chat-console 0.3.5__py3-none-any.whl → 0.3.7__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__ = "0.3.5"
+__version__ = "0.3.7"

app/api/base.py

@@ -38,27 +38,41 @@ class BaseModelClient(ABC):
         
         logger = logging.getLogger(__name__)
         
+        # Safety check for None or empty string
+        if not model_name:
+            logger.warning("Empty model name passed to get_client_type_for_model")
+            return None
+            
         # Get model info and provider
         model_info = CONFIG["available_models"].get(model_name)
         model_name_lower = model_name.lower()
         
+        # Debug log the model name
+        logger.info(f"Getting client type for model: {model_name}")
+        
         # If model is in config, use its provider
         if model_info:
             provider = model_info["provider"]
+            logger.info(f"Found model in config with provider: {provider}")
         # For custom models, try to infer provider
         else:
-            # First try Ollama for known model names or if selected from Ollama UI
-            if (any(name in model_name_lower for name in ["llama", "mistral", "codellama", "gemma"]) or
-                model_name in [m["id"] for m in CONFIG.get("ollama_models", [])]):
-                provider = "ollama"
-            # Then try other providers
-            elif any(name in model_name_lower for name in ["gpt", "text-", "davinci"]):
+            # First check for OpenAI models - these should ALWAYS use OpenAI client
+            if any(name in model_name_lower for name in ["gpt", "text-", "davinci"]):
                 provider = "openai"
+                logger.info(f"Identified as OpenAI model: {model_name}")
+            # Then check for Anthropic models - these should ALWAYS use Anthropic client
             elif any(name in model_name_lower for name in ["claude", "anthropic"]):
                 provider = "anthropic"
+                logger.info(f"Identified as Anthropic model: {model_name}")
+            # Then try Ollama for known model names or if selected from Ollama UI
+            elif (any(name in model_name_lower for name in ["llama", "mistral", "codellama", "gemma"]) or
+                  model_name in [m["id"] for m in CONFIG.get("ollama_models", [])]):
+                provider = "ollama"
+                logger.info(f"Identified as Ollama model: {model_name}")
             else:
                 # Default to Ollama for unknown models
                 provider = "ollama"
+                logger.info(f"Unknown model type, defaulting to Ollama: {model_name}")
         
         # Return appropriate client class
         if provider == "ollama":
@@ -81,6 +95,14 @@ class BaseModelClient(ABC):
         
         logger = logging.getLogger(__name__)
         
