Private-AI

Running

App Files Files Community

seawolf2357 commited on Jun 13

Commit

235fd3c

verified ·

1 Parent(s): 78b5e61

Update app.py

Browse files

Files changed (1) hide show

app.py +44 -674

app.py CHANGED Viewed

@@ -162,20 +162,6 @@ HTML_CONTENT = """<!DOCTYPE html>
             gap: 15px;
             margin-bottom: 15px;
         }
-        .interpretation-section {
-            display: flex;
-            flex-direction: column;
-            gap: 15px;
-            padding: 15px;
-            background-color: var(--dark-bg);
-            border-radius: 8px;
-            margin-top: 15px;
-        }
-        .interpretation-info {
-            font-size: 13px;
-            color: #999;
-            margin-top: 5px;
-        }
         .setting-item {
             display: flex;
             align-items: center;
@@ -252,46 +238,6 @@ HTML_CONTENT = """<!DOCTYPE html>
             resize: vertical;
             min-height: 80px;
         }
-        /* Multi-language selection */
-        .language-selection-grid {
-            display: grid;
-            grid-template-columns: repeat(2, 1fr);
-            gap: 10px;
-            margin-top: 10px;
-            max-height: 200px;
-            overflow-y: auto;
-            padding: 10px;
-            background-color: var(--dark-bg);
-            border-radius: 6px;
-            border: 1px solid var(--border-color);
-        }
-        .language-checkbox {
-            display: flex;
-            align-items: center;
-            gap: 8px;
-            font-size: 13px;
-            cursor: pointer;
-            padding: 5px;
-            border-radius: 4px;
-            transition: background-color 0.2s;
-        }
-        .language-checkbox:hover {
-            background-color: rgba(111, 66, 193, 0.1);
-        }
-        .language-checkbox input[type="checkbox"] {
-            width: 16px;
-            height: 16px;
-            cursor: pointer;
-        }
-        .language-checkbox.default {
-            font-weight: 500;
-            color: var(--primary-color);
-        }
-        .selected-languages {
-            margin-top: 10px;
-            font-size: 12px;
-            color: #999;
-        }
         .chat-container {
             border-radius: 12px;
             background-color: var(--card-bg);
@@ -358,63 +304,6 @@ HTML_CONTENT = """<!DOCTYPE html>
             padding: 10px;
             margin-bottom: 10px;
         }
-        .message.assistant.interpretation {
-            background: linear-gradient(135deg, #1a5a3e, #2e7d32);
-            font-style: normal;
-        }
-        .interpretation-result {
-            background: linear-gradient(135deg, #1e3a5f, #2c5282);
-            padding: 15px;
-            margin: 10px 0;
-            border-radius: 8px;
-            border: 1px solid rgba(66, 153, 225, 0.3);
-        }
-        .interpretation-header {
-            font-weight: bold;
-            color: #90cdf4;
-            margin-bottom: 10px;
-            display: flex;
-            align-items: center;
-            gap: 10px;
-        }
-        .interpretation-original {
-            font-size: 14px;
-            color: #cbd5e0;
-            margin-bottom: 15px;
-            padding: 10px;
-            background-color: rgba(0, 0, 0, 0.2);
-            border-radius: 4px;
-        }
-        .interpretation-translations {
-            display: flex;
-            flex-direction: column;
-            gap: 8px;
-        }
-        .translation-item {
-            display: flex;
-            align-items: baseline;
-            gap: 10px;
-            padding: 8px 12px;
-            background-color: rgba(255, 255, 255, 0.05);
-            border-radius: 4px;
-            border-left: 3px solid var(--primary-color);
-        }
-        .translation-lang {
-            font-weight: 500;
-            color: var(--primary-color);
-            min-width: 80px;
-            font-size: 13px;
-        }
-        .translation-text {
-            flex: 1;
-            color: var(--text-color);
-            font-size: 14px;
-        }
-        .interpretation-arrow {
-            color: #4caf50;
-            font-weight: bold;
-            margin: 0 10px;
-        }
         .language-info {
             font-size: 12px;
             color: #888;
@@ -639,7 +528,7 @@ HTML_CONTENT = """<!DOCTYPE html>
                             </div>
                         </div>
                         <div class="setting-item">
-                            <span class="setting-label">자동 번역</span>
                             <select id="language-select">
                                 <option value="">비활성화</option>
                                 <option value="ko">한국어 (Korean)</option>
@@ -674,30 +563,6 @@ HTML_CONTENT = """<!DOCTYPE html>
                             </select>
                         </div>
                     </div>
-                    <div class="interpretation-section">
-                        <div class="setting-item">
-                            <span class="setting-label">자동 통역</span>
-                            <div id="interpretation-toggle" class="toggle-switch">
-                                <div class="toggle-slider"></div>
-                            </div>
-                        </div>
-                        <div id="interpretation-languages-container" style="display: none;">
-                            <div class="setting-label" style="margin-bottom: 5px;">통역 언어 선택 (최대 4개)</div>
-                            <div class="language-selection-grid" id="language-selection-grid">
-                                <!-- Languages will be populated by JavaScript -->
