Update app.py

app.py

@@ -8,7 +8,6 @@ from datetime import datetime
 import re
 from typing import List, Dict, Any, Optional, Tuple, Generator
 from dataclasses import dataclass
-from functools import lru_cache
 import logging
 import gradio as gr
 import torch
@@ -28,12 +27,13 @@ logger = logging.getLogger(__name__)
 # Global event for cancellation
 cancel_event = threading.Event()
 
-# Constants
+# Constants - Handle empty token properly
 ACCESS_TOKEN = os.environ.get('HF_TOKEN', '')
+if ACCESS_TOKEN == '':
+    ACCESS_TOKEN = None  # Convert empty string to None for proper handling
+
 PIPELINES = {}
 SEARCH_TIMEOUT_DEFAULT = 5.0
-MAX_RETRIES = 3
-RETRY_DELAY = 1
 
 # Data classes for better structure
 @dataclass
@@ -271,21 +271,34 @@ class ModelManager:
                 return cls._pipelines[model_name]
             
             repo = MODELS[model_name]["repo_id"]
-            tokenizer = AutoTokenizer.from_pretrained(repo, token=ACCESS_TOKEN)
+            
+            # Load tokenizer without token if not available
+            try:
+                tokenizer = AutoTokenizer.from_pretrained(
+                    repo, 
+                    token=ACCESS_TOKEN if ACCESS_TOKEN else None
+                )
+            except Exception as e:
+                logger.warning(f"Failed to load tokenizer with token, trying without: {e}")
+                tokenizer = AutoTokenizer.from_pretrained(repo)
             
             # Try different dtypes
             for dtype in (torch.bfloat16, torch.float16, torch.float32):
                 try:
-                    pipe = pipeline(
-                        task="text-generation",
-                        model=repo,
-                        tokenizer=tokenizer,
-                        trust_remote_code=True,
-                        dtype=dtype,
-                        device_map="auto",
-                        use_cache=True,
-                        token=ACCESS_TOKEN
-                    )
+                    pipe_kwargs = {
+                        'task': "text-generation",
+                        'model': repo,
+                        'tokenizer': tokenizer,
+                        'trust_remote_code': True,
+                        'dtype': dtype,
+                        'device_map': "auto",
+                        'use_cache': True,
+                    }
+                    # Only add token if it exists
+                    if ACCESS_TOKEN:
+                        pipe_kwargs['token'] = ACCESS_TOKEN
+                    
+                    pipe = pipeline(**pipe_kwargs)
                     cls._pipelines[model_name] = pipe
                     return pipe
                 except Exception as e:
@@ -293,14 +306,18 @@ class ModelManager:
                     continue
             
             # Final fallback
-            pipe = pipeline(
-                task="text-generation",
-                model=repo,
-                tokenizer=tokenizer,
-                trust_remote_code=True,
-                device_map="auto",
-                use_cache=True
-            )
+            pipe_kwargs = {
+                'task': "text-generation",
+                'model': repo,
+                'tokenizer': tokenizer,
+                'trust_remote_code': True,
+                'device_map': "auto",
+                'use_cache': True,
+            }
+            if ACCESS_TOKEN:
+                pipe_kwargs['token'] = ACCESS_TOKEN
+            
+            pipe = pipeline(**pipe_kwargs)
             cls._pipelines[model_name] = pipe
             return pipe
 
@@ -320,179 +337,355 @@ class PromptBuilder:
             )
         else:
             # Fallback for base LMs
-            prompt = f"{system_prompt.strip()}\n"
+            prompt = f"{ prompt = f"{system_prompt.stripsystem_prompt.strip()}\n"
+()}\n"
             for msg in history:
-                role = "User" if msg['role'] == 'user' else "Assistant"
-                prompt += f"{role}: {msg['content'].strip()}\n"
+                role            for msg in history:
+                role = "User" = "User" if msg['role'] == 'user if msg['role'] == 'user' else "Assistant"
+                prompt += f"{role}: {msg['content' else "Assistant"
+                prompt += f"{role}: {msg['content'].strip()}\'].strip()}\n"
             
-            if not prompt.strip().endswith("Assistant:"):
-                prompt += "Assistant: "
+           n"
+            
+            if not prompt.strip if not prompt.strip().endswith("Assistant:"):
+               ().endswith("Assistant:"):
+                prompt += " prompt += "Assistant: "
+            returnAssistant: "
             return prompt
     
     @staticmethod
-    def build_search_context(search_results: List[SearchResult], system_prompt: str, user_query: str) -> str:
+    def prompt
+    
+    @staticmethod
+    def build_search_context(search_results: List build_search_context(search_results: List[SearchResult], system_prompt: str[SearchResult], system_prompt: str, user_query: str) -> str, user_query: str) -> str:
+        """Build enriched prompt with search:
         """Build enriched prompt with search context"""
         if not search_results:
+ context"""
+        if not search_results:
             return system_prompt.strip()
         
-        formatted_results = "\n".join(f"[{i+1}] {r.format()}" for i, r in enumerate(search_results))
+            return system_prompt.strip()
+        
+        formatted_results = "\n".join        formatted_results = "\n".join(f"[{i(f"[{i+1}] {r.format()}" for i, r+1}] {r.format()}" for i, r in enumerate(search_results in enumerate(search_results))
+        
+        return f"""{))
         
         return f"""{system_prompt.strip()}
 
-# SEARCH CONTEXT (TRUSTED SOURCES ONLY)
-Below are search results. Treat them as the ONLY source of truth for answering.
+#system_prompt.strip()}
+
+# SEARCH CONTEXT (TRUSTED SEARCH CONTEXT (TRUSTED SOURCES ONLY)
+Below are search SOURCES ONLY)
+Below are search results. Treat them as the ONLY source results. Treat them as the ONLY source of truth for answering.
+{formatted of truth for answering.
 {formatted_results}
 
