chat-console 0.4.0__py3-none-any.whl → 0.4.3__py3-none-any.whl

app/__init__.py

@@ -3,4 +3,4 @@ Chat CLI
 A command-line interface for chatting with various LLM providers like ChatGPT and Claude.
 """
 
-__version__ = "0.4.0"
+__version__ = "0.4.3"

app/api/ollama.py

@@ -31,8 +31,96 @@ class OllamaClient(BaseModelClient):
         # Track model loading state
         self._model_loading = False
         
+        # Track preloaded models and their last use timestamp
+        self._preloaded_models = {}
+        
+        # Default timeout values (in seconds)
+        self.DEFAULT_TIMEOUT = 30
+        self.MODEL_LOAD_TIMEOUT = 120
+        self.MODEL_PULL_TIMEOUT = 3600  # 1 hour for large models
+        
         # Path to the cached models file
         self.models_cache_path = Path(__file__).parent.parent / "data" / "ollama-models.json"
+        
+    def get_timeout_for_model(self, model_id: str, operation: str = "generate") -> int:
+        """
+        Calculate an appropriate timeout based on model size
+        
+        Parameters:
+        - model_id: The model identifier
+        - operation: The operation type ('generate', 'load', 'pull')
+        
+        Returns:
+        - Timeout in seconds
+        """
+        # Default timeouts by operation
+        default_timeouts = {
+            "generate": self.DEFAULT_TIMEOUT,      # 30s
+            "load": self.MODEL_LOAD_TIMEOUT,       # 2min 
+            "pull": self.MODEL_PULL_TIMEOUT,       # 1h
+            "list": 5,                             # 5s
+            "test": 2                              # 2s
+        }
+        
+        # Parameter size multipliers
+        size_multipliers = {
+            # For models < 3B
+            "1b": 0.5,
+            "2b": 0.7,
+            "3b": 1.0,
+            # For models 3B-10B
+            "5b": 1.2,
+            "6b": 1.3,
+            "7b": 1.5,
+            "8b": 1.7,
+            "9b": 1.8,
+            # For models 10B-20B
+            "13b": 2.0,
+            "14b": 2.0,
+            # For models 20B-50B
+            "27b": 3.0,
+            "34b": 3.5,
+            "40b": 4.0,
+            # For models 50B+
+            "70b": 5.0,
+            "80b": 6.0,
+            "100b": 7.0,
+            "400b": 10.0,
+            "405b": 10.0,
+        }
+        
+        # Get the base timeout for the operation
+        base_timeout = default_timeouts.get(operation, self.DEFAULT_TIMEOUT)
+        
+        # Try to determine the model size from the model ID
+        model_size = "7b"  # Default assumption is 7B parameters
+        model_lower = model_id.lower()
+        
+        # Check for size indicators in the model name
+        for size in size_multipliers.keys():
+            if size in model_lower:
+                model_size = size
+                break
+                
+        # If it's a known large model without size in name
+        if "llama3.1" in model_lower and not any(size in model_lower for size in size_multipliers.keys()):
+            model_size = "8b"  # Default for llama3.1 without size specified
+            
+        # For first generation after model selection, if preloaded, use shorter timeout
+        if operation == "generate" and model_id in self._preloaded_models:
+            # For preloaded models, use a shorter timeout
+            return max(int(base_timeout * 0.7), 20)  # Min 20 seconds
+        
+        # Calculate final timeout with multiplier
+        multiplier = size_multipliers.get(model_size, 1.0)
+        timeout = int(base_timeout * multiplier)
+        
+        # For pull operation, ensure we have a reasonable maximum
+        if operation == "pull":
+            return min(timeout, 7200)  # Max 2 hours
+            
+        logger.info(f"Calculated timeout for {model_id} ({operation}): {timeout}s (base: {base_timeout}s, multiplier: {multiplier})")
+        return timeout
 
