Update app.py

app.py

@@ -45,9 +45,9 @@ class CreativeAgenticAI:
         else:
             raise ValueError(f"Invalid provider or missing API key for {provider}")
         
-    async def _chutes_chat_async(self, messages: List[Dict], temperature: float = 0.7, max_tokens: int = 1024) -> str:
+    async def _chutes_chat_async(self, messages: List[Dict], temperature: float = 0.7, max_tokens: int = 1024, stream: bool = False) -> Dict:
         """
-        Async method for Chutes API chat
+        Async method for Chutes API chat with thinking support
         """
         headers = {
             "Authorization": f"Bearer {self.chutes_api_key}",
@@ -57,7 +57,7 @@ class CreativeAgenticAI:
         body = {
             "model": self.model,
             "messages": messages,
-            "stream": False,  # Set to False for simpler handling
+            "stream": stream,
             "max_tokens": max_tokens,
             "temperature": temperature
         }
@@ -69,13 +69,66 @@ class CreativeAgenticAI:
                 json=body
             ) as response:
                 if response.status == 200:
-                    result = await response.json()
-                    return result['choices'][0]['message']['content']
+                    if stream:
+                        thinking_content = ""
+                        final_content = ""
+                        in_thinking = False
+                        
+                        async for line in response.content:
+                            line = line.decode("utf-8").strip()
+                            if line.startswith("data: "):
+                                data = line[6:]
+                                if data == "[DONE]":
+                                    break
+                                try:
+                                    chunk_data = json.loads(data)
+                                    if 'choices' in chunk_data and len(chunk_data['choices']) > 0:
+                                        delta = chunk_data['choices'][0].get('delta', {})
+                                        content = delta.get('content', '')
+                                        
+                                        if content:
+                                            # Check for thinking tags
+                                            if '<thinking>' in content:
+                                                in_thinking = True
+                                                thinking_content += content.replace('<thinking>', '')
+                                            elif '</thinking>' in content:
+                                                thinking_content += content.replace('</thinking>', '')
+                                                in_thinking = False
+                                            elif in_thinking:
+                                                thinking_content += content
+                                            else:
+                                                final_content += content
+                                                
+                                except json.JSONDecodeError:
+                                    continue
+                        
+                        return {
+                            "thinking": thinking_content.strip(),
+                            "content": final_content.strip()
+                        }
+                    else:
+                        result = await response.json()
+                        full_content = result['choices'][0]['message']['content']
+                        
+                        # Extract thinking and final content from non-streaming response
+                        thinking_content = ""
+                        final_content = full_content
+                        
+                        if '<thinking>' in full_content and '</thinking>' in full_content:
+                            start_idx = full_content.find('<thinking>') + len('<thinking>')
+                            end_idx = full_content.find('</thinking>')
+                            thinking_content = full_content[start_idx:end_idx].strip()
+                            final_content = full_content[end_idx + len('</thinking>'):].strip()
+                        
+                        return {
+                            "thinking": thinking_content,
+                            "content": final_content
+                        }
                 else:
                     error_text = await response.text()
                     raise Exception(f"Chutes API error: {response.status} - {error_text}")
 
-    def _chutes_chat_sync(self, messages: List[Dict], temperature: float = 0.7, max_tokens: int = 1024) -> str:
+    def _chutes_chat_sync(self, messages: List[Dict], temperature: float = 0.7, max_tokens: int = 1024, stream: bool = True) -> Dict:
         """
         Synchronous wrapper for Chutes API chat
         """
@@ -86,7 +139,7 @@ class CreativeAgenticAI:
             asyncio.set_event_loop(loop)
         
         return loop.run_until_complete(
-            self._chutes_chat_async(messages, temperature, max_tokens)
+            self._chutes_chat_async(messages, temperature, max_tokens, stream)
         )
         
     def chat(self, message: str, 
@@ -136,10 +189,13 @@ IMPORTANT: When you search the web and find information, you MUST:
                 
                 Your responses should be well-structured, informative, and properly cited with working links."""
             else:
-                # Simpler system prompt for Chutes (no web search capabilities)
-                system_prompt = """You are a creative and intelligent AI assistant. 
+                # System prompt for Chutes thinking models
+                system_prompt = """You are a creative and intelligent AI assistant with advanced reasoning capabilities. 
+                Think through problems step-by-step, showing your reasoning process clearly.
                 Be helpful, creative, and engaging while maintaining accuracy.
-                Your responses should be well-structured, informative, and comprehensive."""
+                Your responses should be well-structured, informative, and comprehensive.
+                
+                When solving complex problems, break them down into steps and explain your thinking process."""
         