+RULES (VERY_results}
+
 RULES (VERY IMPORTANT):
-- Do NOT use outside knowledge. Do NOT guess or fill missing information.
-- If the answer is not clearly supported by the search results, say: "Not enough information in the provided sources."
-- Every factual statement must be directly supported by at least one citation [citation:X].
-- Do NOT add explanations, examples, or background that are not explicitly present in the sources.
-- Do NOT paraphrase beyond what is necessary for clarity.
+- Do NOT use outside IMPORTANT):
+- Do NOT use outside knowledge. Do NOT knowledge. Do NOT guess or fill missing information.
+- If the answer is guess or fill missing information.
+- If the answer is not clearly supported by not clearly supported by the search results, the search results, say: "Not say: "Not enough information in the provided sources."
+- Every factual statement must enough information in the provided sources."
+- Every factual statement must be directly supported by at least one citation be directly supported by at least one citation [citation:X].
+- Do NOT [citation:X].
+- Do NOT add explanations, examples, or background that add explanations, examples, or background that are not explicitly present in the sources.
+ are not explicitly present in the sources.
+- Do NOT paraphrase beyond what is- Do NOT paraphrase beyond what is necessary for clarity.
+- If sources conflict necessary for clarity.
 - If sources conflict, mention the conflict and cite both.
-- If multiple sources are used, distribute citations per sentence, not only at the end.
+, mention the conflict and cite both.
+- If multiple sources are used, distribute- If multiple sources are used, distribute citations per sentence, not only at the citations per sentence, not only at the end.
+
+CITATION RULES end.
 
 CITATION RULES:
 - Use inline citations like this: [citation:1]
+- If multiple sources support:
+- Use inline citations like this: [citation:1]
 - If multiple sources support a sentence: [citation:1][citation:3]
+ a sentence: [citation:1][citation:3]
 - Never place all citations only at the end.
 
 ANSWER POLICY:
 - Be concise and strictly grounded.
+- Never place all citations only at the end.
+
+ANSWER POLICY:
+- Be concise- No speculation, no assumptions, no "likely", no "probably".
+- and strictly grounded.
 - No speculation, no assumptions, no "likely", no "probably".
-- If the user requests a list, only include items explicitly found in sources.
-- If sources are insufficient, stop and ask for more data instead of guessing.
+- If the user requests a list, only include items explicitly If the user requests a list, only include items explicitly found in sources.
+ found in sources.
+- If sources are insufficient, stop and ask for more data- If sources are insufficient, stop and ask for more data instead of guessing.
 
+ instead of guessing.
+
+DATE CONTEXT:
 DATE CONTEXT:
-- Today is {datetime.now().strftime('%Y-%m-%d')} (use only for time reference, not for assumptions).
+- Today is {- Today is {datetime.now().strftime('%Y-%mdatetime.now().strftime('%Y-%m-%d')} (use only-%d')} (use only for time reference, not for assumptions).
+
+ for time reference, not for assumptions).
+
+USER QUESTION:
+{user_query}"""
 
 USER QUESTION:
 {user_query}"""
 
 class StreamProcessor:
+    """Handlesclass StreamProcessor:
     """Handles streaming token processing"""
     
+    @staticmethod streaming token processing"""
+    
     @staticmethod
-    def process_stream(streamer: TextIteratorStreamer, history: List[Dict]) -> Generator[Tuple[List[Dict], str], None, None]:
+    def process_stream(streamer
+    def process_stream(streamer: TextIteratorStreamer, history: TextIteratorStreamer, history: List[Dict]) -> Generator: List[Dict]) -> Generator[Tuple[List[Dict], str], None, None]:
         """Process streaming tokens and handle thinking tags"""
+        thought[Tuple[List[Dict], str], None, None]:
+        """Process streaming_buf = ''
+        answer_buf = ''
+        in_ tokens and handle thinking tags"""
         thought_buf = ''
         answer_buf = ''
-        in_thought = False
+        inthought = False
+        assistant_message_started = False
+        
+        for chunk in streamer:
+_thought = False
         assistant_message_started = False
         
         for chunk in streamer:
             if cancel_event.is_set():
-                if assistant_message_started and history and history[-1]['role'] == 'assistant':
-                    history[-1]['content'] += " [Generation Canceled]"
+                if assistant_message_start            if cancel_event.is_set():
+                if assistant_message_started and history and history[-1]['role'] == 'ed and history and history[-1]['role'] == 'assistant':
+                   assistant':
+                    history[-1]['content'] += " [Generation Cancel history[-1]['content'] += " [Generation Canceled]"
+                yield history, "Generation canceled by usered]"
                 yield history, "Generation canceled by user."
+               ."
                 break
             
             text = chunk
             
             # Handle thinking tags
-            if not in_thought and '<think>' in text:
-                in_thought = True
-                history.append({'role': 'assistant', 'content': '', 'metadata': {'title': '💭 Thought'}})
+            break
+            
+            text = chunk
+            
+            # Handle thinking tags
+            if not in_ if not in_thought and '<think>' in text:
+thought and '<think>' in text:
+                in_thought                in_thought = True
+                history.append({'role': 'assistant = True
+                history.append({'role': 'assistant', 'content':', 'content': '', 'metadata': {'title': '💭 Thought'}})
+                assistant_message_started = '', 'metadata': {'title': '💭 Thought'}})
                 assistant_message_started = True
                 after = text.split('<think>', 1)[1]
                 thought_buf += after
                 
+                if '</think>' in thought_buf:
+                    before, after2 = thought_buf.split('</think>', 1 True
+                after = text.split('<think>', 1)[1]
+                thought_buf += after
+                
                 if '</think>' in thought_buf:
                     before, after2 = thought_buf.split('</think>', 1)
                     history[-1]['content'] = before.strip()
+                    in)
+                    history[-1]['content'] = before.strip()
                     in_thought = False
+                    answer_buf =_thought = False
                     answer_buf = after2
-                    history.append({'role': 'assistant', 'content': answer_buf})
+                    history.append({'role after2
+                    history.append({'role': 'assistant', 'content':': 'assistant', 'content': answer_buf})
+                answer_buf})
                 else:
+                    history[-1]['content'] = thought else:
                     history[-1]['content'] = thought_buf
                 yield history, ""
+_buf
+                yield history, ""
                 continue
             