-                            </div>
-                            <div class="selected-languages" id="selected-languages-display">
-                                선택된 언어: 없음
-                            </div>
-                        </div>
-                    </div>
-                    <div class="interpretation-info" id="interpretation-info" style="display: none;">
-                        <strong>통역 모드 안내:</strong><br>
-                        • 음성으로 말하면 선택한 언어들로 자동 통역됩니다<br>
-                        • Whisper + GPT-4o-mini를 사용합니다<br>
-                        • 말을 마치고 잠시 기다리면 통역이 시작됩니다<br>
-                        • 번역된 텍스트만 화면에 표시됩니다
-                    </div>
                     <div class="text-input-section">
                         <label for="system-prompt" class="setting-label">시스템 프롬프트:</label>
                         <textarea id="system-prompt" placeholder="AI 어시스턴트의 성격, 역할, 행동 방식을 정의하세요...">You are a helpful assistant. Respond in a friendly and professional manner.</textarea>
@@ -730,8 +595,6 @@ HTML_CONTENT = """<!DOCTYPE html>
         let webrtc_id;
         let webSearchEnabled = false;
         let selectedLanguage = "";
-        let interpretationMode = false;
-        let interpretationLanguages = [];
         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');
@@ -741,113 +604,14 @@ HTML_CONTENT = """<!DOCTYPE html>
         const statusText = document.getElementById('status-text');
         const searchToggle = document.getElementById('search-toggle');
         const languageSelect = document.getElementById('language-select');
-        const interpretationToggle = document.getElementById('interpretation-toggle');
-        const interpretationLanguagesContainer = document.getElementById('interpretation-languages-container');
-        const interpretationInfo = document.getElementById('interpretation-info');
         const systemPromptInput = document.getElementById('system-prompt');
         const textInput = document.getElementById('text-input');
-        const languageSelectionGrid = document.getElementById('language-selection-grid');
-        const selectedLanguagesDisplay = document.getElementById('selected-languages-display');
         let audioLevel = 0;
         let animationFrame;
         let audioContext, analyser, audioSource;
         let dataChannel = null;
         let isVoiceActive = false;
-        // Available languages for interpretation
-        const INTERPRETATION_LANGUAGES = {
-            "en": { name: "English", default: true },
-            "zh": { name: "中文 (Chinese)", default: true },
-            "th": { name: "ไทย (Thai)", default: true },
-            "ru": { name: "Русский (Russian)", default: true },
-            "ja": { name: "日本語 (Japanese)", default: false },
-            "es": { name: "Español (Spanish)", default: false },
-            "fr": { name: "Français (French)", default: false },
-            "de": { name: "Deutsch (German)", default: false },
-            "pt": { name: "Português (Portuguese)", default: false },
-            "ar": { name: "العربية (Arabic)", default: false },
-            "hi": { name: "हिन्दी (Hindi)", default: false },
-            "vi": { name: "Tiếng Việt (Vietnamese)", default: false },
-            "id": { name: "Bahasa Indonesia", default: false },
-            "it": { name: "Italiano (Italian)", default: false },
-            "nl": { name: "Nederlands (Dutch)", default: false },
-            "pl": { name: "Polski (Polish)", default: false },
-            "tr": { name: "Türkçe (Turkish)", default: false },
-            "sv": { name: "Svenska (Swedish)", default: false },
-            "da": { name: "Dansk (Danish)", default: false },
-            "no": { name: "Norsk (Norwegian)", default: false },
-            "fi": { name: "Suomi (Finnish)", default: false },
-            "he": { name: "עברית (Hebrew)", default: false },
-            "uk": { name: "Українська (Ukrainian)", default: false },
-            "cs": { name: "Čeština (Czech)", default: false },
-            "el": { name: "Ελληνικά (Greek)", default: false },
-            "ro": { name: "Română (Romanian)", default: false },
-            "hu": { name: "Magyar (Hungarian)", default: false },
-            "ms": { name: "Bahasa Melayu (Malay)", default: false }
-        };
-        // Initialize language selection grid
-        function initializeLanguageSelection() {
-            languageSelectionGrid.innerHTML = '';
-            // Sort languages: defaults first, then alphabetically
-            const sortedLanguages = Object.entries(INTERPRETATION_LANGUAGES).sort((a, b) => {
-                if (a[1].default && !b[1].default) return -1;
-                if (!a[1].default && b[1].default) return 1;
-                return a[1].name.localeCompare(b[1].name);
-            });
-            sortedLanguages.forEach(([code, lang]) => {
-                const label = document.createElement('label');
-                label.className = 'language-checkbox' + (lang.default ? ' default' : '');
-                const checkbox = document.createElement('input');
-                checkbox.type = 'checkbox';
-                checkbox.value = code;
-                checkbox.checked = lang.default;
-                checkbox.addEventListener('change', onLanguageCheckboxChange);
-                const text = document.createElement('span');
-                text.textContent = lang.name;
-                label.appendChild(checkbox);
-                label.appendChild(text);
-                languageSelectionGrid.appendChild(label);
-            });
-            // Initialize with default languages