         # Build messages
         messages = [{"role": "system", "content": system_prompt}]
@@ -174,6 +230,7 @@ IMPORTANT: When you search the web and find information, you MUST:
             
             return {
                 "content": error_msg,
+                "thinking": "",
                 "timestamp": datetime.now().isoformat(),
                 "model": self.model,
                 "provider": self.provider,
@@ -215,6 +272,7 @@ IMPORTANT: When you search the web and find information, you MUST:
         # Create response object
         return {
             "content": processed_content,
+            "thinking": "",  # Groq doesn't have thinking process
             "timestamp": datetime.now().isoformat(),
             "model": self.model,
             "provider": "groq",
@@ -228,16 +286,20 @@ IMPORTANT: When you search the web and find information, you MUST:
         }
 
     def _handle_chutes_chat(self, messages: List[Dict], temperature: float, max_tokens: int, original_message: str) -> Dict:
-        """Handle Chutes API chat"""
-        content = self._chutes_chat_sync(messages, temperature, max_tokens)
+        """Handle Chutes API chat with thinking support"""
+        result = self._chutes_chat_sync(messages, temperature, max_tokens, stream=True)
         
-        # Add to conversation history
+        thinking_content = result.get("thinking", "")
+        final_content = result.get("content", "")
+        
+        # Add to conversation history (only store final content)
         self.conversation_history.append({"role": "user", "content": original_message})
-        self.conversation_history.append({"role": "assistant", "content": content})
+        self.conversation_history.append({"role": "assistant", "content": final_content})
         
         # Create response object
         return {
-            "content": content,
+            "content": final_content,
+            "thinking": thinking_content,
             "timestamp": datetime.now().isoformat(),
             "model": self.model,
             "provider": "chutes",
@@ -381,14 +443,15 @@ def validate_api_keys(groq_api_key: str, chutes_api_key: str, provider: str, mod
             test_response = test_ai._chutes_chat_sync(
                 [{"role": "user", "content": "Hello"}], 
                 temperature=0.7, 
-                max_tokens=10
+                max_tokens=10,
+                stream=False  # Use non-streaming for validation
             )
             
             # Create AI instance
             ai_instance = CreativeAgenticAI(chutes_api_key=chutes_api_key, provider="chutes", model=model)
             api_key_status["chutes"] = "Valid ✅"
             
-            return f"✅ Chutes API Key Valid! Creative AI is ready.\n\n**Provider:** Chutes\n**Model:** {model}\n**Status:** Connected (text generation focused)!"
+            return f"✅ Chutes API Key Valid! Creative AI with Thinking is ready.\n\n**Provider:** Chutes\n**Model:** {model}\n**Status:** Connected with thinking model capabilities!"
             
         except Exception as e:
             api_key_status["chutes"] = "Invalid ❌"
@@ -400,7 +463,7 @@ def get_available_models(provider: str) -> List[str]:
     if provider == "groq":
         return ["compound-beta", "compound-beta-mini"]
     elif provider == "chutes":
-        return ["openai/gpt-oss-20b", "zai-org/GLM-4.5-Air", "Qwen/Qwen3-8B"]
+        return ["openai/gpt-oss-20b", "meta-llama/llama-3.1-8b-instruct", "anthropic/claude-3-sonnet"]
     return []
 
 def update_model_choices(provider: str):
@@ -414,8 +477,9 @@ def chat_with_ai(message: str,
                 system_prompt: str,
                 temperature: float,
                 max_tokens: int,
-                history: List) -> tuple:
-    """Main chat function"""
+                history: List,
+                show_thinking: bool = True) -> tuple:
+    """Main chat function with thinking support"""
     global ai_instance, current_provider
     
     if not ai_instance:
@@ -444,8 +508,20 @@ def chat_with_ai(message: str,
             max_tokens=int(max_tokens)
         )
         
-        # Format response
-        ai_response = response["content"]
+        # Format response with thinking (if available and enabled)
+        ai_response = ""
+        
+        # Add thinking section for Chutes models
+        if current_provider == "chutes" and response.get("thinking") and show_thinking:
+            thinking_content = response["thinking"].strip()
+            if thinking_content:
+                ai_response += f"### 🤔 **Model's Thinking Process:**\n\n"
+                ai_response += f"*The model is reasoning through your question...*\n\n"
+                ai_response += f"```thinking\n{thinking_content}\n```\n\n"
+                ai_response += "---\n\n### 💡 **Final Response:**\n\n"
+        
+        # Add main content
+        ai_response += response["content"]
         
         # Add enhanced tool usage info (Groq only)
         if response.get("tool_usage") and current_provider == "groq":
@@ -474,8 +550,15 @@ def chat_with_ai(message: str,
             
             ai_response += f"\n\n*🌐 Domain filtering applied: {' | '.join(filter_info)}*"
         