+            if in_thought                continue
+            
             if in_thought:
                 thought_buf += text
-                if '</think>' in thought_buf:
-                    before, after2 = thought_buf.split('</think>', 1)
+                if '</think>' in thought:
+                thought_buf += text
+                if '</think>' in thought_b_buf:
+                    beforeuf:
+                    before, after, after2 = thought_buf.split2 = thought_buf.split('</think>('</think>', 1)
+                    history[-1', 1)
                     history[-1]['content'] = before.strip()
+                   ]['content'] = before.strip()
                     in_thought = False
+                    answer in_thought = False
                     answer_buf = after2
-                    history.append({'role': 'assistant', 'content': answer_buf})
+                    history_buf = after2
+                    history.append({'role': 'assistant',.append({'role': 'assistant', 'content': answer_buf})
                 else:
+ 'content': answer_buf})
+                                   history else:
                     history[-1]['content'] = thought_buf
+                yield[-1]['content'] = thought_buf
                 yield history, ""
+                history, ""
                 continue
             
+            # continue
+            
             # Stream answer
-            if not assistant_message_started:
-                history.append({'role': 'assistant', 'content': ''})
-                assistant_message_started = True
+ Stream answer
+            if not assistant            if not assistant_message_started:
+_message_started:
+                history.append({'role': 'assistant', 'content                history.append({'role': 'assistant', 'content': ''})
+               ': ''})
+                assistant_message_started assistant_message_started = True
+            
+ = True
             
             answer_buf += text
             history[-1]['content'] = answer_buf.strip()
+                       answer_buf += text
+            history[-1]['content'] yield history, ""
+
+# Main chat = answer_buf.strip()
             yield history, ""
 
 # Main chat function
+def chat_response(
+    user function
 def chat_response(
     user_msg: str,
+    chat_history:_msg: str,
     chat_history: List[Dict],
+    system_prompt List[Dict],
     system_prompt: str,
+    enable_search: bool: str,
     enable_search: bool,
     max_results: int,
+   ,
+    max_results: int,
     max_chars: int,
+    model max_chars: int,
     model_name: str,
+    max_tokens_name: str,
     max_tokens: int,
     temperature: float,
     top_k: int,
+    top: int,
+    temperature: float,
+    top_k: int,
     top_p: float,
     repeat_penalty: float,
+_p: float,
+    repeat_penalty: float,
     search_timeout: float
+) -> Generator[Tuple    search_timeout: float
 ) -> Generator[Tuple[List[Dict], str], None, None]:
+    """[List[Dict], str], None, None]:
     """Generate streaming chat responses with search integration"""
     
+    cancel_eventGenerate streaming chat responses with search integration"""
+    
     cancel_event.clear()
     history = list(chat_history or [])
     history.append({'role': 'user', 'content': user_msg})
     
+    # Perform search.clear()
+    history = list(chat_history or [])
+    history.append({'role': 'user', 'content': user_msg})
+    
     # Perform search if enabled
+    if enabled
     search_results: List[SearchResult] = []
+    search_debug = " search_results: List[SearchResult] = []
     search_debug = "Web search disabled."
     
+    if enableWeb search disabled."
+    
     if enable_search:
-        search_debug = "🔍 Searching across multiple engines..."
+        search_debug = "_search:
+        search_debug = "🔍 Searching across multiple engines🔍 Searching across multiple engines..."
+        try:
+            search_results = SearchManager..."
         try:
             search_results = SearchManager.search(
                 user_msg, 
+               .search(
+                user_msg, 
                 int(max_results), 
+                int(max int(max_results), 
                 int(max_chars), 
                 float(search_timeout)
             )
             
-            if search_results:
-                search_debug = f"✅ Search completed - Found {len(search_results)} results\n\n" + "\n".join(
-                    f"- {r.format(int(max_chars))}" for r in search_results
+           _chars), 
+                float(search_timeout)
+            )
+            
+            if search if search_results:
+                search_results:
+                search_debug = f"✅ Search completed - Found {len_debug = f"✅ Search completed - Found {len(search_results)} results(search_results)} results\n\n" + "\n".join(
+                    f"-\n\n" + "\n".join(
+                    f"- {r.format(int(max_chars))}" for r {r.format(int(max_chars))}" for r in search_results
+                )
+            else:
+                search_debug = "❌ No search results found. Check internet connection or try again in search_results
                 )
             else:
                 search_debug = "❌ No search results found. Check internet connection or try again."
+        except Exception as e:
+            search_debug = f"❌ Search failed: {."
         except Exception as e:
             search_debug = f"❌ Search failed: {str(e)}"
+            logger.error(f"Search error:str(e)}"
             logger.error(f"Search error: {e}")
     
+    try:
+        {e}")
+    
     try:
         # Build prompt
         if enable_search and search_results:
+ # Build prompt
+        if enable_search and search_results:
             enriched_prompt = PromptBuilder.build_search_context(
                 search_results, 
-                system_prompt, 
+                           enriched_prompt = PromptBuilder.build_search_context(
+                search_results, 
+                system_prompt, system_prompt, 
+                user_msg
+            )
+        else:
+            enriched 
                 user_msg
             )
         else:
             enriched_prompt = system_prompt.strip()
         
+        # Load_prompt = system_prompt.strip()
+        
         # Load model
+        pipe = ModelManager.load_pipeline(model_name model
         pipe = ModelManager.load_pipeline(model_name)
         
+        #)
+        
         # Format prompt
         prompt = PromptBuilder.format_conversation(history, enriched_prompt, pipe.tokenizer)
-        prompt_debug = f"\n\n--- Prompt Preview ---\n```\n{prompt[:500]}...\n```" if len(prompt) > 500 else f"\n\n--- Prompt Preview ---\n```\n{prompt}\n```"
+        Format prompt
+        prompt = PromptBuilder.format_conversation(history, enriched_prompt, pipe.tokenizer)
+        prompt_debug = f"\n\n--- Prompt Preview ---\n``` prompt_debug = f"\n\n--- Prompt Preview ---\n```\n{prompt[:500]}...\n```" if len(prompt) > 500 else f"\n\n--- Prompt Preview ---\n```\n{prompt}\n```"
+        
+        # Configure generation
+        config = GenerationConfig(
+            max_tokens\n{prompt[:500]}...\n```" if len(prompt) > 500 else f"\n\n--- Prompt Preview ---\n```\n{prompt}\n```"
         
