Update app.py

app.py

@@ -32,13 +32,8 @@ except Exception as e:
     print("Continuing with limited functionality...")
 
 # ----------------------- Global Variables ----------------------- #
-VOICE_CHOICES = {
-    '🇺🇸 Female (Default)': 'af',
-    '🇺🇸 Bella': 'af_bella',
-    '🇺🇸 Sarah': 'af_sarah',
-    '🇺🇸 Nicole': 'af_nicole'
-}
-TTS_ENABLED = False  # 기본적으로 TTS 모듈 초기화 실패 시 비활성화
+# 음성 관련 변수는 더 이상 사용하지 않음
+# VOICE_CHOICES = { ... }  --> 제거
 
 # ----------------------- Model and Tokenizer Initialization ----------------------- #
 model_name = "deepseek-ai/DeepSeek-R1-Distill-Llama-8B"
@@ -56,16 +51,8 @@ def init_models():
     return model
 
 # ----------------------- Kokoro TTS Initialization ----------------------- #
-try:
-    import sys
-    sys.path.append('Kokoro-82M')
-    from models import build_model
-    from kokoro import generate
-
-    TTS_ENABLED = True
-except Exception as e:
-    print(f"Warning: Could not initialize Kokoro TTS: {str(e)}")
-    TTS_ENABLED = False
+# 음성 기능 제거: TTS 초기화 관련 코드는 더 이상 사용하지 않음
+TTS_ENABLED = False
 
 # ----------------------- Web Search Functions ----------------------- #
 def get_web_results(query, max_results=5):
@@ -82,19 +69,18 @@ def get_web_results(query, max_results=5):
         return []
 
 def format_prompt(query, context):
+    """웹 검색 결과를 바탕으로 간결하고 요약된 답변을 생성하도록 프롬프트를 구성합니다."""
     current_time = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
-    context_lines = '\n'.join([f'- [{res["title"]}]: {res["snippet"]}' for res in context])
-    return f"""You are an intelligent search assistant. Answer the user's query using the provided web context.
+    context_lines = '\n'.join([f'- {res["title"]}: {res["snippet"]}' for res in context])
+    return f"""You are an intelligent search assistant. Your task is to provide a concise, clear summary answer to the user's query based solely on the provided web context.
 Current Time: {current_time}
 
-Important: For election-related queries, please distinguish clearly between different election years and types (presidential vs. non-presidential). Only use information from the provided web context.
-
 Query: {query}
 
 Web Context:
 {context_lines}
 
-Provide a detailed answer in markdown format. Include relevant information from sources and cite them using [1], [2], etc. If the query is about elections, clearly specify which year and type of election you're discussing.
+Please provide a summary answer in markdown format, including citations such as [1], [2], etc. in your answer if needed.
 Answer:"""
 
 def format_sources(web_results):
@@ -144,96 +130,36 @@ def generate_answer(prompt):
     )
     return tokenizer.decode(outputs[0], skip_special_tokens=True)
 
-@spaces.GPU(duration=60)
-def generate_speech_with_gpu(text, voice_name='af'):
-    try:
-        device = 'cuda'
-        TTS_MODEL = build_model('Kokoro-82M/kokoro-v0_19.pth', device)
-        VOICEPACK = torch.load(f'Kokoro-82M/voices/{voice_name}.pt', weights_only=True).to(device)
-        
-        clean_text = ' '.join([line for line in text.split('\n') if not line.startswith('#')])
-        clean_text = clean_text.replace('[', '').replace(']', '').replace('*', '')
-        
-        max_chars = 1000
-        if len(clean_text) > max_chars:
-            sentences = clean_text.split('.')
-            chunks = []
-            current_chunk = ""
-            for sentence in sentences:
-                if len(current_chunk) + len(sentence) < max_chars:
-                    current_chunk += sentence + "."
-                else:
-                    if current_chunk:
-                        chunks.append(current_chunk)
-                    current_chunk = sentence + "."
-            if current_chunk:
-                chunks.append(current_chunk)
-        else:
-            chunks = [clean_text]
-        
-        audio_chunks = []
-        for chunk in chunks:
-            if chunk.strip():
-                chunk_audio, _ = generate(TTS_MODEL, chunk.strip(), VOICEPACK, lang='a')
-                if isinstance(chunk_audio, torch.Tensor):
-                    chunk_audio = chunk_audio.cpu().numpy()
-                audio_chunks.append(chunk_audio)
-        
-        if audio_chunks:
-            final_audio = np.concatenate(audio_chunks) if len(audio_chunks) > 1 else audio_chunks[0]
-            return (24000, final_audio)
-        return None
-        
-    except Exception as e:
-        print(f"Error generating speech: {str(e)}")
-        import traceback
-        traceback.print_exc()
-        return None
-
-def process_query(query, history, selected_voice='af'):
+# ----------------------- Process Query and Output Summary ----------------------- #
+def process_query(query, history):
     try:
         if history is None:
             history = []
             
+        # 웹 검색 결과 가져오기
         web_results = get_web_results(query)
         sources_html = format_sources(web_results)
         