     @classmethod
     async def create(cls) -> 'OllamaClient':
@@ -61,7 +149,29 @@ class OllamaClient(BaseModelClient):
             style_instructions = self._get_style_instructions(style)
             debug_log(f"Adding style instructions: {style_instructions[:50]}...")
             formatted_messages.append(style_instructions)
+            
+        # Special case for title generation - check if this is a title generation message
+        is_title_generation = False
+        for msg in messages:
+            if msg.get("role") == "system" and "generate a brief, descriptive title" in msg.get("content", "").lower():
+                is_title_generation = True
+                debug_log("Detected title generation prompt")
+                break
+        
+        # For title generation, use a direct approach
+        if is_title_generation:
+            debug_log("Using specialized formatting for title generation")
+            # Find the user message containing the input for title generation
+            user_msg = next((msg for msg in messages if msg.get("role") == "user"), None)
+            if user_msg and "content" in user_msg:
+                # Create a direct prompt
+                prompt = "You must generate a short, descriptive title (maximum 40 characters) for this conversation. ONLY output the title with no additional text, no quotes, and no explanation. Do not start with phrases like 'Here's a title' or 'Title:'. RESPOND ONLY WITH THE TITLE TEXT for the following message:\n\n" + user_msg["content"]
+                debug_log(f"Created title generation prompt: {prompt[:100]}...")
+                return prompt
+            else:
+                debug_log("Could not find user message for title generation, using standard formatting")
         
+        # Standard processing for normal chat messages
         # Add message content, preserving conversation flow
         for i, msg in enumerate(messages):
             try:
@@ -185,6 +295,7 @@ class OllamaClient(BaseModelClient):
             try:
                 async with aiohttp.ClientSession() as session:
                     logger.debug(f"Sending request to {self.base_url}/api/generate")
+                    gen_timeout = self.get_timeout_for_model(model, "generate")
                     async with session.post(
                         f"{self.base_url}/api/generate",
                         json={
@@ -193,12 +304,16 @@ class OllamaClient(BaseModelClient):
                             "temperature": temperature,
                             "stream": False
                         },
-                        timeout=30
+                        timeout=gen_timeout
                     ) as response:
                         response.raise_for_status()
                         data = await response.json()
                         if "response" not in data:
                             raise Exception("Invalid response format from Ollama server")
+                            
+                        # Update the model usage timestamp to keep it hot
+                        self.update_model_usage(model)
+                        
                         return data["response"]
                         
             except aiohttp.ClientConnectorError:
@@ -324,10 +439,11 @@ class OllamaClient(BaseModelClient):
                                 "stream": False
                             }
                             
+                        test_timeout = self.get_timeout_for_model(model, "test")
                         async with session.post(
                             f"{self.base_url}/api/generate",
                             json=test_payload,
-                            timeout=2
+                            timeout=test_timeout
                         ) as response:
                             if response.status != 200:
                                 logger.warning(f"Model test request failed with status {response.status}")
@@ -361,10 +477,11 @@ class OllamaClient(BaseModelClient):
                             debug_log(f"Error preparing pull payload: {str(pull_err)}, using default")
                             pull_payload = {"name": "gemma:2b"}  # Safe default
                             
+                        pull_timeout = self.get_timeout_for_model(model, "pull")
                         async with session.post(
                             f"{self.base_url}/api/pull",
                             json=pull_payload,
-                            timeout=60
+                            timeout=pull_timeout
                         ) as pull_response:
                             if pull_response.status != 200:
                                 logger.error("Failed to pull model")
@@ -415,10 +532,11 @@ class OllamaClient(BaseModelClient):
                         }
                     
                     debug_log(f"Sending request to Ollama API")
+                    gen_timeout = self.get_timeout_for_model(model, "generate")
                     response = await session.post(
                         f"{self.base_url}/api/generate",
                         json=request_payload,
-                        timeout=60  # Longer timeout for actual generation
+                        timeout=gen_timeout
                     )
                     response.raise_for_status()
                     debug_log(f"Response status: {response.status}")
@@ -426,6 +544,9 @@ class OllamaClient(BaseModelClient):
                     # Use a simpler async iteration pattern that's less error-prone
                     debug_log("Starting to process response stream")
                     
+                    # Update the model usage timestamp to keep it hot
+                    self.update_model_usage(model)
+                    
                     # Set a flag to track if we've yielded any content
                     has_yielded_content = False
                     