         # Configure generation
         config = GenerationConfig(
@@ -500,71 +693,141 @@ def chat_response(
             temperature=temperature,
             top_k=top_k,
             top_p=top_p,
-            repetition_penalty=repeat_penalty
+           =max_tokens,
+            temperature=temperature,
+            top_k=top_k,
+            top_p=top_p,
+            repetition_penalty= repetition_penalty=repeat_penalty
+        )
+        
+        # Setuprepeat_penalty
         )
         
         # Setup streamer
         streamer = TextIteratorStreamer(
-            pipe.tokenizer,
+ streamer
+        streamer = TextIteratorStreamer(
+            pipe.tokenizer            pipe.tokenizer,
+            skip_prompt=True,
+            skip_special_t,
             skip_prompt=True,
             skip_special_tokens=True
         )
         
         # Start generation in background thread
-        gen_kwargs = config.to_dict()
-        gen_kwargs['streamer'] = streamer
-        gen_kwargs['return_full_text'] = False
+        genokens=True
+        )
+        
+        # Start generation in background thread
+        gen_kwargs =_kwargs = config.to_dict()
+ config.to_dict()
+        gen_kwargs['streamer'] = streamer        gen_kwargs['streamer'] = streamer
+        gen_k
+        gen_kwargs['returnwargs['return_full_text'] = False
+        
+        gen_thread = threading.Thread_full_text'] = False
         
         gen_thread = threading.Thread(
+            target=(
             target=pipe,
+            argspipe,
             args=(prompt,),
+            kwargs=gen_kwargs=(prompt,),
             kwargs=gen_kwargs
         )
+       
+        )
         gen_thread.start()
+ gen_thread.start()
         
+        # Yield        
         # Yield initial state
+        yield history, search_debug initial state
         yield history, search_debug
         
         # Process stream
-        for history_update, debug_update in StreamProcessor.process_stream(streamer, history):
+
+        
+        # Process stream
+        for history_update, debug_update in        for history_update, debug_update in StreamProcessor.process_stream(streamer, history):
+            yield history_update StreamProcessor.process_stream(streamer, history):
             yield history_update, debug_update
         
+        # Wait for, debug_update
+        
         # Wait for completion
         gen_thread.join(timeout=5.0)
-        yield history, search_debug + prompt_debug
+        yield history completion
+        gen_thread.join(timeout=5.0)
+, search_debug + prompt_de        yield history, search_debug + prompt_debug
+        
+    exceptbug
         
     except GeneratorExit:
         logger.info("Generation cancelled by user")
+ GeneratorExit:
+        logger.info("Generation cancelled by user")
         return
-    except Exception as e:
+           return
+    except Exception as e except Exception as e:
+        logger.error:
         logger.error(f"Generation error: {e}")
-        history.append({'role': 'assistant', 'content': f"Error: {str(e)}"})
+        history.append({'role': 'ass(f"Generation error: {e}")
+        history.append({'role': 'assistant', 'content': f"Erroristant', 'content': f"Error: {str(e)}"})
+       : {str(e)}"})
         yield history, search_debug
+    yield history, search_debug
     finally:
         gc.collect()
 
+# finally:
+        gc.collect()
+
 # Utility functions
+def get_model_size(model_name: str) -> Utility functions
 def get_model_size(model_name: str) -> float:
+    """Get model size in billions float:
     """Get model size in billions of parameters"""
-    return MODELS.get(model_name, {}).get("params_b", 4.0)
+    return MODELS.get(model_name of parameters"""
+    return MODELS.get(model_name, {}).get("params_b",, {}).get("params_b", 4.0)
+
+def get_d 4.0)
 
 def get_duration_estimate(
     model_name: str,
     enable_search: bool,
     max_tokens: int,
     search_timeout: float
+) -> float:
+    """Calculate estimated GPU duration"""
+    model_size = get_modeluration_estimate(
+    model_name: str,
+    enable_search: bool,
+    max_tokens: int,
+    search_timeout: float
 ) -> float:
     """Calculate estimated GPU duration"""
     model_size = get_model_size(model_name)
     use_aot = model_size >= 2
     
+    base_duration = 20 if not use_aot else 40
+    token_duration = max_tokens * 0.005_size(model_name)
+    use_aot = model_size >= 2
+    
     base_duration = 20 if not use_aot else 40
     token_duration = max_tokens * 0.005
     search_duration = 10 if enable_search else 0
+    aot_compilation = 20
+    search_duration = 10 if enable_search else 0
     aot_compilation = 20 if use_aot else 0
     
     return base_duration + token_duration + search_duration + aot_compilation
 
+def update_duration_estimate(
+    if use_aot else 0
+    
+    return base_duration + token_duration + search_duration + aot_compilation
+
 def update_duration_estimate(
     model_name: str,
     enable_search: bool,
@@ -580,32 +843,76 @@ def update_duration_estimate(
         
         return f"""⏱️ **Estimated GPU Time: {duration:.1f} seconds**
 
+ model_name: str,
+    enable_search: bool,
+    max_results: int,
+    max_chars: int,
+    max_tokens: int,
+    search_timeout: float
+) -> str:
+    """Format duration estimate for display"""
+    try:
+        duration = get_duration_estimate(model_name, enable_search, max_tokens, search_timeout)
+        model_size = get_model_size(model_name)
+        
+        return f"""⏱️ **Estimated GPU Time: {duration:.1f} seconds**
+
 📊 **Model Size:** {model_size:.1f}B parameters
-🔍 **Web Search:** {'Enabled (Multi-Engine)' if enable_search else 'Disabled'}"""
+📊 **Model Size:** {model_size:.1f}B parameters
+🔍 **Web Search:** {'Enabled (Multi-Engine)' if enable_search else🔍 **Web Search:** {'Enabled (Multi-Engine)' if enable_search else 'Disabled'}"""
+    except Exception as 'Disabled'}"""
     except Exception as e:
+        logger.error(f"Error e:
         logger.error(f"Error calculating estimate: {e}")
-        return f"⚠️ Error calculating estimate: {e}"
+        return calculating estimate: {e}")
+        return f"⚠️ Error calculating estimate: f"⚠️ Error calculating estimate: {e}"
+
+def update_default_p {e}"
 
-def update_default_prompt(enable_search: bool) -> str:
+def update_default_prompt(enable_search: bool) ->rompt(enable_search: bool) -> str:
+    """Generate default system str:
     """Generate default system prompt"""
     return "You are a helpful assistant."
 
