndata: {json.dumps(data)}\n\n"