@@ -535,6 +656,123 @@ class OllamaClient(BaseModelClient):
     def is_loading_model(self) -> bool:
         """Check if Ollama is currently loading a model"""
         return self._model_loading
+    
+    async def preload_model(self, model_id: str) -> bool:
+        """
+        Preload a model to keep it hot/ready for use
+        Returns True if successful, False otherwise
+        """
+        from datetime import datetime
+        import asyncio
+        
+        logger.info(f"Preloading model: {model_id}")
+        
+        # First, check if the model is already preloaded
+        if model_id in self._preloaded_models:
+            # Update timestamp if already preloaded
+            self._preloaded_models[model_id] = datetime.now()
+            logger.info(f"Model {model_id} already preloaded, updated timestamp")
+            return True
+        
+        try:
+            # We'll use a minimal prompt to load the model
+            warm_up_prompt = "hello"
+            
+            # Set model loading state
+            old_loading_state = self._model_loading
+            self._model_loading = True
+            
+            async with aiohttp.ClientSession() as session:
+                # First try pulling the model if needed
+                try:
+                    logger.info(f"Ensuring model {model_id} is pulled")
+                    pull_payload = {"name": model_id}
+                    pull_timeout = self.get_timeout_for_model(model_id, "pull")
+                    async with session.post(
+                        f"{self.base_url}/api/pull",
+                        json=pull_payload,
+                        timeout=pull_timeout
+                    ) as pull_response:
+                        # We don't need to process the full pull, just initiate it
+                        if pull_response.status != 200:
+                            logger.warning(f"Pull request for model {model_id} failed with status {pull_response.status}")
+                except Exception as e:
+                    logger.warning(f"Error during model pull check: {str(e)}")
+                
+                # Now send a small generation request to load the model into memory
+                logger.info(f"Sending warm-up request for model {model_id}")
+                gen_timeout = self.get_timeout_for_model(model_id, "load")
+                async with session.post(
+                    f"{self.base_url}/api/generate",
+                    json={
+                        "model": model_id,
+                        "prompt": warm_up_prompt,
+                        "temperature": 0.7,
+                        "stream": False
+                    },
+                    timeout=gen_timeout
+                ) as response:
+                    if response.status != 200:
+                        logger.error(f"Failed to preload model {model_id}, status: {response.status}")
+                        self._model_loading = old_loading_state
+                        return False
+                    
+                    # Read the response to ensure the model is fully loaded
+                    await response.json()
+                    
+                    # Update preloaded models with timestamp
+                    self._preloaded_models[model_id] = datetime.now()
+                    logger.info(f"Successfully preloaded model {model_id}")
+                    return True
+        except Exception as e:
+            logger.error(f"Error preloading model {model_id}: {str(e)}")
+            return False
+        finally:
+            # Reset model loading state
+            self._model_loading = old_loading_state
+    
+    def get_preloaded_models(self) -> Dict[str, datetime]:
+        """Return the dict of preloaded models and their last use times"""
+        return self._preloaded_models
+    
+    def update_model_usage(self, model_id: str) -> None:
+        """Update the timestamp for a model that is being used"""
+        if model_id and model_id in self._preloaded_models:
+            from datetime import datetime
+            self._preloaded_models[model_id] = datetime.now()
+            logger.info(f"Updated usage timestamp for model {model_id}")
+    
+    async def release_inactive_models(self, max_inactive_minutes: int = 30) -> List[str]:
+        """
+        Release models that have been inactive for more than the specified time
+        Returns a list of model IDs that were released
+        """
+        from datetime import datetime, timedelta
+        
+        if not self._preloaded_models:
+            return []
+            
+        now = datetime.now()
+        inactive_threshold = timedelta(minutes=max_inactive_minutes)
+        models_to_release = []
+        
+        # Find models that have been inactive for too long
+        for model_id, last_used in list(self._preloaded_models.items()):
+            if now - last_used > inactive_threshold:
+                models_to_release.append(model_id)
+                
+        # Release the models
+        released_models = []
+        for model_id in models_to_release:
+            try:
+                logger.info(f"Releasing inactive model: {model_id} (inactive for {(now - self._preloaded_models[model_id]).total_seconds() / 60:.1f} minutes)")
+                # We don't have an explicit "unload" API in Ollama, but we can remove it from our tracking
+                del self._preloaded_models[model_id]
+                released_models.append(model_id)
+            except Exception as e:
+                logger.error(f"Error releasing model {model_id}: {str(e)}")
+                
+        return released_models
             