 # ------------------------------
-# Gradio UI
+# Grad prompt"""
+    return "You are a helpful assistant."
+
+# ------------------------------
+io UI
+# ------------------------------
+with gr.Blocks(
+    title="# Gradio UI
 # ------------------------------
 with gr.Blocks(
     title="LLM Inference",
+    theme=gr.themes.SoftLLM Inference",
     theme=gr.themes.Soft(
         primary_hue="blue",
         secondary_hue="blue",
         neutral_hue="slate",
+       (
+        primary_hue="blue",
+        secondary_hue="blue",
+        neutral_hue="slate",
         radius_size="lg",
+        font=[gr.themes.GoogleFont("Syne"), "Arial", "s radius_size="lg",
         font=[gr.themes.GoogleFont("Syne"), "Arial", "sans-serif"]
     ),
     css="""
-        .duration-estimate { background: linear-gradient(135deg, #667eea15 0%, #764ba215 100%); border-left: 4px solid #667eea; padding: 12px; border-radius: 8px; margin: 16px 0; }
-        .chatbot { border-radius: 12px; box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.1); }
+        .durationans-serif"]
+    ),
+    css="""
+        .duration-estimate { background: linear-gradient(135deg, #-estimate { background: linear-gradient(667eea15 0%, #764ba215 135deg, #667eea15 0%, #764ba215 100%); border-left100%); border-left: 4px solid #667eea; padding: 12px; border-radius: 8px; margin: 16px 0; }
+        .chatbot: 4px solid #667eea; padding: 12px; border-radius: 8px; margin: 16px 0; }
+ { border-radius: 12px; box-shadow: 0 4px        .chatbot { border-radius: 12px; box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.1 6px -1px rgba(0, 0, 0,); }
         button.primary { font-weight: 600; }
+        0.1); }
+        button.primary { font-weight: 600 .gradio-accordion { margin-bottom: 12px; }
+    """
+) as demo:
+    # Header
+    gr.Mark; }
         .gradio-accordion { margin-bottom: 12px; }
     """
 ) as demo:
@@ -614,40 +921,79 @@ with gr.Blocks(
     # 🧠 LLM Inference with Multi-Engine Search
     """)
     
+   down("""
+    # 🧠 LLM Inference with Multi-Engine Search
+    """)
+    
     with gr.Row():
         # Left Panel - Configuration
+        with gr.Column(scale=3):
+            # Core with gr.Row():
+        # Left Panel - Configuration
         with gr.Column(scale=3):
             # Core Settings (Always Visible)
+            with gr.Group():
+                gr.Markdown(" Settings (Always Visible)
             with gr.Group():
                 gr.Markdown("### ⚙️ Core Settings")
+                model_dd = gr.Drop### ⚙️ Core Settings")
                 model_dd = gr.Dropdown(
                     label="🤖 Model",
                     choices=list(MODELS.keys()),
                     value="Qwen3-1.7B",
                     info="Select the language model to use"
                 )
+                search_chk = gr.Checkbox(
+                    label="🔍 Enabledown(
+                    label="🤖 Model",
+                    choices=list(MODELS.keys()),
+                    value="Qwen3-1.7B",
+                    info="Select the language model to use"
+                )
                 search_chk = gr.Checkbox(
                     label="🔍 Enable Web Search",
                     value=False,
+                    info="Search across Google Web Search",
+                    value=False,
                     info="Search across Google, DuckDuckGo, and Bing (no API required)"
                 )
+                sys_prompt = gr.Textbox(label="📝, DuckDuckGo, and Bing (no API required)"
+                )
                 sys_prompt = gr.Textbox(label="📝 System Prompt", lines=3, value=update_default_prompt(False), placeholder="Define the assistant's behavior and personality...")
             
             # Duration Estimate
             duration_display = gr.Markdown(
-                value=update_duration_estimate("Qwen3-1.7B", False, 4, 50, 1024, 5.0),
+                value=update_duration_estimate(" System Prompt", lines=3, value=update_default_prompt(False), placeholder="Define the assistant's behavior and personality...")
+            
+            # Duration Estimate
+            duration_display = gr.Markdown(
+                value=update_duration_estimate("Qwen3-Qwen3-1.7B", False, 4, 50, 1024, 5.0),
+                elem_classes="duration-estimate"
+            )
+1.7B", False, 4, 50, 1024, 5.0),
                 elem_classes="duration-estimate"
             )
             