-        # Add provider info
-        ai_response += f"\n\n*🤖 Powered by: {current_provider.title()} ({response.get('model', 'unknown')})*"
+        # Add provider and thinking info
+        provider_info = f"🤖 Powered by: {current_provider.title()} ({response.get('model', 'unknown')})"
+        if current_provider == "chutes" and response.get("thinking"):
+            if show_thinking:
+                provider_info += " | 🤔 Thinking process shown"
+            else:
+                provider_info += " | 🤔 Thinking process hidden"
+        
+        ai_response += f"\n\n*{provider_info}*"
         
         # Add to history
         history.append([message, ai_response])
@@ -546,40 +629,78 @@ def create_gradio_app():
         padding: 15px;
         margin: 10px 0;
     }
-    #neuroscope-accordion {
-        background: linear-gradient(to right, #00ff94, #00b4db); 
+    .thinking-info {
+        background-color: #e2e3e5;
+        border: 1px solid #d6d8db;
         border-radius: 8px;
+        padding: 15px;
+        margin: 10px 0;
     }
-    #neuroscope-accordion2 {
+    #neuroscope-accordion {
         background: linear-gradient(to right, #00ff94, #00b4db); 
         border-radius: 8px;
-        margin-top: 10px;
     }
     """
     
-    with gr.Blocks(css=css, title="🤖 Multi-Provider Creative Agentic AI Chat", theme=gr.themes.Ocean()) as app:
+    with gr.Blocks(css=css, title="🤖 Multi-Provider Creative Agentic AI Chat with Thinking", theme=gr.themes.Ocean()) as app:
         
         # Header
         gr.HTML("""
         <div class="header">
-            <h1>🤖 NeuroScope-AI Enhanced</h1>
-            <p>Multi-Provider AI Chat Tool - Powered by Groq's Compound Models & Chutes API</p>
+            <h1>🤖 NeuroScope-AI Enhanced with Thinking Models</h1>
+            <p>Multi-Provider AI Chat Tool - Groq's Compound Models & Chutes Thinking Models</p>
         </div>
         """)
 
         # Provider Selection
         with gr.Group():
-            with gr.Accordion("🤖 Multi-Provider NeuroScope AI", open=False, elem_id="neuroscope-accordion"):
+            with gr.Accordion("🤖 Multi-Provider NeuroScope AI with Thinking", open=False, elem_id="neuroscope-accordion"):
                 gr.Markdown("""
-                    **Enhanced with Multiple AI Providers:**
-                    - 🧠 Intelligence (Neuro) - Now supports Groq & Chutes
-                    - 🔍 Advanced capabilities (Scope) - Web search with Groq, powerful text generation with Chutes
-                    - 🤖 AI capabilities (AI) - Multiple model options
+                    **Enhanced with Multiple AI Providers & Thinking Models:**
+                    - 🧠 Intelligence (Neuro) - Now supports Groq & Chutes Thinking Models
+                    - 🔍 Advanced capabilities (Scope) - Web search with Groq, reasoning traces with Chutes
+                    - 🤖 AI capabilities (AI) - Multiple model options including thinking models
                     - ⚡ Precision & Speed (Scope) - Choose the best provider for your needs
+                    - 🤔 **NEW**: Thinking process visualization for Chutes models
+                """)
+
+        # Thinking Models Info
+        with gr.Group():
+            with gr.Accordion("🤔 About Chutes Thinking Models", open=False, elem_id="neuroscope-accordion"):
+                gr.Markdown("""
+                <div class="thinking-info">
+                <h3>🧠 What are Thinking Models?</h3>
+                <p><strong>Chutes Thinking Models</strong> are advanced AI systems that show their reasoning process before providing the final answer.</p>
+                
+                <h4>🔍 How They Work:</h4>
+                <ul>
+                    <li><strong>Step-by-Step Reasoning:</strong> Models think through problems systematically</li>
+                    <li><strong>Transparent Process:</strong> You can see exactly how the AI reaches its conclusions</li>
+                    <li><strong>Better Accuracy:</strong> The thinking process often leads to more accurate and well-reasoned responses</li>
+                    <li><strong>Educational Value:</strong> Learn from the AI's problem-solving approach</li>
+                </ul>
+                
+                <h4>🎯 Available Thinking Models:</h4>
+                <ul>
+                    <li><strong>openai/gpt-oss-20b:</strong> Large-scale reasoning and analysis</li>
+                    <li><strong>meta-llama/llama-3.1-8b-instruct:</strong> Efficient thinking and instruction following</li>
+                    <li><strong>anthropic/claude-3-sonnet:</strong> Advanced reasoning and creative thinking</li>
+                </ul>
+                
+                <h4>💡 Best For:</h4>
+                <ul>
+                    <li>Complex problem-solving tasks</li>
+                    <li>Mathematical and logical reasoning</li>
+                    <li>Step-by-step analysis</li>
+                    <li>Educational explanations</li>
+                    <li>Creative writing with detailed planning</li>
+                </ul>
+                </div>
                 """)
 