     async def get_model_details(self, model_id: str) -> Dict[str, Any]:
         """Get detailed information about a specific Ollama model"""

app/config.py

@@ -151,7 +151,9 @@ DEFAULT_CONFIG = {
     "max_history_items": 100,
     "highlight_code": True,
     "auto_save": True,
-    "generate_dynamic_titles": True
+    "generate_dynamic_titles": True,
+    "ollama_model_preload": True,
+    "ollama_inactive_timeout_minutes": 30
 }
 
 def validate_config(config):

app/main.py

@@ -363,7 +363,13 @@ class SimpleChatApp(App): # Keep SimpleChatApp class definition
         self.selected_model = resolve_model_id(default_model_from_config)
         self.selected_style = CONFIG["default_style"] # Keep SimpleChatApp __init__
         self.initial_text = initial_text # Keep SimpleChatApp __init__
-        # Removed self.input_widget instance variable
+        
+        # Task for model cleanup
+        self._model_cleanup_task = None
+        
+        # Inactivity threshold in minutes before releasing model resources
+        # Read from config, default to 30 minutes
+        self.MODEL_INACTIVITY_THRESHOLD = CONFIG.get("ollama_inactive_timeout_minutes", 30)
 
     def compose(self) -> ComposeResult: # Modify SimpleChatApp compose
         """Create the simplified application layout."""
@@ -420,6 +426,11 @@ class SimpleChatApp(App): # Keep SimpleChatApp class definition
             pass # Silently ignore if widget not found yet
 
         self.update_app_info()  # Update the model info
+        
+        # Start the background task for model cleanup if model preloading is enabled
+        if CONFIG.get("ollama_model_preload", True):
+            self._model_cleanup_task = asyncio.create_task(self._check_inactive_models())
+            debug_log("Started background task for model cleanup")
 
         # Check API keys and services # Keep SimpleChatApp on_mount
         api_issues = [] # Keep SimpleChatApp on_mount
@@ -675,29 +686,98 @@ class SimpleChatApp(App): # Keep SimpleChatApp class definition
 