+            # Advanced Settings (Collapsible)
+            with gr.Accordion("🎛️ Advanced Generation Parameters", open=False):
+                max_tok = gr.Slider(
+                    64, 16384, value            
             # Advanced Settings (Collapsible)
             with gr.Accordion("🎛️ Advanced Generation Parameters", open=False):
                 max_tok = gr.Slider(
                     64, 16384, value=1024, step=32,
+                    label="Max=1024, step=32,
                     label="Max Tokens",
                     info="Maximum length of generated response"
                 )
                 temp = gr.Slider(
-                    0.1, 2.0, value=0.7, step=0.1,
+                    0.1, 2.0, value=0. Tokens",
+                    info="Maximum length of generated response"
+                )
+                temp = gr.Slider(
+                    0.1, 2.0,7, step=0.1,
+                    label="Temperature",
+                    info="Higher = more creative, Lower value=0.7, step=0.1,
                     label="Temperature",
                     info="Higher = more creative, Lower = more focused"
                 )
@@ -658,10 +1004,24 @@ with gr.Blocks(
                         info="Number of top tokens to consider"
                     )
                     p = gr.Slider(
+ = more focused"
+                )
+                with gr.Row():
+                    k = gr.Slider(
+                        1, 100, value=40, step=1,
+                        label="Top-K",
+                        info="Number of top tokens to consider"
+                    )
+                    p = gr                        0.1, 1.0, value=0.9, step=0.05,
+                        label="Top-P",
+.Slider(
                         0.1, 1.0, value=0.9, step=0.05,
                         label="Top-P",
                         info="Nucleus sampling threshold"
                     )
+                rp = gr.Slider(
+                    1.0, 2.0, value=1.2, step=0.                        info="Nucleus sampling threshold"
+                    )
                 rp = gr.Slider(
                     1.0, 2.0, value=1.2, step=0.1,
                     label="Repetition Penalty",
@@ -669,23 +1029,48 @@ with gr.Blocks(
                 )
             
             # Web Search Settings (Collapsible)
-            with gr.Accordion("🌐 Web Search Settings", open=False, visible=False) as search_settings:
+            with gr.Acc1,
+                    label="Repetition Penalty",
+                    info="Penalize repeated tokens"
+                )
+            
+            # Web Search Settings (Collapsible)
+            with gr.Accordion("ordion("🌐 Web Search Settings", open=False, visible=False) as search_settings:
+                mr = gr.Number(
+                    value=4, precision=0,
+                    label="Max🌐 Web Search Settings", open=False, visible=False) as search_settings:
                 mr = gr.Number(
                     value=4, precision=0,
                     label="Max Results",
+                    info="Number of search results to Results",
                     info="Number of search results to retrieve"
                 )
                 mc = gr.Number(
                     value=50, precision=0,
                     label="Max Chars/Result",
-                    info="Character limit per search result"
+                    info="Character limit per search retrieve"
+                )
+                mc = gr.Number(
+                    value=50, precision=0,
+                    label="Max Chars/Result",
+                    info=" result"
+                )
+                st = grCharacter limit per search result"
                 )
                 st = gr.Slider(
+                    minimum=0.0, maximum=30.0, step=0.5, value=.Slider(
                     minimum=0.0, maximum=30.0, step=0.5, value=5.0,
                     label="Search Timeout (s)",
+                    info5.0,
+                    label="Search Timeout (s)",
                     info="Maximum time to wait for search results"
                 )
-                gr.Markdown("""
+                gr="Maximum time to wait for search results"
+                )
+                gr.Markdown("".Markdown("""
+                ⚠️ **Search Engines:**
+                - Google (primary)
+                - DuckD"
                 ⚠️ **Search Engines:**
                 - Google (primary)
                 - DuckDuckGo (fallback)
@@ -696,145 +1081,301 @@ with gr.Blocks(
             
             # Actions
             with gr.Row():
-                clr = gr.Button("🗑️ Clear Chat", variant="secondary", scale=1)
+                clr = gr.Button("uckGo (fallback)
+                - Bing (fallback)
+                
+                SafeSearch is **OFF** for comprehensive results.
+                """)
+            
+            # Actions
+            with gr.Row():
+                clr =🗑️ Clear Chat", variant="secondary", scale=1 gr.Button("🗑️ Clear Chat", variant="secondary", scale=1)
+        
+        # Right Panel - Chat)
         
         # Right Panel - Chat Interface
         with gr.Column(scale=7):
             chat = gr.Chatbot(
                 type="messages",
-                height=600,
+ Interface
+        with gr.Column(scale=7):
+            chat = gr.Chatbot(
+                type="messages                height=600,
                 label="💬 Conversation",
                 show_copy_button=True,
                 avatar_images=(
-                    "data:image/svg+xml,%3Csvg xmlns='http://www.w3.org/2000/svg' width='40' height='40'%3E%3Crect width='40' height='40' rx='20' fill='%23f093fb'/%3E%3Ctext x='20' y='28' text-anchor='middle' font-size='20' fill='white' font-family='Arial'%3E👤%3C/text%3E%3C/svg%3E",
-                    "data:image/svg+xml,%3Csvg xmlns='http://www.w3.org/2000/svg' width='40' height='40'%3E%3Crect width='40' height='40' rx='20' fill='%23667eea'/%3E%3Ctext x='20' y='28' text-anchor='middle' font-size='20' fill='white' font-family='Arial'%3E🤖%3C/text%3E%3C/svg%3E"
+                   ",
+                height=600,
+                label="💬 Conversation",
+                show_copy_button=True,
+                avatar "data:image/svg+xml,%3Csvg xmlns='http://www.w3.org/2000/svg' width='40' height='40'%3E%3C_images=(
+                    "data:image/svg+xml,%3Csvg xmlns='http://www.w3.org/2000/svg' width='40' height='40'%3E%3Crect width='40rect width='40' height='40' rx='20' fill='%' height='40' rx='20' fill='%23f093fb23f093fb'/%3E%3Ctext x='20' y='28' text-anchor='middle' font-size='20''/%3E%3Ctext x='20' y='28' text-anchor='middle' font-size='20' fill='white' font-family='Arial'%3E👤%3C/text%3E%3 fill='white' font-family='Arial'%3E👤%3C/text%3E%3C/svg%3E",
+                    "data:image/svg+xml,%3Csvg xmlns='http://www.w3.org/2000/svg' width='40' height='40'%3E%3Crect width='40' height='40' rx='20' fill='%23667eea'/%3E%3Ctext x='20' y='28' text-anchor='middle' font-size='20' fill='white'C/svg%3E",
+                    "data:image/svg+xml,%3Csvg xmlns='http://www.w3.org/2000/svg' width='40' height='40'%3E%3Crect width='40' height='40' rx='20' fill='%23667eea'/%3E%3Ctext x='20' y='28' text-anchor='middle' font-size='20' fill=' font-family='Arial'%3E🤖%3C/text%3E%3C/svg%3E"
+                ),
+                bubble_fullwhite' font-family='Arial'%3E🤖%3C/text%3E%3C/svg%3E"
                 ),
                 bubble_full_width=False,
                 render_markdown=True,
+                sanit_width=False,
+                render_markdown=True,
                 sanitize_html=False
+ize_html=False
             )
             