-            updateSelectedLanguages();
-        }
-        function onLanguageCheckboxChange() {
-            const checkedBoxes = languageSelectionGrid.querySelectorAll('input[type="checkbox"]:checked');
-            // Limit to 4 languages
-            if (checkedBoxes.length > 4) {
-                this.checked = false;
-                showError('최대 4개 언어까지 선택할 수 있습니다.');
-                return;
-            }
-            updateSelectedLanguages();
-        }
-        function updateSelectedLanguages() {
-            const checkedBoxes = languageSelectionGrid.querySelectorAll('input[type="checkbox"]:checked');
-            interpretationLanguages = Array.from(checkedBoxes).map(cb => cb.value);
-            if (interpretationLanguages.length === 0) {
-                selectedLanguagesDisplay.textContent = '선택된 언어: 없음';
-            } else {
-                const langNames = interpretationLanguages.map(code =>
-                    INTERPRETATION_LANGUAGES[code].name
-                ).join(', ');
-                selectedLanguagesDisplay.textContent = `선택된 언어 (${interpretationLanguages.length}/4): ${langNames}`;
-            }
-            console.log('Selected interpretation languages:', interpretationLanguages);
-        }
         // Web search toggle functionality
         searchToggle.addEventListener('click', () => {
             webSearchEnabled = !webSearchEnabled;
@@ -861,90 +625,6 @@ HTML_CONTENT = """<!DOCTYPE html>
             console.log('Selected language:', selectedLanguage);
         });
-        // Interpretation mode toggle - 수정된 버전
-        interpretationToggle.addEventListener('click', async () => {
-            if (!interpretationMode) {
-                // Turning ON interpretation mode
-                interpretationLanguagesContainer.style.display = 'block';
-                interpretationInfo.style.display = 'block';
-                // Check if any languages are selected
-                if (interpretationLanguages.length === 0) {
-                    showError('통역 언어를 선택해주세요.');
-                    interpretationToggle.classList.remove('active');
-                    return;
-                }
-                // Enable interpretation mode
-                interpretationMode = true;
-                interpretationToggle.classList.add('active');
-                // Disable other features
-                languageSelect.value = '';
-                selectedLanguage = '';
-                languageSelect.disabled = true;
-                searchToggle.classList.remove('active');
-                webSearchEnabled = false;
-                searchToggle.style.opacity = '0.5';
-                searchToggle.style.pointerEvents = 'none';
-                textInput.disabled = true;
-                textInput.placeholder = '통역 모드에서는 텍스트 입력이 지원되지 않습니다';
-                sendButton.style.display = 'none';
-                console.log('[FRONTEND] Interpretation mode enabled with languages:', interpretationLanguages);
-                // If connected, stop and restart with interpretation mode
-                if (peerConnection && peerConnection.connectionState === 'connected') {
-                    console.log('[FRONTEND] Stopping current connection for interpretation mode');
-                    showError('통역 모드 설정을 적용하기 위해 연결을 다시 시작합니다.');
-                    // Force stop the connection
-                    stop();
-                    // Wait a bit longer to ensure cleanup
-                    await new Promise(resolve => setTimeout(resolve, 1000));
-                    // Start new connection with interpretation mode
-                    console.log('[FRONTEND] Starting new connection with interpretation mode');
-                    setupWebRTC();
-                }
-            } else {
-                // Turning OFF interpretation mode
-                interpretationMode = false;
-                interpretationToggle.classList.remove('active');
-                interpretationLanguagesContainer.style.display = 'none';
-                interpretationInfo.style.display = 'none';
-                // Re-enable other features
-                languageSelect.disabled = false;
-                searchToggle.style.opacity = '1';
-                searchToggle.style.pointerEvents = 'auto';
-                textInput.disabled = false;
-                textInput.placeholder = '텍스트 메시지를 입력하세요...';
-                sendButton.style.display = 'block';
-                console.log('[FRONTEND] Interpretation mode disabled');
-                // If connected, restart to apply normal mode
-                if (peerConnection && peerConnection.connectionState === 'connected') {
-                    console.log('[FRONTEND] Stopping current connection for normal mode');
-                    showError('일반 모드로 전환하기 위해 연결을 다시 시작합니다.');
-                    // Force stop the connection
-                    stop();
-                    // Wait a bit longer to ensure cleanup
-                    await new Promise(resolve => setTimeout(resolve, 1000));
-                    // Start new connection with normal mode
-                    console.log('[FRONTEND] Starting new connection with normal mode');
-                    setupWebRTC();
-                }
-            }
-            console.log('[FRONTEND] Final interpretation mode:', interpretationMode);
-        });
         // System prompt update
         systemPromptInput.addEventListener('input', () => {
             systemPrompt = systemPromptInput.value || "You are a helpful assistant. Respond in a friendly and professional manner.";
@@ -964,12 +644,6 @@ HTML_CONTENT = """<!DOCTYPE html>