+        # Safety check for None or empty string
+        if not model_name:
+            logger.warning("Empty model name passed to get_client_for_model")
+            raise ValueError("Model name cannot be empty")
+            
+        # Log the model name we're getting a client for
+        logger.info(f"Getting client for model: {model_name}")
+        
         # Get model info and provider
         model_info = CONFIG["available_models"].get(model_name)
         model_name_lower = model_name.lower()
@@ -88,31 +110,35 @@ class BaseModelClient(ABC):
         # If model is in config, use its provider
         if model_info:
             provider = model_info["provider"]
+            logger.info(f"Found model in config with provider: {provider}")
             if not AVAILABLE_PROVIDERS[provider]:
                 raise Exception(f"Provider '{provider}' is not available. Please check your configuration.")
         # For custom models, try to infer provider
         else:
-            # First try Ollama for known model names or if selected from Ollama UI
-            if (any(name in model_name_lower for name in ["llama", "mistral", "codellama", "gemma"]) or
-                model_name in [m["id"] for m in CONFIG.get("ollama_models", [])]):
-                if not AVAILABLE_PROVIDERS["ollama"]:
-                    raise Exception("Ollama server is not running. Please start Ollama and try again.")
-                provider = "ollama"
-                logger.info(f"Using Ollama for model: {model_name}")
-            # Then try other providers if they're available
-            elif any(name in model_name_lower for name in ["gpt", "text-", "davinci"]):
+            # First check for OpenAI models - these should ALWAYS use OpenAI client
+            if any(name in model_name_lower for name in ["gpt", "text-", "davinci"]):
                 if not AVAILABLE_PROVIDERS["openai"]:
                     raise Exception("OpenAI API key not found. Please set OPENAI_API_KEY environment variable.")
                 provider = "openai"
+                logger.info(f"Identified as OpenAI model: {model_name}")
+            # Then check for Anthropic models - these should ALWAYS use Anthropic client
             elif any(name in model_name_lower for name in ["claude", "anthropic"]):
                 if not AVAILABLE_PROVIDERS["anthropic"]:
                     raise Exception("Anthropic API key not found. Please set ANTHROPIC_API_KEY environment variable.")
                 provider = "anthropic"
+                logger.info(f"Identified as Anthropic model: {model_name}")
+            # Then try Ollama for known model names or if selected from Ollama UI
+            elif (any(name in model_name_lower for name in ["llama", "mistral", "codellama", "gemma"]) or
+                  model_name in [m["id"] for m in CONFIG.get("ollama_models", [])]):
+                if not AVAILABLE_PROVIDERS["ollama"]:
+                    raise Exception("Ollama server is not running. Please start Ollama and try again.")
+                provider = "ollama"
+                logger.info(f"Identified as Ollama model: {model_name}")
             else:
                 # Default to Ollama for unknown models
                 if AVAILABLE_PROVIDERS["ollama"]:
                     provider = "ollama"
-                    logger.info(f"Defaulting to Ollama for unknown model: {model_name}")
+                    logger.info(f"Unknown model type, defaulting to Ollama: {model_name}")
                 else:
                     raise Exception(f"Unknown model: {model_name}")
         

app/main.py

@@ -707,10 +707,18 @@ class SimpleChatApp(App): # Keep SimpleChatApp class definition
                         else:
                             raise Exception("No valid API clients available for title generation")
 
-                # Generate title
+                # Generate title - make sure we're using the right client for the model
                 print(f"Calling generate_conversation_title with model: {model}")
                 log(f"Calling generate_conversation_title with model: {model}")
-                debug_log(f"Calling generate_conversation_title with model: {model}")
+                debug_log(f"Calling generate_conversation_title with model: {model}, client type: {type(client).__name__}")
+                
+                # Double-check that we're using the right client for this model
+                expected_client_type = BaseModelClient.get_client_type_for_model(model)
+                if expected_client_type and not isinstance(client, expected_client_type):
+                    debug_log(f"Warning: Client type mismatch. Expected {expected_client_type.__name__}, got {type(client).__name__}")
+                    debug_log("Creating new client with correct type")
+                    client = await BaseModelClient.get_client_for_model(model)
+                
                 title = await generate_conversation_title(content, model, client)
                 debug_log(f"Generated title: {title}")
                 log(f"Generated title: {title}")
@@ -729,11 +737,9 @@ class SimpleChatApp(App): # Keep SimpleChatApp class definition
                 # Update conversation object
                 self.current_conversation.title = title
                 
-                # IMPORTANT: Save the successful model for consistency
-                # If the title was generated with a different model than initially selected,
-                # update the selected_model to match so the response uses the same model
-                debug_log(f"Using same model for chat response: '{model}'")
-                self.selected_model = model
+                # DO NOT update the selected model here - keep the user's original selection
+                # This was causing issues with model mixing
+                debug_log(f"Keeping original selected model: '{self.selected_model}'")
 
                 self.notify(f"Conversation title set to: {title}", severity="information", timeout=3)
 
@@ -805,17 +811,23 @@ class SimpleChatApp(App): # Keep SimpleChatApp class definition
             style = self.selected_style
             
             debug_log(f"Using model: '{model}', style: '{style}'")
+            
+            # Determine the expected client type for this model
+            expected_client_type = BaseModelClient.get_client_type_for_model(model)
+            debug_log(f"Expected client type for {model}: {expected_client_type.__name__ if expected_client_type else 'None'}")
 