         # Provider and API Key Section
         with gr.Row():
+                        
             with gr.Column():
                 provider_selection = gr.Radio(
                     choices=["groq", "chutes"],
@@ -604,8 +725,7 @@ def create_gradio_app():
                     info="Get your API key from: https://chutes.ai/",
                     visible=False
                 )
-            
-            with gr.Column():
+                
                 model_selection = gr.Radio(
                     choices=get_available_models("groq"),
                     label="🧠 Groq Models",
@@ -613,6 +733,14 @@ def create_gradio_app():
                     info="compound-beta: More powerful | compound-beta-mini: Faster"
                 )
                 
+                # Thinking toggle for Chutes
+                show_thinking = gr.Checkbox(
+                    label="🤔 Show Thinking Process",
+                    value=True,
+                    info="Display the model's reasoning process (Chutes only)",
+                    visible=False
+                )
+                
                 connect_btn = gr.Button("🔗 Connect", variant="primary", size="lg")
         
         # Status display
@@ -632,26 +760,29 @@ def create_gradio_app():
                 - ✅ **Citation System** - Automatic source linking and references
                 - ⚡ **Ultra-fast inference** - Groq's hardware acceleration
                 - 🧠 **Models**: compound-beta, compound-beta-mini
+                - ❌ **No thinking process** - Direct responses without visible reasoning
                 
-                **🎯 Chutes**
+                **🎯 Chutes (Thinking Models)**
                 - ✅ **Multiple Model Access** - Various open-source and commercial models
-                - ✅ **Cost-effective** - Competitive pricing
-                - ✅ **High-quality text generation** - Excellent for creative writing and analysis
+                - ✅ **Thinking Process** - See the model's step-by-step reasoning
+                - ✅ **High-quality reasoning** - Better accuracy through visible thinking
+                - ✅ **Educational Value** - Learn from AI's problem-solving approach
                 - ⚡ **Good performance** - Reliable and fast responses
-                - 🧠 **Models**: GPT-OSS-20B, Llama 3.1, Claude 3 Sonnet, and more
+                - 🧠 **Models**: GPT-OSS-20B, Llama 3.1, Claude 3 Sonnet (all with thinking)
                 - ❌ **No web search** - Relies on training data only
                 
                 **💡 Use Groq when you need:**
                 - Real-time information and web search
                 - Research with source citations
                 - Domain-specific searches
-                - Agentic AI capabilities
+                - Ultra-fast responses
                 
                 **💡 Use Chutes when you need:**
-                - Pure text generation and analysis
-                - Creative writing tasks
-                - Cost-effective AI access
-                - Variety of model options
+                - Complex problem-solving with visible reasoning
+                - Educational explanations
+                - Mathematical and logical analysis
+                - Creative planning with detailed thinking
+                - Understanding AI's reasoning process
                 </div>
                 """)
         
@@ -666,15 +797,14 @@ def create_gradio_app():
                 gr.update(visible=chutes_visible),  # chutes_api_key
                 gr.update(choices=models, value=models[0] if models else None, 
                          label=f"🧠 {provider.title()} Models"),  # model_selection
-                gr.update(visible=groq_visible),  # domain filtering sections
-                gr.update(visible=groq_visible),  # include domains
-                gr.update(visible=groq_visible)   # exclude domains
+                gr.update(visible=chutes_visible),  # show_thinking
+                gr.update(visible=groq_visible),  # domain_group (will be defined later)
             )
         
         provider_selection.change(
             fn=update_provider_ui,
             inputs=[provider_selection],
-            outputs=[groq_api_key, chutes_api_key, model_selection]  # We'll add domain filtering updates later
+            outputs=[groq_api_key, chutes_api_key, model_selection, show_thinking, domain_group]  # domain_group will be added later
         )
         
         # Connect button functionality
@@ -687,8 +817,8 @@ def create_gradio_app():
         # Main Chat Interface
         with gr.Tab("💬 Chat"):
             chatbot = gr.Chatbot(
-                label="Multi-Provider Creative AI Assistant",
-                height=500,
+                label="Multi-Provider Creative AI Assistant with Thinking",
+                height=600,
                 show_label=True,
                 bubble_full_width=False,
                 show_copy_button=True
@@ -732,7 +862,7 @@ def create_gradio_app():
             )    
         