             const message = textInput.value.trim();
             if (!message) return;
-            // Don't allow text messages in interpretation mode
-            if (interpretationMode) {
-                showError('통역 모드에서는 텍스트 입력이 지원되지 않습니다.');
-                return;
-            }
             // Add user message to chat
             addMessage('user', message);
             textInput.value = '';
@@ -1023,18 +697,14 @@ HTML_CONTENT = """<!DOCTYPE html>
             statusDot.className = 'status-dot ' + state;
             if (state === 'connected') {
                 statusText.textContent = '연결됨';
-                if (!interpretationMode) {
-                    sendButton.style.display = 'block';
-                }
                 isVoiceActive = true;
             } else if (state === 'connecting') {
                 statusText.textContent = '연결 중...';
                 sendButton.style.display = 'none';
             } else {
                 statusText.textContent = '연결 대기 중';
-                if (!interpretationMode) {
-                    sendButton.style.display = 'block';  // Show send button even when disconnected for text chat
-                }
                 isVoiceActive = false;
             }
         }
@@ -1180,9 +850,7 @@ HTML_CONTENT = """<!DOCTYPE html>
                     webrtc_id: webrtc_id,
                     web_search_enabled: webSearchEnabled,
                     target_language: selectedLanguage,
-                    system_prompt: systemPrompt,
-                    interpretation_mode: interpretationMode,
-                    interpretation_languages: interpretationLanguages
                 });
                 const response = await fetch('/webrtc/offer', {
@@ -1194,9 +862,7 @@ HTML_CONTENT = """<!DOCTYPE html>
                         webrtc_id: webrtc_id,
                         web_search_enabled: webSearchEnabled,
                         target_language: selectedLanguage,
-                        system_prompt: systemPrompt,
-                        interpretation_mode: interpretationMode,
-                        interpretation_languages: interpretationLanguages
                     })
                 });
                 const serverResponse = await response.json();
@@ -1213,25 +879,10 @@ HTML_CONTENT = """<!DOCTYPE html>
                     const eventJson = JSON.parse(event.data);
                     let content = eventJson.content;
-                    // Debug logging for interpretation mode
-                    if (interpretationMode) {
-                        console.log('[INTERPRETATION OUTPUT]', {
-                            content: content,
-                            mode: eventJson.mode,
-                            translations: eventJson.translations
-                        });
-                    }
                     if (selectedLanguage && eventJson.language) {
                         content += ` <span class="language-info">[${eventJson.language}]</span>`;
-                    } else if (interpretationMode && eventJson.mode === 'interpretation') {
-                        // Handle interpretation mode output
-                        if (eventJson.original && eventJson.translations) {
-                            addInterpretationResult(eventJson.original, eventJson.translations);
-                        }
-                    } else {
-                        addMessage("assistant", content);
                     }
                 });
                 eventSource.addEventListener("search", (event) => {
                     const eventJson = JSON.parse(event.data);
@@ -1259,44 +910,6 @@ HTML_CONTENT = """<!DOCTYPE html>
             chatMessages.scrollTop = chatMessages.scrollHeight;
         }
-        function addInterpretationResult(original, translations) {
-            const resultDiv = document.createElement('div');
-            resultDiv.className = 'interpretation-result';
-            // Header
-            const headerDiv = document.createElement('div');
-            headerDiv.className = 'interpretation-header';
-            headerDiv.innerHTML = `
-                <span>🌐</span>
-                <span>자동 통역 결과</span>
-            `;
-            resultDiv.appendChild(headerDiv);
-            // Original text
-            const originalDiv = document.createElement('div');
-            originalDiv.className = 'interpretation-original';
-            originalDiv.innerHTML = `<strong>원문:</strong> ${original}`;
-            resultDiv.appendChild(originalDiv);
-            // Translations
-            const translationsDiv = document.createElement('div');
-            translationsDiv.className = 'interpretation-translations';
-            translations.forEach(trans => {
-                const transItem = document.createElement('div');
-                transItem.className = 'translation-item';
-                transItem.innerHTML = `
-                    <span class="translation-lang">${trans.language}:</span>
-                    <span class="translation-text">${trans.text}</span>
-                `;
-                translationsDiv.appendChild(transItem);
-            });
-            resultDiv.appendChild(translationsDiv);
-            chatMessages.appendChild(resultDiv);
-            chatMessages.scrollTop = chatMessages.scrollHeight;
-        }
         function stop() {
             console.log('[STOP] Stopping connection...');
@@ -1387,7 +1000,6 @@ HTML_CONTENT = """<!DOCTYPE html>
         // Initialize on page load
         window.addEventListener('DOMContentLoaded', () => {
             sendButton.style.display = 'block';
-            initializeLanguageSelection();