+            # Input Area
+            with            )
+            
             # Input Area
             with gr.Row():
+                txt gr.Row():
                 txt = gr.Textbox(
-                    placeholder="💭 Type your message here... (Press Enter to send)",
+                    placeholder="💭 Type your = gr.Textbox(
+                    placeholder="💭 Type your message here... ( message here... (Press Enter to sendPress Enter to send)",
+                    scale)",
                     scale=9,
                     container=False,
+                   =9,
+                    container=False,
                     show_label=False,
+                    lines=1 show_label=False,
                     lines=1,
                     max_lines=5
+               ,
+                    max_lines=5
                 )
+                with gr.Column(scale= )
                 with gr.Column(scale=1, min_width=120):
-                    submit_btn = gr.Button("📤 Send", variant="primary", size="lg")
+                   1, min_width=120):
+                    submit_btn = gr.Button("📤 Send", variant="primary", size submit_btn = gr.Button("📤 Send", variant="primary", size="lg")
+                    cancel_btn = gr.Button("⏹️ Stop", variant="stop", visible=False="lg")
                     cancel_btn = gr.Button("⏹️ Stop", variant="stop", visible=False, size="lg")
             
+            #, size="lg")
+            
             # Example Prompts
+            gr.Examples(
+                examples=[
+                    ["Explain Example Prompts
             gr.Examples(
                 examples=[
                     ["Explain quantum computing in simple terms"],
+                    ["Write a Python function to quantum computing in simple terms"],
                     ["Write a Python function to calculate fibonacci numbers"],
-                    ["What are the latest developments in AI? (Enable web search)"],
+                    ["What are the calculate fibonacci numbers"],
+                    ["What are the latest developments in AI? (Enable latest developments in AI? (Enable web search)"],
                     ["Tell me a creative story about a time traveler"],
+ web search)"],
+                    ["Tell me a creative story about a time traveler"],
+                    ["Help me debug this code: def add(a,b): return a+b+1"]
                     ["Help me debug this code: def add(a,b): return a+b+1"]
                 ],
+                inputs                ],
                 inputs=txt,
                 label="💡 Example Prompts"
+=txt,
+                label="💡 Example Prompts"
             )
             
             # Debug/Status Info (Collapsible)
             with gr.Accordion("🔍 Debug Info", open=False):
                 dbg = gr.Markdown()
     
+    # Footer
+    gr.Markdown("""
+    ---
+    💡 **Tips            )
+            
+            # Debug/Status Info (Collapsible)
+            with gr.Accordion("🔍 Debug Info", open=False):
+                dbg = gr.Markdown()
+    
     # Footer
     gr.Markdown("""
     ---
     💡 **Tips:** 
     - Use **Advanced Parameters** to fine-tune creativity and response length
-    - Enable **Web Search** for real-time information (uses multiple search engines)
+    - Enable **Web Search:** 
+    - Use **Advanced Parameters** to fine-tune creativity and response length
+    - Enable **Web Search** for real-time information (uses** for real-time information (uses multiple search engines)
+    - SafeSearch is **OFF** multiple search engines)
     - SafeSearch is **OFF** for comprehensive results
-    - Try different **models** for various tasks (reasoning, coding, general chat)
-    - Click the **Copy** button on responses to save them to your clipboard
+    - Try different ** for comprehensive results
+    - Try different **models** for various tasks (reasonmodels** for various tasks (reasoning, coding, general chat)
+   ing, coding, general chat)
+    - Click the **Copy** button on - Click the **Copy** button on responses to save them to your clipboard
+ responses to save them to your clipboard
     """, elem_classes="footer")
 
     # --- Event Listeners ---
 
-    # Group all inputs for cleaner event handling
-    chat_inputs = [txt, chat, sys_prompt, search_chk, mr, mc, model_dd, max_tok, temp, k, p, rp, st]
-    # Group all UI components that can be updated.
-    ui_components = [chat, dbg, txt, submit_btn, cancel_btn]
+       """, elem_classes="footer")
+
+    # --- Event Listeners ---
+
+    # Group # Group all inputs for cleaner event handling
+    chat_inputs = [txt, chat, sys_p all inputs for cleaner event handling
+    chat_inputs = [txt, chat, sys_prompt, search_chk,rompt, search_chk, mr, mc, model_dd, max mr, mc, model_dd, max_tok, temp, k, p_tok, temp, k, p, rp, st]
+    #, rp, st]
+    # Group all UI components that can Group all UI components that can be updated.
+    ui_components = be updated.
+    ui_components = [chat, dbg, txt, submit [chat, dbg, txt, submit_btn, cancel_btn, cancel_btn]
+
+    def submit_and_manage_ui(user_btn]
 
-    def submit_and_manage_ui(user_msg, chat_history, *args):
+    def submit_and_manage_ui(user_msg, chat_history_msg, chat_history, *args):
+, *args):
         """
-        Orchestrator function that manages UI state and calls the backend chat function.
+        Orche        """
+        Orchestrator function that manages UI state andstrator function that manages UI state and calls the backend chat function.
+        """
+ calls the backend chat function.
         """
         if not user_msg.strip():
+                   if not user_msg.strip():
             yield {}
             return
 
+        # yield {}
+            return
+
         # Update UI to "generating" state
         yield {
-            txt: gr.update(value="", interactive=False),
+            txt: gr.update(value="", interactive Update UI to "generating" state
+        yield {
+            txt: gr.update(value="",=False),
+            submit_btn: gr.update(inter interactive=False),
             submit_btn: gr.update(interactive=False),
+            cancel_active=False),
             cancel_btn: gr.update(visible=True),
         }
 