         # Domain Filtering Section (Groq only)
-        with gr.Group():
+        with gr.Group() as domain_group:
             with gr.Accordion("🌐 Domain Filtering (Groq Web Search Only)", open=False, elem_id="neuroscope-accordion"):
                 gr.Markdown("""
                 <div class="domain-info">
@@ -761,7 +891,7 @@ def create_gradio_app():
                         info="Never search these domains"
                     )
                     
-            with gr.Accordion("🔗 Common Domain Examples", open=False, elem_id="neuroscope-accordion2"):
+            with gr.Accordion("🔗 Common Domain Examples", open=False, elem_id="neuroscope-accordion"):
                 gr.Markdown("""
                 **Academic & Research:**
                 - `arxiv.org`, `*.edu`, `scholar.google.com`, `researchgate.net`
@@ -779,7 +909,7 @@ def create_gradio_app():
                 - `nature.com`, `science.org`, `pubmed.ncbi.nlm.nih.gov`, `who.int`
                 """)
 
-        # Update provider UI function with domain filtering
+        # Update provider UI function with domain filtering and thinking toggle
         def update_provider_ui_complete(provider):
             groq_visible = provider == "groq"
             chutes_visible = provider == "chutes"
@@ -790,25 +920,27 @@ def create_gradio_app():
                 gr.update(visible=chutes_visible),  # chutes_api_key
                 gr.update(choices=models, value=models[0] if models else None, 
                          label=f"🧠 {provider.title()} Models"),  # model_selection
+                gr.update(visible=chutes_visible),  # show_thinking
                 gr.update(visible=groq_visible),  # domain_group
             )
         
         provider_selection.change(
             fn=update_provider_ui_complete,
             inputs=[provider_selection],
-            outputs=[groq_api_key, chutes_api_key, model_selection]
+            outputs=[groq_api_key, chutes_api_key, model_selection, show_thinking, domain_group]
         )
         
-        # IMPORTANT Section with Citation Info
+        # IMPORTANT Section with Citation Info and Thinking
         with gr.Group():
-            with gr.Accordion("📚 IMPORTANT - Citations & Multi-Provider Features!", open=False, elem_id="neuroscope-accordion"):
+            with gr.Accordion("📚 IMPORTANT - Citations & Thinking Models!", open=False, elem_id="neuroscope-accordion"):
                 gr.Markdown("""
                 <div class="citation-info">
-                <h3>🆕 Multi-Provider Enhancement</h3>
+                <h3>🆕 Multi-Provider Enhancement with Thinking Models</h3>
                 <p>This enhanced version now supports both Groq and Chutes AI providers:</p>
                 <ul>
                     <li><strong>🚀 Groq Integration:</strong> Agentic AI with web search, citations, and tool usage</li>
-                    <li><strong>🎯 Chutes Integration:</strong> Multiple AI models for text generation and analysis</li>
+                    <li><strong>🎯 Chutes Integration:</strong> Multiple thinking models with visible reasoning processes</li>
+                    <li><strong>🤔 Thinking Process:</strong> See step-by-step reasoning from Chutes models</li>
                     <li><strong>🔄 Easy Switching:</strong> Switch between providers based on your needs</li>
                     <li><strong>📊 Provider Comparison:</strong> Clear information about each provider's strengths</li>
                 </ul>
@@ -822,13 +954,14 @@ def create_gradio_app():
                     <li><strong>Search Query Tracking:</strong> Shows what queries were made to find information</li>
                 </ul>
                 
-                <h3>🎯 Chutes Model Access</h3>
+                <h3>🤔 Chutes Thinking Model Features</h3>
                 <p>When using Chutes, you get access to:</p>
                 <ul>
-                    <li><strong>Multiple Models:</strong> GPT-OSS-20B, Llama 3.1, Claude 3 Sonnet</li>
-                    <li><strong>Cost-Effective:</strong> Competitive pricing for high-quality AI</li>
-                    <li><strong>Specialized Tasks:</strong> Optimized for creative writing and analysis</li>
-                    <li><strong>Reliable Performance:</strong> Consistent and fast responses</li>
+                    <li><strong>Thinking Process Visualization:</strong> See exactly how the AI reasons through problems</li>
+                    <li><strong>Step-by-Step Analysis:</strong> Watch the model break down complex questions</li>
+                    <li><strong>Toggle Thinking Display:</strong> Choose to show or hide the reasoning process</li>
+                    <li><strong>Educational Value:</strong> Learn from AI's problem-solving approaches</li>
+                    <li><strong>Multiple Thinking Models:</strong> Different models with unique reasoning styles</li>
                 </ul>
                 </div>
                 
@@ -849,10 +982,20 @@ def create_gradio_app():
                   - Useful for **filtering out unreliable or unwanted sources**.
                   - Allows broad search with **targeted exclusions**.
                 