             # Determine title client and model based on available keys
             if OPENAI_API_KEY:
+                # For highest success rate, use OpenAI for title generation when available
                 from app.api.openai import OpenAIClient
                 title_client = await OpenAIClient.create()
                 title_model = "gpt-3.5-turbo"
                 debug_log("Using OpenAI for background title generation")
             elif ANTHROPIC_API_KEY:
+                # Next best option is Anthropic
                 from app.api.anthropic import AnthropicClient
                 title_client = await AnthropicClient.create()
                 title_model = "claude-3-haiku-20240307"
                 debug_log("Using Anthropic for background title generation")
             else:
                 # Fallback to the currently selected model's client if no API keys
+                # Get client type first to ensure we correctly identify Ollama models
+                from app.api.ollama import OllamaClient
                 selected_model_resolved = resolve_model_id(self.selected_model)
-                title_client = await BaseModelClient.get_client_for_model(selected_model_resolved)
-                title_model = selected_model_resolved
-                debug_log(f"Using selected model's client ({type(title_client).__name__}) for background title generation")
+                client_type = BaseModelClient.get_client_type_for_model(selected_model_resolved)
+                
+                # For Ollama models, special handling is required
+                if client_type == OllamaClient:
+                    debug_log(f"Title generation with Ollama model detected: {selected_model_resolved}")
+                    
+                    # Always try to use smalllm2:135m first, then fall back to other small models
+                    try:
+                        # Check if we have smalllm2:135m or other smaller models available
+                        ollama_client = await OllamaClient.create()
+                        available_models = await ollama_client.get_available_models()
+                        
+                        # Use smalllm2:135m if available (extremely small and fast)
+                        preferred_model = "smalllm2:135m"
+                        fallback_models = ["tinyllama", "gemma:2b", "phi3:mini", "llama3:8b", "orca-mini:3b", "phi2"]
+                        
+                        # First check for our preferred smallest model
+                        small_model_found = False
+                        if any(model["id"] == preferred_model for model in available_models):
+                            debug_log(f"Found optimal small model for title generation: {preferred_model}")
+                            title_model = preferred_model
+                            small_model_found = True
+                        
+                        # If not found, try fallbacks in order
+                        if not small_model_found:
+                            for model_name in fallback_models:
+                                if any(model["id"] == model_name for model in available_models):
+                                    debug_log(f"Found alternative small model for title generation: {model_name}")
+                                    title_model = model_name
+                                    small_model_found = True
+                                    break
+                                    
+                        if not small_model_found:
+                            # Use the current model if no smaller models found
+                            title_model = selected_model_resolved
+                            debug_log(f"No smaller models found, using current model: {title_model}")
+                            
+                        # Always create a fresh client instance to avoid interference with model preloading
+                        title_client = ollama_client
+                        debug_log(f"Created dedicated Ollama client for title generation with model: {title_model}")
+                    except Exception as e:
+                        debug_log(f"Error finding optimized Ollama model for title generation: {str(e)}")
+                        # Fallback to standard approach
+                        title_client = await OllamaClient.create()
+                        title_model = selected_model_resolved
+                else:
+                    # For other providers, use normal client acquisition
+                    title_client = await BaseModelClient.get_client_for_model(selected_model_resolved)
+                    title_model = selected_model_resolved
+                    debug_log(f"Using selected model's client ({type(title_client).__name__}) for background title generation")
 
             if not title_client or not title_model:
                 raise Exception("Could not determine a client/model for title generation.")
 
             # Call the utility function
             from app.utils import generate_conversation_title # Import locally if needed
-            new_title = await generate_conversation_title(content, title_model, title_client)
-            debug_log(f"Background generated title: {new_title}")
+            
+            # Add timeout handling for title generation to prevent hangs
+            try:
+                # Create a task with timeout
+                import asyncio
+                title_generation_task = asyncio.create_task(
+                    generate_conversation_title(content, title_model, title_client)
+                )
+                
+                # Wait for completion with timeout (30 seconds)
+                new_title = await asyncio.wait_for(title_generation_task, timeout=30)
+                debug_log(f"Background generated title: {new_title}")
+            except asyncio.TimeoutError:
+                debug_log("Title generation timed out after 30 seconds")
+                # Use default title in case of timeout
+                new_title = f"Conversation ({datetime.now().strftime('%Y-%m-%d %H:%M')})"
+                # Try to cancel the task
+                if not title_generation_task.done():
+                    title_generation_task.cancel()
+                    debug_log("Cancelled timed out title generation task")
 
             # Check if title generation returned the default or a real title
             if new_title and not new_title.startswith("Conversation ("):
@@ -718,8 +798,8 @@ class SimpleChatApp(App): # Keep SimpleChatApp class definition
                         title_widget.update(new_title)
                         self.current_conversation.title = new_title # Update local object too
                         log(f"Background title update successful: {new_title}")
-                        # Maybe a subtle notification? Optional.
-                        # self.notify(f"Title set: {new_title}", severity="information", timeout=2)
+                        # Subtle notification to show title was updated
+                        self.notify(f"Conversation titled: {new_title}", severity="information", timeout=2)
                     else:
                         log("Conversation changed before background title update could apply.")
                 else:
@@ -1226,6 +1306,94 @@ class SimpleChatApp(App): # Keep SimpleChatApp class definition
             log(f"Stored selected provider: {self.selected_provider} for model: {self.selected_model}")
         