         });
     </script>
 </body>
@@ -1566,8 +1178,7 @@ 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,
-                 interpretation_mode: bool = False, interpretation_languages: List[str] = None) -> None:
         super().__init__(
             expected_layout="mono",
             output_sample_rate=SAMPLE_RATE,
@@ -1584,18 +1195,9 @@ class OpenAIHandler(AsyncStreamHandler):
         self.web_search_enabled = web_search_enabled
         self.target_language = target_language
         self.system_prompt = system_prompt
-        self.interpretation_mode = interpretation_mode
-        self.interpretation_languages = interpretation_languages or []
-        # For interpretation mode
-        self.audio_buffer = []
-        self.is_recording = False
-        self.silence_frames = 0
-        self.silence_threshold = 20  # Reduced for faster response (20 frames = ~0.4 seconds)
-        self.min_audio_length = 10  # Minimum frames to consider as speech
-        print(f"[INIT] Handler created with interpretation_mode={interpretation_mode}, "
-              f"interpretation_languages={interpretation_languages}")
     def copy(self):
         # Get the most recent settings
@@ -1610,20 +1212,16 @@ class OpenAIHandler(AsyncStreamHandler):
                 # Log the settings being copied
                 print(f"[COPY] Copying settings from {recent_id}:")
-                print(f"[COPY] interpretation_mode={settings.get('interpretation_mode', False)}")
-                print(f"[COPY] interpretation_languages={settings.get('interpretation_languages', [])}")
                 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,
-                    interpretation_mode=settings.get('interpretation_mode', False),
-                    interpretation_languages=settings.get('interpretation_languages', [])
                 )
         print(f"[COPY] No settings found, creating default handler")
-        return OpenAIHandler(web_search_enabled=False, interpretation_mode=False)
     async def search_web(self, query: str) -> str:
         """Perform web search and return formatted results"""
@@ -1658,173 +1256,9 @@ class OpenAIHandler(AsyncStreamHandler):
             )
             await self.connection.response.create()
-    async def _interpretation_loop(self):
-        """Keep the handler alive for interpretation mode"""
-        while self.interpretation_mode:
-            await asyncio.sleep(0.1)
-    async def process_interpretation(self):
-        """Process audio buffer for interpretation - text only output"""
-        if not self.audio_buffer or not self.interpretation_languages:
-            return
-        try:
-            print(f"[INTERPRETATION] Processing audio buffer with {len(self.audio_buffer)} frames")
-            # Convert audio buffer to WAV format
-            audio_data = np.concatenate(self.audio_buffer)
-            # Create WAV file in memory
-            wav_buffer = io.BytesIO()
-            with wave.open(wav_buffer, 'wb') as wav_file:
-                wav_file.setnchannels(1)  # Mono
-                wav_file.setsampwidth(2)   # 16-bit
-                wav_file.setframerate(SAMPLE_RATE)
-                wav_file.writeframes(audio_data.tobytes())
-            wav_buffer.seek(0)
-            wav_buffer.name = "audio.wav"
-            # 1. Transcribe with Whisper
-            print("[INTERPRETATION] Transcribing with Whisper...")
-            transcript = await self.client.audio.transcriptions.create(
-                model="whisper-1",
-                file=wav_buffer,
-                response_format="text"  # Get only text, no timestamps
-            )
-            user_text = transcript.text.strip()
-            print(f"[INTERPRETATION] Transcribed: {user_text}")
-            if not user_text:
-                return
-            # 2. Translate to all selected languages
-            translations = []
-            for lang_code in self.interpretation_languages:
-                target_lang_name = SUPPORTED_LANGUAGES.get(lang_code, lang_code)
-                # Create very explicit translation prompt
-                translation_examples = {
-                    "en": {
-                        "안녕하세요": "Hello",
-                        "감사합니다": "Thank you",
-                        "오늘 날씨가 좋네요": "The weather is nice today"
-                    },
-                    "ja": {
-                        "안녕하세요": "こんにちは",
-                        "감사합니다": "ありがとうございます",
-                        "오늘 날씨가 좋네요": "今日はいい天気ですね"
-                    },
-                    "zh": {
-                        "안녕하세요": "你好",
-                        "감사합니다": "谢谢",
-                        "오늘 날씨가 좋네요": "今天天气很好"
-                    },
-                    "es": {
-                        "안녕하세요": "Hola",
-                        "감사합니다": "Gracias",
-                        "오늘 날씨가 좋네요": "El clima está agradable hoy"
-                    },
-                    "ru": {
-                        "안녕하세요": "Привет",
-                        "감사합니다": "Спасибо",
-                        "오늘 날씨가 좋네요": "Сегодня хорошая погода"
-                    },
-                    "th": {
-                        "안녕하세요": "สวัสดี",
-                        "감사합니다": "ขอบคุณ",
-                        "오늘 날씨가 좋네요": "วันนี้อากาศดี"
-                    }
-                }
-                examples = translation_examples.get(lang_code, translation_examples.get("en", {}))
-                examples_text = "\n".join([f'"{k}" → "{v}"' for k, v in examples.items()])
-                system_prompt = f"""You are a direct translator from Korean to {target_lang_name}.