-                **Both Include and Exclude Domains Specified:**
-                  - **Only the include domains** are used for searching.
-                  - **Exclude list is ignored** because the include list already restricts search scope.
-                  - Guarantees AI pulls content **exclusively from whitelisted domains**, regardless of the excluded ones.
+                ### 🤔 **Thinking Models Behavior (Chutes Only)**
+                
+                **How Thinking Models Work:**
+                  - Models first **reason through the problem** in a thinking section
+                  - You can see the **step-by-step thought process**
+                  - The model then provides its **final polished answer**
+                  - **Toggle thinking display** on/off as needed
+                
+                **Best Use Cases for Thinking Models:**
+                  - Complex mathematical problems
+                  - Logical reasoning tasks
+                  - Creative writing with planning
+                  - Educational explanations
+                  - Problem-solving scenarios
                 
                 ---
                 
@@ -863,21 +1006,23 @@ def create_gradio_app():
                   - A **professional business consultant**
                   - A **coding mentor**
                   - A **creative writer**
+                  - A **step-by-step tutor** (especially effective with thinking models)
                   - A **specific character or persona**
                 - Provides full control to **reshape the AI's tone, expertise, and conversational style** with a single prompt.
                 """)
         
         # How to Use Section
-        with gr.Accordion("📖 How to Use This Enhanced Multi-Provider App", open=False, elem_id="neuroscope-accordion"):
+        with gr.Accordion("📖 How to Use This Enhanced Multi-Provider App with Thinking", open=False, elem_id="neuroscope-accordion"):
             gr.Markdown("""
             ### 🚀 Getting Started
-            1. **Choose your AI Provider** - Select between Groq (web search + agentic) or Chutes (text generation)
+            1. **Choose your AI Provider** - Select between Groq (web search + agentic) or Chutes (thinking models)
             2. **Enter your API Key** - 
                - Groq: Get one from [console.groq.com](https://console.groq.com/)
                - Chutes: Get one from [chutes.ai](https://chutes.ai/)
             3. **Select a model** - Choose from provider-specific model options
-            4. **Click Connect** - Validate your key and connect to the AI
-            5. **Start chatting!** - Type your message and get intelligent responses
+            4. **Configure thinking display** - For Chutes, decide if you want to see the reasoning process
+            5. **Click Connect** - Validate your key and connect to the AI
+            6. **Start chatting!** - Type your message and get intelligent responses
             
             ### 🎯 Key Features
             **🚀 Groq Features:**
@@ -887,10 +1032,11 @@ def create_gradio_app():
             - **Ultra-fast**: Groq's hardware-accelerated inference
             
             **🎯 Chutes Features:**
-            - **Multiple Models**: Access to various open-source and commercial models
-            - **Cost-Effective**: Competitive pricing for AI access
-            - **High Quality**: Excellent text generation and analysis
-            - **Model Variety**: Choose the best model for your specific task
+            - **Thinking Models**: See the model's step-by-step reasoning process
+            - **Multiple Models**: Access to GPT-OSS-20B, Llama 3.1, Claude 3 Sonnet
+            - **Educational**: Learn from AI's problem-solving approaches
+            - **Toggle Thinking**: Show/hide the reasoning process as needed
+            - **High Quality**: Excellent reasoning and analysis capabilities
             
             **🔄 Universal Features:**
             - **Memory**: Maintains conversation context throughout the session
@@ -904,73 +1050,83 @@ def create_gradio_app():
             - Check the "Sources Used" section for all references
             - Try different domain combinations to see varied results
             
-            **For Chutes:**
-            - Experiment with different models for different tasks
-            - Use higher temperatures for creative tasks
-            - Leverage the variety of available models (GPT, Llama, Claude)
-            - Perfect for tasks that don't require real-time information
+            **For Chutes (Thinking Models):**
+            - Ask complex, multi-step questions to see rich thinking processes
+            - Use for educational purposes - the thinking is very instructive
+            - Try mathematical problems, logical puzzles, or creative planning tasks
+            - Toggle thinking display based on whether you want to see the process
+            - Different models have different thinking styles - experiment!
             