         self.update_app_info()  # Update the displayed model info
+        
+        # Preload the model if it's an Ollama model and preloading is enabled
+        if self.selected_provider == "ollama" and CONFIG.get("ollama_model_preload", True):
+            # Start the background task to preload the model
+            debug_log(f"Starting background task to preload Ollama model: {self.selected_model}")
+            asyncio.create_task(self._preload_ollama_model(self.selected_model))
+    
+    async def _preload_ollama_model(self, model_id: str) -> None:
+        """Preload an Ollama model in the background"""
+        from app.api.ollama import OllamaClient
+        
+        debug_log(f"Preloading Ollama model: {model_id}")
+        # Show a subtle notification to the user
+        self.notify("Preparing model for use...", severity="information", timeout=3)
+        
+        try:
+            # Initialize the client
+            client = await OllamaClient.create()
+            
+            # Update the loading indicator to show model loading
+            loading = self.query_one("#loading-indicator")
+            loading.remove_class("hidden")
+            loading.add_class("model-loading")
+            loading.update(f"⚙️ Loading Ollama model...")
+            
+            # Preload the model
+            success = await client.preload_model(model_id)
+            
+            # Hide the loading indicator
+            loading.add_class("hidden")
+            loading.remove_class("model-loading")
+            
+            if success:
+                debug_log(f"Successfully preloaded model: {model_id}")
+                self.notify(f"Model ready for use", severity="success", timeout=2)
+            else:
+                debug_log(f"Failed to preload model: {model_id}")
+                # No need to notify the user about failure - will happen naturally on first use
+        except Exception as e:
+            debug_log(f"Error preloading model: {str(e)}")
+            # Make sure to hide the loading indicator
+            try:
+                loading = self.query_one("#loading-indicator")
+                loading.add_class("hidden")
+                loading.remove_class("model-loading")
+            except Exception:
+                pass
+                
+    async def _check_inactive_models(self) -> None:
+        """Background task to check for and release inactive models"""
+        from app.api.ollama import OllamaClient
+        
+        # How often to check for inactive models (in seconds)
+        CHECK_INTERVAL = 600  # 10 minutes
+        
+        debug_log(f"Starting inactive model check task with interval {CHECK_INTERVAL}s")
+        
+        try:
+            while True:
+                await asyncio.sleep(CHECK_INTERVAL)
+                
+                debug_log("Checking for inactive models...")
+                
+                try:
+                    # Initialize the client
+                    client = await OllamaClient.create()
+                    
+                    # Get the threshold from instance variable
+                    threshold = getattr(self, "MODEL_INACTIVITY_THRESHOLD", 30)
+                    
+                    # Check and release inactive models
+                    released_models = await client.release_inactive_models(threshold)
+                    
+                    if released_models:
+                        debug_log(f"Released {len(released_models)} inactive models: {released_models}")
+                    else:
+                        debug_log("No inactive models to release")
+                        
+                except Exception as e:
+                    debug_log(f"Error checking for inactive models: {str(e)}")
+                    # Continue loop even if this check fails
+                    
+        except asyncio.CancelledError:
+            debug_log("Model cleanup task cancelled")
+            # Normal task cancellation, clean exit
+        except Exception as e:
+            debug_log(f"Unexpected error in model cleanup task: {str(e)}")
+            # Log but don't crash
 
     def on_style_selector_style_selected(self, event: StyleSelector.StyleSelected) -> None: # Keep SimpleChatApp on_style_selector_style_selected
         """Handle style selection""" # Keep SimpleChatApp on_style_selector_style_selected docstring

app/utils.py

@@ -32,6 +32,11 @@ async def generate_conversation_title(message: str, model: str, client: Any) ->
     
     # Try-except the entire function to ensure we always return a title
     try:
+        # Check if we're using an Ollama client
+        from app.api.ollama import OllamaClient
+        is_ollama_client = isinstance(client, OllamaClient)
+        debug_log(f"Client is Ollama: {is_ollama_client}")
+        
         # Pick a reliable title generation model - prefer OpenAI if available
         from app.config import OPENAI_API_KEY, ANTHROPIC_API_KEY
         