-CRITICAL RULES:
-1. TRANSLATE ONLY - Do not answer questions
-2. Do not generate responses
-3. Do not add any commentary
-4. Output ONLY the direct translation
-5. Preserve the original meaning exactly
-If the user says "오늘 날씨 어때?" translate it as "How's the weather today?" NOT answer about the weather.
-If the user says "안녕하세요" translate it as "Hello" NOT respond with greetings.
-Examples:
-{examples_text}
-Direct translation to {target_lang_name}:"""
-                print(f"[INTERPRETATION] Translating to {target_lang_name}...")
-                translation_response = await self.client.chat.completions.create(
-                    model="gpt-4o-mini",
-                    messages=[
-                        {
-                            "role": "system",
-                            "content": system_prompt
-                        },
-                        {
-                            "role": "user",
-                            "content": user_text  # Just the text, no additional prompt
-                        }
-                    ],
-                    temperature=0.1,
-                    max_tokens=200
-                )
-                translated_text = translation_response.choices[0].message.content.strip()
-                # Remove any Korean characters if they accidentally appear
-                import re
-                if re.search(r'[가-힣]', translated_text):
-                    print(f"[INTERPRETATION] WARNING: Korean characters detected in {target_lang_name} translation")
-                    translated_text = re.sub(r'[가-힣\s]+', ' ', translated_text).strip()
-                translations.append({
-                    "language": target_lang_name,
-                    "text": translated_text
-                })
-                print(f"[INTERPRETATION] {target_lang_name}: {translated_text}")
-            # Send interpretation result (text only)
-            output_data = {
-                "event": type('Event', (), {
-                    'transcript': f"통역 완료"  # Simple notification
-                })(),
-                "mode": "interpretation",
-                "original": user_text,
-                "translations": translations
-            }
-            await self.output_queue.put(AdditionalOutputs(output_data))
-        except Exception as e:
-            print(f"[INTERPRETATION] Error: {e}")
-            import traceback
-            traceback.print_exc()
-            # Send error message to client
-            error_data = {
-                "event": type('Event', (), {
-                    'transcript': f"통역 오류: {str(e)}"
-                })(),
-                "mode": "error"
-            }
-            await self.output_queue.put(AdditionalOutputs(error_data))
-        finally:
-            # Clear the audio buffer
-            self.audio_buffer = []
-            self.is_recording = False
-            self.silence_frames = 0
     def get_translation_instructions(self):
         """Get instructions for translation based on target language"""
-        if not self.target_language or self.interpretation_mode:
             return ""
         language_name = SUPPORTED_LANGUAGES.get(self.target_language, self.target_language)
@@ -1835,7 +1269,7 @@ Direct translation to {target_lang_name}:"""
         )
     async def start_up(self):
-        """Connect to realtime API or setup interpretation mode"""
         # First check if we have the most recent settings
         if connection_settings and self.webrtc_id:
             if self.webrtc_id in connection_settings:
@@ -1843,27 +1277,13 @@ Direct translation to {target_lang_name}:"""
                 self.web_search_enabled = settings.get('web_search_enabled', False)
                 self.target_language = settings.get('target_language', '')
                 self.system_prompt = settings.get('system_prompt', '')
-                self.interpretation_mode = settings.get('interpretation_mode', False)
-                self.interpretation_languages = settings.get('interpretation_languages', [])
                 print(f"[START_UP] Updated settings from storage for {self.webrtc_id}")
-                print(f"[START_UP] interpretation_mode={self.interpretation_mode}")
-                print(f"[START_UP] interpretation_languages={self.interpretation_languages}")
-        print(f"[START_UP] Starting with interpretation_mode={self.interpretation_mode}")
         self.client = openai.AsyncOpenAI()
-        # If in interpretation mode, don't connect to Realtime API
-        if self.interpretation_mode:
-            print(f"[INTERPRETATION MODE] Active - Skipping Realtime API connection")
-            print(f"[INTERPRETATION MODE] Using Whisper + GPT-4o-mini (text only)")
-            print(f"[INTERPRETATION MODE] Target languages: {self.interpretation_languages}")
-            # Do NOT connect to Realtime API
-            # Just keep the handler ready to process audio
-            return
         # Normal mode - connect to Realtime API
         print(f"[NORMAL MODE] Connecting to Realtime API...")
