Update app.py

app.py

@@ -33,11 +33,12 @@ except Exception as e:
     print(f"Warning: Initial setup error: {str(e)}")
     print("Continuing with limited functionality...")
 
+
 # --- Initialization (Do this ONCE) ---
+
 model_name = "deepseek-ai/DeepSeek-R1-Distill-Llama-8B"
 tokenizer = AutoTokenizer.from_pretrained(model_name)
 tokenizer.pad_token = tokenizer.eos_token
-
 # Initialize DeepSeek model
 model = AutoModelForCausalLM.from_pretrained(
     model_name,
@@ -78,12 +79,12 @@ try:
         TTS_ENABLED = True
     else:
         print("Warning: Kokoro-82M directory not found. TTS disabled.")
-
 except Exception as e:
     print(f"Warning: Could not initialize Kokoro TTS: {str(e)}")
     TTS_ENABLED = False
 
 
+
 def get_web_results(query: str, max_results: int = 5) -> List[Dict[str, str]]:
     """Get web search results using DuckDuckGo"""
     try:
@@ -99,19 +100,27 @@ def get_web_results(query: str, max_results: int = 5) -> List[Dict[str, str]]:
         print(f"Error in web search: {e}")
         return []
 
+
 def format_prompt(query: str, context: List[Dict[str, str]]) -> str:
     """Format the prompt with web context"""
     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.
+
 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.
+
 Answer:"""
 
+
 def format_sources(web_results: List[Dict[str, str]]) -> str:
     """Format sources with more details"""
     if not web_results:
@@ -134,6 +143,7 @@ def format_sources(web_results: List[Dict[str, str]]) -> str:
     sources_html += "</div>"
     return sources_html
 
+
 @spaces.GPU(duration=30)
 def generate_answer(prompt: str) -> str:
     """Generate answer using the DeepSeek model"""
@@ -158,10 +168,11 @@ def generate_answer(prompt: str) -> str:
     )
     return tokenizer.decode(outputs[0], skip_special_tokens=True)
 
+
+
 @spaces.GPU(duration=30)
 def generate_speech_with_gpu(text: str, voice_name: str = 'af', tts_model = TTS_MODEL, voicepack = VOICEPACK) -> Tuple[int, np.ndarray] | None:
     """Generate speech from text using Kokoro TTS model."""
-
     if not TTS_ENABLED or tts_model is None:
         print("TTS is not enabled or model is not loaded.")
         return None
@@ -172,7 +183,6 @@ def generate_speech_with_gpu(text: str, voice_name: str = 'af', tts_model = TTS_
             device = 'cuda' if torch.cuda.is_available() else 'cpu'
             voicepack = torch.load(f'Kokoro-82M/voices/{voice_name}.pt', map_location=device, weights_only=True)
 
-
         # Clean the text
         clean_text = ' '.join([line for line in text.split('\n') if not line.startswith('#')])
         clean_text = clean_text.replace('[', '').replace(']', '').replace('*', '')
@@ -211,12 +221,14 @@ def generate_speech_with_gpu(text: str, voice_name: str = 'af', tts_model = TTS_
         else:
             return None
 
-
     except Exception as e:
         print(f"Error generating speech: {str(e)}")
         import traceback
         traceback.print_exc()
         return None
+
+
+
 def process_query(query: str, history: List[List[str]], selected_voice: str = 'af') -> Dict[str, Any]:
     """Process user query with streaming effect"""
     try:
@@ -228,6 +240,7 @@ def process_query(query: str, history: List[List[str]], selected_voice: str = 'a
         sources_html = format_sources(web_results)
 
         current_history = history + [[query, "*Searching...*"]]
+
         yield {
             answer_output: gr.Markdown("*Searching & Thinking...*"),
             sources_output: gr.HTML(sources_html),
@@ -244,6 +257,7 @@ def process_query(query: str, history: List[List[str]], selected_voice: str = 'a
         # Update history *before* TTS (important for correct display)
         updated_history = history + [[query, final_answer]]
 
+
         # Generate speech from the answer (only if enabled)
         if TTS_ENABLED:
             yield {  # Intermediate update before TTS
@@ -261,6 +275,8 @@ def process_query(query: str, history: List[List[str]], selected_voice: str = 'a
         else:
             audio = None
 
+
+
         yield {
             answer_output: gr.Markdown(final_answer),
             sources_output: gr.HTML(sources_html),
@@ -273,7 +289,6 @@ def process_query(query: str, history: List[List[str]], selected_voice: str = 'a
         error_message = str(e)
         if "GPU quota" in error_message:
             error_message = "⚠️ GPU quota exceeded.  Please try again later when the daily quota resets."
-
         yield {
             answer_output: gr.Markdown(f"Error: {error_message}"),
             sources_output: gr.HTML(sources_html), #Still show sources on error