@@ -46,16 +51,22 @@ async def generate_conversation_title(message: str, model: str, client: Any) ->
             title_model = "claude-3-haiku-20240307"
             debug_log("Using Anthropic for title generation")
         else:
-            # Use the passed client if no API keys available
-            title_client = client
-            title_model = model
-            debug_log(f"Using provided {type(client).__name__} for title generation")
+            # For Ollama clients, ensure we have a clean instance to avoid conflicts with preloaded models
+            if is_ollama_client:
+                debug_log("Creating fresh Ollama client instance for title generation")
+                title_client = await OllamaClient.create()
+                title_model = model
+            else:
+                # Use the passed client for other providers
+                title_client = client
+                title_model = model
+            debug_log(f"Using {type(title_client).__name__} for title generation with model {title_model}")
         
         # Create a special prompt for title generation
         title_prompt = [
             {
                 "role": "system", 
-                "content": "Generate a brief, descriptive title (maximum 40 characters) for a conversation that starts with the following message. Return only the title text with no additional explanation or formatting."
+                "content": "Generate a brief, descriptive title (maximum 40 characters) for a conversation that starts with the following message. ONLY output the title text. DO NOT include phrases like 'Sure, here's a title' or any additional formatting, explanation, or quotes."
             },
             {
                 "role": "user",
@@ -85,12 +96,31 @@ async def generate_conversation_title(message: str, model: str, client: Any) ->
                 max_tokens=60
             )
         
-        # Sanitize the title
+        # Sanitize the title - remove quotes, extra spaces and unwanted prefixes
         title = title.strip().strip('"\'').strip()
+        
+        # Remove common LLM prefixes like "Title:", "Sure, here's a title:", etc.
+        prefixes_to_remove = [
+            "title:", "here's a title:", "here is a title:", 
+            "a title for this conversation:", "sure,", "certainly,",
+            "the title is:", "suggested title:"
+        ]
+        
+        # Case-insensitive prefix removal
+        title_lower = title.lower()
+        for prefix in prefixes_to_remove:
+            if title_lower.startswith(prefix):
+                title = title[len(prefix):].strip()
+                title_lower = title.lower()  # Update lowercase version after removal
+        
+        # Remove any remaining quotes
+        title = title.strip('"\'').strip()
+        
+        # Enforce length limit
         if len(title) > 40:
             title = title[:37] + "..."
             
-        debug_log(f"Generated title: {title}")
+        debug_log(f"Generated title (after sanitization): {title}")
         return title
         
     except Exception as e:

{chat_console-0.4.0.dist-info → chat_console-0.4.3.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: chat-console
-Version: 0.4.0
+Version: 0.4.3
 Summary: A command-line interface for chatting with LLMs, storing chats and (future) rag interactions
 Home-page: https://github.com/wazacraftrfid/chat-console
 Author: Johnathan Greenaway