@@ -2080,18 +1500,18 @@ RULES:
                         ),
                     )
-                # Handle function calls (only in non-interpretation mode)
-                elif event.type == "response.function_call_arguments.start" and not self.interpretation_mode:
                     print(f"Function call started")
                     self.function_call_in_progress = True
                     self.current_function_args = ""
                     self.current_call_id = getattr(event, 'call_id', None)
-                elif event.type == "response.function_call_arguments.delta" and not self.interpretation_mode:
                     if self.function_call_in_progress:
                         self.current_function_args += event.delta
-                elif event.type == "response.function_call_arguments.done" and not self.interpretation_mode:
                     if self.function_call_in_progress:
                         print(f"Function call done, args: {self.current_function_args}")
                         try:
@@ -2127,82 +1547,41 @@ RULES:
                             self.current_call_id = None
     async def receive(self, frame: tuple[int, np.ndarray]) -> None:
-        if self.interpretation_mode:
-            print(f"[RECEIVE] Processing in interpretation mode")
-            # In interpretation mode, buffer audio and process with Whisper
             _, array = frame
             array = array.squeeze()
-            # Simple voice activity detection
-            audio_level = np.abs(array).mean()
-            if audio_level > 200:  # Lower threshold for better detection
-                if not self.is_recording:
-                    print(f"[INTERPRETATION] Started recording, level: {audio_level:.1f}")
-                self.is_recording = True
-                self.silence_frames = 0
-                self.audio_buffer.append(array)
-            elif self.is_recording:
-                self.silence_frames += 1
-                self.audio_buffer.append(array)
-                # If we've had enough silence, process the audio
-                if self.silence_frames > self.silence_threshold and len(self.audio_buffer) > self.min_audio_length:
-                    print(f"[INTERPRETATION] Silence detected after {len(self.audio_buffer)} frames")
-                    # Process in the background to avoid blocking
-                    asyncio.create_task(self.process_interpretation())
-        else:
-            # Normal mode - use Realtime API
-            if not self.connection:
-                print(f"[RECEIVE] No connection in normal mode, skipping")
-                return
-            try:
-                _, array = frame
-                array = array.squeeze()
-                audio_message = base64.b64encode(array.tobytes()).decode("utf-8")
-                await self.connection.input_audio_buffer.append(audio=audio_message)
-            except Exception as e:
-                print(f"Error in receive: {e}")
     async def emit(self) -> tuple[int, np.ndarray] | AdditionalOutputs | None:
-        # In interpretation mode, we need to keep checking for outputs
-        if self.interpretation_mode:
-            # Use a short timeout to prevent blocking
-            try:
-                item = await asyncio.wait_for(wait_for_item(self.output_queue), timeout=0.05)
-                return item
-            except asyncio.TimeoutError:
-                # Return None to keep the stream alive
-                return None
-        else:
-            # Normal mode
-            item = await wait_for_item(self.output_queue)
-            # Check if it's a dict with text message
-            if isinstance(item, dict) and item.get('type') == 'text_message':
-                await self.process_text_message(item['content'])
-                return None
-            return item
     async def shutdown(self) -> None:
-        print(f"[SHUTDOWN] Called with interpretation_mode={self.interpretation_mode}")
-        if self.interpretation_mode:
-            # Clean up interpretation mode
-            self.audio_buffer = []
-            self.is_recording = False
-            print("[INTERPRETATION MODE] Shutdown complete")
-        else:
-            # 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
-handler = OpenAIHandler(web_search_enabled=False, interpretation_mode=False)
 # Create components
 chatbot = gr.Chatbot(type="messages")
@@ -2235,12 +1614,8 @@ async def custom_offer(request: Request):
     web_search_enabled = body.get("web_search_enabled", False)
     target_language = body.get("target_language", "")
     system_prompt = body.get("system_prompt", "")
-    interpretation_mode = body.get("interpretation_mode", False)
-    interpretation_languages = body.get("interpretation_languages", [])
     print(f"[OFFER] Received offer with webrtc_id: {webrtc_id}")
-    print(f"[OFFER] interpretation_mode: {interpretation_mode}")
-    print(f"[OFFER] interpretation_languages: {interpretation_languages}")
     print(f"[OFFER] web_search_enabled: {web_search_enabled}")
     print(f"[OFFER] target_language: {target_language}")
@@ -2250,8 +1625,6 @@ async def custom_offer(request: Request):
             'web_search_enabled': web_search_enabled,
             'target_language': target_language,
             'system_prompt': system_prompt,
-            'interpretation_mode': interpretation_mode,
-            'interpretation_languages': interpretation_languages,
             'timestamp': asyncio.get_event_loop().time()
         }
@@ -2335,10 +1708,7 @@ async def outputs(webrtc_id: str):
                         data = {
                             "role": "assistant",
                             "content": event_data['event'].transcript,
-                            "language": event_data.get('language', ''),
-                            "mode": event_data.get('mode', 'normal'),
-                            "original": event_data.get('original', ''),
-                            "translations": event_data.get('translations', [])
                         }
                         yield f"event: output\ndata: {json.dumps(data)}\n\n"

             gap: 15px;
             margin-bottom: 15px;
         }
         .setting-item {