+        # 중간 상태 표시
         current_history = history + [[query, "*Searching...*"]]
         yield {
-            answer_output: gr.Markdown("*Searching & Thinking...*"),
+            answer_output: gr.Markdown("*Searching & Summarizing...*"),
             sources_output: gr.HTML(sources_html),
             search_btn: gr.Button("Searching...", interactive=False),
-            chat_history_display: current_history,
-            audio_output: None
+            chat_history_display: current_history
         }
         
+        # 프롬프트 생성: 웹 결과를 요약하는 형태로 구성
         prompt_text = format_prompt(query, web_results)
         answer = generate_answer(prompt_text)
         final_answer = answer.split("Answer:")[-1].strip()
         
-        if TTS_ENABLED:
-            try:
-                yield {
-                    answer_output: gr.Markdown(final_answer),
-                    sources_output: gr.HTML(sources_html),
-                    search_btn: gr.Button("Generating audio...", interactive=False),
-                    chat_history_display: history + [[query, final_answer]],
-                    audio_output: None
-                }
-                audio = generate_speech_with_gpu(final_answer, selected_voice)
-            except Exception as e:
-                print(f"Error in speech generation: {str(e)}")
-                audio = None
-        else:
-            audio = None
-        
         updated_history = history + [[query, final_answer]]
         yield {
             answer_output: gr.Markdown(final_answer),
             sources_output: gr.HTML(sources_html),
             search_btn: gr.Button("Search", interactive=True),
-            chat_history_display: updated_history,
-            audio_output: audio if audio is not None else gr.Audio(value=None)
+            chat_history_display: updated_history
         }
     except Exception as e:
         error_message = str(e)
@@ -242,10 +168,9 @@ def process_query(query, history, selected_voice='af'):
         
         yield {
             answer_output: gr.Markdown(f"Error: {error_message}"),
-            sources_output: gr.HTML(sources_html),
+            sources_output: gr.HTML(""),
             search_btn: gr.Button("Search", interactive=True),
-            chat_history_display: history + [[query, f"*Error: {error_message}*"]],
-            audio_output: None
+            chat_history_display: history + [[query, f"*Error: {error_message}*"]]
         }
 
 # ----------------------- Custom CSS for Bright UI ----------------------- #
@@ -260,7 +185,7 @@ css = """
 #header {
     text-align: center;
     padding: 2rem 0;
-    background: #f0f8ff;
+    background: #e3f2fd;
     border-radius: 12px;
     color: #333333;
     margin-bottom: 2rem;
@@ -385,20 +310,6 @@ css = """
     border: 1px solid #e0e0e0;
 }
 
-.voice-selector {
-    margin-top: 1rem;
-    background: #ffffff;
-    border-radius: 8px;
-    padding: 0.5rem;
-    border: 1px solid #cccccc;
-}
-
-.voice-selector select {
-    background: #ffffff !important;
-    color: #333333 !important;
-    border: 1px solid #cccccc !important;
-}
-
 footer {
     text-align: center;
     padding: 1rem 0;
@@ -413,7 +324,7 @@ with gr.Blocks(title="AI Search Assistant", css=css) as demo:
     
     with gr.Column(elem_id="header"):
         gr.Markdown("# 🔍 AI Search Assistant")
-        gr.Markdown("### Powered by DeepSeek & Real-time Web Results with Voice")
+        gr.Markdown("### Powered by DeepSeek & Real-time Web Results")
     
     with gr.Column(elem_classes="search-container"):
         with gr.Row(elem_classes="search-box"):
@@ -424,19 +335,12 @@ with gr.Blocks(title="AI Search Assistant", css=css) as demo:
                 container=False
             )
             search_btn = gr.Button("Search", variant="primary", scale=1)
-            voice_select = gr.Dropdown(
-                choices=list(VOICE_CHOICES.items()),
-                value='af',
-                label="Select Voice",
-                elem_classes="voice-selector"
-            )
         
         with gr.Row(elem_classes="results-container"):
             with gr.Column(scale=2):
                 with gr.Column(elem_classes="answer-box"):
                     answer_output = gr.Markdown()
-                    audio_output = gr.Audio(label="Voice Response")
-                with gr.Accordion("Chat History", open=False, elem_classes="accordion"):
+                with gr.Accordion("Chat History", open=False):
                     chat_history_display = gr.Chatbot(elem_classes="chat-history")
             with gr.Column(scale=1):
                 with gr.Column():
@@ -457,13 +361,13 @@ with gr.Blocks(title="AI Search Assistant", css=css) as demo:
     
     search_btn.click(
         fn=process_query,
-        inputs=[search_input, chat_history, voice_select],
-        outputs=[answer_output, sources_output, search_btn, chat_history_display, audio_output]
+        inputs=[search_input, chat_history],
+        outputs=[answer_output, sources_output, search_btn, chat_history_display]
     )
     search_input.submit(
         fn=process_query,
-        inputs=[search_input, chat_history, voice_select],
-        outputs=[answer_output, sources_output, search_btn, chat_history_display, audio_output]
+        inputs=[search_input, chat_history],
+        outputs=[answer_output, sources_output, search_btn, chat_history_display]
     )
 
 if __name__ == "__main__":