             **General:**
             - Adjust temperature: higher for creativity, lower for precision
             - Try different system prompts for different conversation styles
             - Use the provider that best fits your current task
+            - For learning: use Chutes with thinking display enabled
+            - For research: use Groq with appropriate domain filtering
             """)
         
         # Sample Examples Section
-        with gr.Accordion("🎯 Sample Examples to Test Both Providers", open=False, elem_id="neuroscope-accordion"):
+        with gr.Accordion("🎯 Sample Examples to Test Both Providers & Thinking", open=False, elem_id="neuroscope-accordion"):
             gr.Markdown("""
             <div class="example-box">
             <h4>🆚 Provider Comparison Examples</h4>
             <p>Try the same prompts with both providers to see the difference:</p>
             
-            <h4>🔬 Research & Analysis</h4>
+            <h4>🤔 Perfect for Thinking Models (Chutes)</h4>
             <ul>
-                <li><strong>Groq (with web search):</strong> "What are the latest breakthroughs in quantum computing in 2024?"</li>
-                <li><strong>Chutes (knowledge-based):</strong> "Explain the fundamental principles of quantum computing"</li>
-                <li><strong>Groq with domains:</strong> Same question with "arxiv.org, *.edu" in include domains</li>
+                <li><strong>Math & Logic:</strong> "Solve this step by step: If a train travels 120 miles in 2 hours, and then 180 miles in 3 hours, what's the average speed for the entire journey?"</li>
+                <li><strong>Problem Solving:</strong> "I have a budget of $1000 for a home office setup. Help me plan the best allocation across desk, chair, computer, and lighting."</li>
+                <li><strong>Creative Planning:</strong> "Plan a short story about a time traveler who accidentally changes history. Walk through the plot structure."</li>
+                <li><strong>Analysis:</strong> "Compare and contrast the pros and cons of remote work vs office work, considering productivity, collaboration, and work-life balance."</li>
             </ul>
             
-            <h4>💻 Programming & Tech</h4>
+            <h4>🔬 Research & Real-time Info (Groq)</h4>
             <ul>
-                <li><strong>Groq:</strong> "What are the current best practices for React 18 in 2024?"</li>
-                <li><strong>Chutes:</strong> "Write a comprehensive React component with hooks and best practices"</li>
-                <li><strong>Groq filtered:</strong> Same with "github.com, stackoverflow.com" included</li>
+                <li><strong>Current Events:</strong> "What are the latest developments in AI research in 2024?"</li>
+                <li><strong>Tech Updates:</strong> "What are the newest features in React 19?"</li>
+                <li><strong>Market Analysis:</strong> "Current trends in cryptocurrency markets with sources"</li>
+                <li><strong>Scientific Updates:</strong> "Recent breakthroughs in quantum computing research"</li>
             </ul>
             
-            <h4>🎨 Creative Tasks (Great for Chutes)</h4>
+            <h4>💻 Programming & Tech (Compare Both)</h4>
             <ul>
-                <li>"Write a short story about AI and humans working together"</li>
-                <li>"Create a marketing plan for a sustainable fashion brand"</li>
-                <li>"Generate ideas for a mobile app that helps with mental health"</li>
-                <li>"Write a poem about the beauty of code"</li>
+                <li><strong>Groq:</strong> "What are the current best practices for React 18 in 2024?" (with web search)</li>
+                <li><strong>Chutes:</strong> "Explain how to build a React component with useState, and walk through your reasoning for the design choices"</li>
             </ul>
             
-            <h4>📊 Business & Analysis</h4>
+            <h4>🎨 Creative Tasks (Great for Thinking Models)</h4>
             <ul>
-                <li><strong>Groq:</strong> "What are the current trends in cryptocurrency markets?"</li>
-                <li><strong>Chutes:</strong> "Analyze the pros and cons of different investment strategies"</li>
-                <li><strong>Groq filtered:</strong> Crypto question with "bloomberg.com, wsj.com" included</li>
+                <li>"Write a marketing strategy for a new eco-friendly product, showing your planning process"</li>
+                <li>"Create a study plan for learning Python in 3 months, explaining your reasoning for each phase"</li>
+                <li>"Design a mobile app concept for meditation, walking through your design thinking"</li>
             </ul>
             
-            <h4>🧠 Model-Specific Testing (Chutes)</h4>
+            <h4>🧠 Model-Specific Testing (Chutes Thinking)</h4>
             <ul>
-                <li><strong>GPT-OSS-20B:</strong> "Explain complex scientific concepts in simple terms"</li>
-                <li><strong>Llama 3.1:</strong> "Help me debug this Python code and explain the solution"</li>
-                <li><strong>Claude 3 Sonnet:</strong> "Analyze this business scenario and provide strategic recommendations"</li>
+                <li><strong>GPT-OSS-20B:</strong> "Explain quantum entanglement in simple terms, showing your thought process for making it accessible"</li>
+                <li><strong>Llama 3.1:</strong> "Debug this Python code and explain your debugging approach: [code with intentional errors]"</li>
+                <li><strong>Claude 3 Sonnet:</strong> "Analyze this business scenario and provide strategic recommendations, showing your analytical framework"</li>
+            </ul>
+            
+            <h4>📊 Side-by-Side Comparisons</h4>
+            <ul>
+                <li><strong>Same Question, Different Providers:</strong> Ask "How do neural networks work?" to both providers</li>
+                <li><strong>Groq Result:</strong> Fast response with potential web sources and current information</li>
+                <li><strong>Chutes Result:</strong> Detailed thinking process showing how the model breaks down the explanation</li>
             </ul>
             </div>
             """)
         