             display: flex;
             align-items: center;
             resize: vertical;
             min-height: 80px;
         }
         .chat-container {
             border-radius: 12px;
             background-color: var(--card-bg);
             padding: 10px;
             margin-bottom: 10px;
         }
         .language-info {
             font-size: 12px;
             color: #888;
                             </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>
                             </select>
                         </div>
                     </div>
                     <div class="text-input-section">
                         <label for="system-prompt" class="setting-label">시스템 프롬프트:</label>
                         <textarea id="system-prompt" placeholder="AI 어시스턴트의 성격, 역할, 행동 방식을 정의하세요...">You are a helpful assistant. Respond in a friendly and professional manner.</textarea>
         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 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;
             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.";
             const message = textInput.value.trim();
             if (!message) return;
             // Add user message to chat
             addMessage('user', message);
             textInput.value = '';
             statusDot.className = 'status-dot ' + state;
             if (state === 'connected') {
                 statusText.textContent = '연결됨';
+                sendButton.style.display = 'block';
                 isVoiceActive = true;
             } else if (state === 'connecting') {
                 statusText.textContent = '연결 중...';
                 sendButton.style.display = 'none';
             } else {
                 statusText.textContent = '연결 대기 중';
+                sendButton.style.display = 'block';  // Show send button even when disconnected for text chat
                 isVoiceActive = false;
             }
         }
                     webrtc_id: webrtc_id,
                     web_search_enabled: webSearchEnabled,
                     target_language: selectedLanguage,
+                    system_prompt: systemPrompt
                 });
                 const response = await fetch('/webrtc/offer', {
                         webrtc_id: webrtc_id,
                         web_search_enabled: webSearchEnabled,
                         target_language: selectedLanguage,
+                        system_prompt: systemPrompt
                     })
                 });
                 const serverResponse = await response.json();
                     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);
                 });
                 eventSource.addEventListener("search", (event) => {
                     const eventJson = JSON.parse(event.data);
             chatMessages.scrollTop = chatMessages.scrollHeight;
         }
         function stop() {
             console.log('[STOP] Stopping connection...');
         // Initialize on page load
         window.addEventListener('DOMContentLoaded', () => {
             sendButton.style.display = 'block';
         });
     </script>
 </body>
 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,
         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
                 # Log the settings being copied
                 print(f"[COPY] Copying settings from {recent_id}:")
                 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")
+        return OpenAIHandler(web_search_enabled=False)
     async def search_web(self, query: str) -> str:
         """Perform web search and return formatted results"""
             )
             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)
         )
     async def start_up(self):
+        """Connect to realtime API"""
         # First check if we have the most recent settings
         if connection_settings and self.webrtc_id:
             if self.webrtc_id in connection_settings:
                 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...")
                         ),
                     )
+                # Handle function calls
+                elif event.type == "response.function_call_arguments.start":
                     print(f"Function call started")
                     self.function_call_in_progress = True
                     self.current_function_args = ""
                     self.current_call_id = getattr(event, 'call_id', None)
+                elif event.type == "response.function_call_arguments.delta":
                     if self.function_call_in_progress:
                         self.current_function_args += event.delta
+                elif event.type == "response.function_call_arguments.done":
                     if self.function_call_in_progress:
                         print(f"Function call done, args: {self.current_function_args}")
                         try:
                             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
             array = array.squeeze()
+            audio_message = base64.b64encode(array.tobytes()).decode("utf-8")
+            await self.connection.input_audio_buffer.append(audio=audio_message)
+        except Exception as e:
+            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
+        if isinstance(item, dict) and item.get('type') == 'text_message':
+            await self.process_text_message(item['content'])
+            return None
+        return item
     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
+handler = OpenAIHandler(web_search_enabled=False)
 # Create components
 chatbot = gr.Chatbot(type="messages")
     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}")
             'web_search_enabled': web_search_enabled,
             'target_language': target_language,
             'system_prompt': system_prompt,
             'timestamp': asyncio.get_event_loop().time()
         }
                         data = {
                             "role": "assistant",
                             "content": event_data['event'].transcript,
+                            "language": event_data.get('language', '')
                         }
                         yield f"event: output\ndata: {json.dumps(data)}\n\n"