{chat_console-0.4.0.dist-info → chat_console-0.4.3.dist-info}/RECORD

@@ -1,13 +1,13 @@
-app/__init__.py,sha256=wlOzzHK2mYbpcY6MSNp2l_vlKYdCrr8nzakp8kl3ffU,130
-app/config.py,sha256=tuTdJfmRsGx2-6PwkpbIsuP2hnB3ZPXRJdCvgzFd8PQ,8391
+app/__init__.py,sha256=T3d41tTB1sDy6ix7bg43dp4zvNqqJku0JGlwMGKFGHo,130
+app/config.py,sha256=F-0hO3NT5kRJxZelGLxaeUmnwx8i0LPHzYtNftL6CwM,8468
 app/database.py,sha256=nt8CVuDpy6zw8mOYqDcfUmNw611t7Ln7pz22M0b6-MI,9967
-app/main.py,sha256=WOcMP6yRwoEzftTSHf0e3zVK1aEuBgKMAsNbzHyKgiA,77427
+app/main.py,sha256=8UU9GcPJINu_TmbKKKFBZXIgLHNDf6vabyupKjj3Img,86297
 app/models.py,sha256=4-y9Lytay2exWPFi0FDlVeRL3K2-I7E-jBqNzTfokqY,2644
-app/utils.py,sha256=G8e7ucCuIa-M8tpPDEd9PeWKIb8hN1FPpZnk_RHeRKo,40283
+app/utils.py,sha256=-L38KGP8TlVl5vtZl5QgTiEAdhLcDsIXm7e62nnXgP8,41765
 app/api/__init__.py,sha256=A8UL84ldYlv8l7O-yKzraVFcfww86SgWfpl4p7R03-w,62
 app/api/anthropic.py,sha256=uInwNvGLJ_iPUs4BjdwaqXTU6NfmK1SzX7498Pt44fI,10667
 app/api/base.py,sha256=valBWV5So76r8tjrgU5-sLfY73WaViTrszdCy8Rimjo,10314
-app/api/ollama.py,sha256=eFG24nI2MlF57z9EHiA97v02NgFJ0kxaPUX26xAXFsg,66154
+app/api/ollama.py,sha256=364PcXoPLJq9jLMF-HhPyQvaBp87U6FzNHDWx4g_Cvc,76925
 app/api/openai.py,sha256=XuHJHpD7tN_ZHLkRpNUcL1VxTtsXOVk1hDPXX8JnBxQ,15322
 app/ui/__init__.py,sha256=RndfbQ1Tv47qdSiuQzvWP96lPS547SDaGE-BgOtiP_w,55
 app/ui/chat_interface.py,sha256=oSDZi0Jgj_L8WnBh1RuJpIeIcN-RQ38CNejwsXiWTVg,18267
@@ -16,9 +16,9 @@ app/ui/model_browser.py,sha256=pdblLVkdyVF0_Bo02bqbErGAtieyH-y6IfhMOPEqIso,71124
 app/ui/model_selector.py,sha256=2G0TOXfcNodrXZOhLeaJJ2iG3Nck4c_NN1AvUAmaF3M,19172
 app/ui/search.py,sha256=b-m14kG3ovqW1-i0qDQ8KnAqFJbi5b1FLM9dOnbTyIs,9763
 app/ui/styles.py,sha256=04AhPuLrOd2yenfRySFRestPeuTPeMLzhmMB67NdGvw,5615
-chat_console-0.4.0.dist-info/licenses/LICENSE,sha256=srHZ3fvcAuZY1LHxE7P6XWju2njRCHyK6h_ftEbzxSE,1057
-chat_console-0.4.0.dist-info/METADATA,sha256=5vwV6wZ4lkp7Puo12-JP6Do3Ap3ZG7Z75lnsn7VGqnI,3810
-chat_console-0.4.0.dist-info/WHEEL,sha256=wXxTzcEDnjrTwFYjLPcsW_7_XihufBwmpiBeiXNBGEA,91
-chat_console-0.4.0.dist-info/entry_points.txt,sha256=kkVdEc22U9PAi2AeruoKklfkng_a_aHAP6VRVwrAD7c,67
-chat_console-0.4.0.dist-info/top_level.txt,sha256=io9g7LCbfmTG1SFKgEOGXmCFB9uMP2H5lerm0HiHWQE,4
-chat_console-0.4.0.dist-info/RECORD,,
+chat_console-0.4.3.dist-info/licenses/LICENSE,sha256=srHZ3fvcAuZY1LHxE7P6XWju2njRCHyK6h_ftEbzxSE,1057
+chat_console-0.4.3.dist-info/METADATA,sha256=y01SPyzTsYIyCbc5gEpCx6dlQ_ROG16T1gTZPOScia8,3810
+chat_console-0.4.3.dist-info/WHEEL,sha256=GHB6lJx2juba1wDgXDNlMTyM13ckjBMKf-OnwgKOCtA,91
+chat_console-0.4.3.dist-info/entry_points.txt,sha256=kkVdEc22U9PAi2AeruoKklfkng_a_aHAP6VRVwrAD7c,67
+chat_console-0.4.3.dist-info/top_level.txt,sha256=io9g7LCbfmTG1SFKgEOGXmCFB9uMP2H5lerm0HiHWQE,4
+chat_console-0.4.3.dist-info/RECORD,,

{chat_console-0.4.0.dist-info → chat_console-0.4.3.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (80.1.0)
+Generator: setuptools (80.3.0)
 Root-Is-Purelib: true
 Tag: py3-none-any