-        cancelled = False
+btn: gr.update(visible=True),
+        }
+
+        cancelled = False        cancelled = False
+        try:
+            backend_args = [user_msg,
         try:
-            backend_args = [user_msg, chat_history] + list(args)
+            backend_args = [user_msg, chat_history] + chat_history] + list(args)
+            for response_chunk in chat_response(*backend_args):
+                yield {
+                    chat list(args)
             for response_chunk in chat_response(*backend_args):
                 yield {
                     chat: response_chunk[0],
+                    dbg: response: response_chunk[0],
                     dbg: response_chunk[1],
                 }
         except GeneratorExit:
+_chunk[1],
+                }
+        except GeneratorExit:
             cancelled = True
+            print("Generation cancelled by user            cancelled = True
             print("Generation cancelled by user.")
             raise
         except Exception as e:
-            print(f"An error occurred during generation: {e}")
+            print.")
+            raise
+        except Exception as e:
+            print(f"An error occurred during generation: {e(f"An error occurred during generation: {e}")
+            error_history = (}")
             error_history = (chat_history or []) + [
+               chat_history or []) + [
                 {'role': 'user', 'content': user_msg},
+                {'role': 'assistant', 'content': f {'role': 'user', 'content': user_msg},
                 {'role': 'assistant', 'content': f"**An error occurred:** {str(e)}"}
             ]
+            yield {"**An error occurred:** {str(e)}"}
+            ]
             yield {chat: error_history}
         finally:
+chat: error_history}
+        finally:
             if not cancelled:
+                print("            if not cancelled:
                 print("Resetting UI state.")
                 yield {
-                    txt: gr.update(interactive=True),
+Resetting UI state.")
+                yield {
+                    txt: gr.update(inter                    txt: gr.update(interactive=True),
+                    submit_btn: gractive=True),
                     submit_btn: gr.update(interactive=True),
-                    cancel_btn: gr.update(visible=False),
+.update(interactive=True),
+                    cancel_btn: gr.update(visible=False                    cancel_btn: gr.update(visible=False),
+                }
+
+    def set_c),
                 }
 
     def set_cancel_flag():
+        """Called by the cancel button, sets the global eventancel_flag():
         """Called by the cancel button, sets the global event."""
         cancel_event.set()
+        print("Cancellation signal."""
+        cancel_event.set()
         print("Cancellation signal sent.")
     
+    def reset_ui_after_cancel sent.")
+    
     def reset_ui_after_cancel():
         """Reset UI components after cancellation."""
         cancel_event.clear()
         print("UI reset after cancellation.")
         return {
             txt: gr.update(interactive=True),
-            submit_btn: gr.update(interactive=True),
+           ():
+        """Reset UI components after cancellation."""
+        cancel_event.clear()
+        print("UI reset after cancellation.")
+        return {
+            txt: gr.update(interactive=True),
+            submit_btn: submit_btn: gr.update(interactive=True),
+            cancel_btn: gr.update(visible gr.update(interactive=True),
             cancel_btn: gr.update(visible=False),
         }
 
-    # Event for submitting text via Enter key or Submit button
+    # Event for=False),
+        }
+
+    # Event for submitting text via Enter key or Submit submitting text via Enter key or Submit button
+    submit_event = txt button
     submit_event = txt.submit(
-        fn=submit_and_manage_ui,
+        fn=submit_and.submit(
+        fn=submit_and_manage_ui_manage_ui,
+        inputs=chat_inputs,
+        outputs=ui,
         inputs=chat_inputs,
         outputs=ui_components,
     )
+    submit_btn.click(
+        fn=submit_and_components,
+    )
     submit_btn.click(
         fn=submit_and_manage_ui,
-        inputs=chat_inputs,
+        inputs_manage_ui,
+        inputs==chat_inputs,
+        outputs=ui_components,
+    )
+
+    # Eventchat_inputs,
         outputs=ui_components,
     )
 
     # Event for the "Cancel" button.
-    cancel_btn.click(
+    for the "Cancel" button.
+    cancel_btn.click cancel_btn.click(
+        fn=set_cancel_flag,
+        cancels=[submit_event]
+    ).then(
+        fn=reset_ui_after_cancel,
+        outputs=ui_components
+    )
+
+    # Listeners for updating the duration estimate
+    duration_inputs = [model_dd, search_chk, mr, mc, max_tok, st]
+    for(
         fn=set_cancel_flag,
         cancels=[submit_event]
     ).then(
@@ -845,19 +1386,35 @@ with gr.Blocks(
     # Listeners for updating the duration estimate
     duration_inputs = [model_dd, search_chk, mr, mc, max_tok, st]
     for component in duration_inputs:
-        component.change(fn=update_duration_estimate, inputs=duration_inputs, outputs=duration_display)
+        component.change(fn=update_duration_ component in duration_inputs:
+        component.change(fn=update_duration_estimate, inputs=duration_inputs, outputs=duration_destimate, inputs=duration_inputs, outputs=duration_display)
+
+    #isplay)
 
-    # Toggle web search settings visibility
+    # Toggle web search Toggle web search settings visibility
     def toggle_search_settings(enabled):
-        return gr.update(visible=enabled)
+        settings visibility
+    def toggle_search_settings(enabled):
+        return gr.update( return gr.update(visible=enabled)
+    
+    search_chvisible=enabled)
     
     search_chk.change(
-        fn=lambda enabled: (update_default_prompt(enabled), gr.update(visible=enabled)),
-        inputs=search_chk,
+        fn=lambda enabled:k.change(
+        fn=lambda enabled: (update_default_prompt(enabled), gr.update(visible (update_default_prompt(enabled), gr.update(visible=enabled)),
+       =enabled)),
+        inputs=search inputs=search_chk,
+        outputs_chk,
         outputs=[sys_prompt, search_settings]
     )
+=[sys_prompt, search_settings]
+    )
     
+    # Clear    
     # Clear chat action
-    clr.click(fn=lambda: ([], "", ""), outputs=[chat, txt, dbg])
+ chat action
+    clr.click(f    clr.click(fn=lambda: ([], "", ""n=lambda: ([], "", ""), outputs=[chat, txt, db), outputs=[chat, txt, dbg])
+    
+    demo.launchg])
     
     demo.launch(share=True)