-        # Event handlers
+        # Event handlers - Updated to include thinking toggle
         send_btn.click(
             fn=chat_with_ai,
-            inputs=[msg, include_domains, exclude_domains, system_prompt, temperature, max_tokens, chatbot],
+            inputs=[msg, include_domains, exclude_domains, system_prompt, temperature, max_tokens, chatbot, show_thinking],
             outputs=[chatbot, msg]
         )
         
         msg.submit(
             fn=chat_with_ai,
-            inputs=[msg, include_domains, exclude_domains, system_prompt, temperature, max_tokens, chatbot],
+            inputs=[msg, include_domains, exclude_domains, system_prompt, temperature, max_tokens, chatbot, show_thinking],
             outputs=[chatbot, msg]
         )
         
@@ -980,16 +1136,16 @@ def create_gradio_app():
         )
         
         # Footer
-        with gr.Accordion("🚀 About This Enhanced Multi-Provider Tool", open=True, elem_id="neuroscope-accordion"):
+        with gr.Accordion("🚀 About This Enhanced Multi-Provider Tool with Thinking", open=True, elem_id="neuroscope-accordion"):
             gr.Markdown("""
-            **Enhanced Multi-Provider Creative Agentic AI Chat Tool** with dual API support:
+            **Enhanced Multi-Provider Creative Agentic AI Chat Tool** with thinking model support:
             
-            **🆕 New Multi-Provider Features:**
-            - 🚀 **Groq Integration**: Agentic AI with web search, citations, and tool usage
-            - 🎯 **Chutes Integration**: Multiple AI models for diverse text generation tasks
-            - 🔄 **Provider Switching**: Easy switching between different AI providers
-            - 📊 **Provider Comparison**: Clear information about each provider's strengths
-            - 🧠 **Multiple Models**: Access to various AI models through both providers
+            **🆕 New Thinking Model Features:**
+            - 🤔 **Thinking Process Visualization**: See step-by-step reasoning from Chutes models
+            - 🧠 **Multiple Thinking Models**: GPT-OSS-20B, Llama 3.1, Claude 3 Sonnet with reasoning
+            - 🔄 **Toggle Thinking Display**: Choose to show or hide the reasoning process
+            - 📚 **Educational Value**: Learn from AI's problem-solving approaches
+            - 🎯 **Better Accuracy**: Thinking process often leads to more accurate responses
             
             **🚀 Groq Features:**
             - 🔗 **Automatic Source Citations**: Every response includes clickable links to sources
@@ -1001,10 +1157,11 @@ def create_gradio_app():
             - 🧠 Advanced AI reasoning with tool usage
             
             **🎯 Chutes Features:**
-            - 🤖 **Multiple AI Models**: GPT-OSS-20B, Llama 3.1, Claude 3 Sonnet
+            - 🤖 **Multiple Thinking Models**: Various AI models with visible reasoning
             - 💰 **Cost-Effective**: Competitive pricing for AI access
-            - 🎨 **Creative Excellence**: Optimized for writing and analysis tasks
-            - ⚡ **Reliable Performance**: Consistent and fast responses
+            - 🎨 **Creative Excellence**: Optimized for reasoning and analysis tasks
+            - ⚡ **Reliable Performance**: Consistent and thoughtful responses
+            - 📖 **Learning Tool**: Perfect for understanding AI reasoning
             
             **🔄 Universal Features:**
             - 💬 Conversational memory and context
@@ -1014,7 +1171,14 @@ def create_gradio_app():
             
             **💡 Choose Your Provider:**
             - **Use Groq** when you need real-time information, web search, and citations
-            - **Use Chutes** when you need pure text generation, creative writing, or cost-effective AI access
+            - **Use Chutes** when you want to see the thinking process, need complex reasoning, or want educational explanations
+            - **Toggle thinking display** in Chutes to customize your experience
+            
+            **🎓 Perfect for:**
+            - Students learning problem-solving approaches
+            - Developers wanting to understand AI reasoning
+            - Researchers needing both current information (Groq) and deep analysis (Chutes)
+            - Anyone curious about how AI thinks through problems
             """)
     
     return app