             # Ensure we have a valid model
             if not model:
                 debug_log("Model is empty, selecting a default model")
-                # Same fallback logic as in autotitling - this ensures consistency
+                # Check which providers are available and select an appropriate default
                 if OPENAI_API_KEY:
                     model = "gpt-3.5-turbo"
-                    debug_log("Falling back to OpenAI gpt-3.5-turbo")
+                    expected_client_type = BaseModelClient.get_client_type_for_model(model)
+                    debug_log(f"Falling back to OpenAI gpt-3.5-turbo with client type {expected_client_type.__name__ if expected_client_type else 'None'}")
                 elif ANTHROPIC_API_KEY:
-                    model = "claude-instant-1.2"
-                    debug_log("Falling back to Anthropic claude-instant-1.2")
+                    model = "claude-3-haiku-20240307"  # Updated to newer Claude model
+                    expected_client_type = BaseModelClient.get_client_type_for_model(model)
+                    debug_log(f"Falling back to Anthropic Claude 3 Haiku with client type {expected_client_type.__name__ if expected_client_type else 'None'}")
                 else:
                     # Check for a common Ollama model
                     try:
@@ -826,11 +838,13 @@ class SimpleChatApp(App): # Keep SimpleChatApp class definition
                             model = models[0].get("id", "llama3")
                         else:
                             model = "llama3"  # Common default
-                        debug_log(f"Falling back to Ollama model: {model}")
+                        expected_client_type = BaseModelClient.get_client_type_for_model(model)
+                        debug_log(f"Falling back to Ollama model: {model} with client type {expected_client_type.__name__ if expected_client_type else 'None'}")
                     except Exception as ollama_err:
                         debug_log(f"Error getting Ollama models: {str(ollama_err)}")
                         model = "llama3"  # Final fallback
-                        debug_log("Final fallback to llama3")
+                        expected_client_type = BaseModelClient.get_client_type_for_model(model)
+                        debug_log(f"Final fallback to llama3 with client type {expected_client_type.__name__ if expected_client_type else 'None'}")
 
             # Convert messages to API format with enhanced error checking
             api_messages = []
@@ -930,70 +944,29 @@ class SimpleChatApp(App): # Keep SimpleChatApp class definition
                 if not self.is_generating:
                     debug_log("update_ui called but is_generating is False, returning.")
                     return
-
-                # Make last_refresh_time accessible in inner scope
-                nonlocal last_refresh_time
                 
                 async with update_lock:
                     try:
                         # Clear thinking indicator on first content
                         if assistant_message.content == "Thinking...":
                             debug_log("First content received, clearing 'Thinking...'")
+                            print("First content received, clearing 'Thinking...'")
                             assistant_message.content = ""
 
                         # Update the message object with the full content
                         assistant_message.content = content
 
-                        # Update UI with the content - the MessageDisplay will now handle its own refresh
-                        # This is a critical change that ensures content is immediately visible
+                        # Update UI with the content
                         await message_display.update_content(content)
                         
-                        # CRITICAL: Force immediate UI refresh after EVERY update 
-                        # This ensures we don't need a second Enter press to see content
+                        # Simple refresh approach - just force a layout refresh
                         self.refresh(layout=True)
-                        
-                        # Always scroll after each update to ensure visibility
                         messages_container.scroll_end(animate=False)
                         
-                        # For longer responses, we can throttle the heavy refreshes
-                        # to reduce visual jitter, but still do light refreshes for every update
-                        content_length = len(content)
-                        
-                        # Define key refresh points that require more thorough updates
-                        new_paragraph = content.endswith("\n") and content.count("\n") > 0
-                        code_block = "```" in content
-                        needs_thorough_refresh = (
-                            content_length < 30 or       # Very aggressive for short responses
-                            content_length % 16 == 0 or  # More frequent periodic updates 
-                            new_paragraph or             # Refresh on paragraph breaks
-                            code_block                   # Refresh when code blocks are detected
-                        )
-                        
-                        # Check if it's been enough time since last heavy refresh
-                        # Reduced from 200ms to 100ms for more responsive UI
-                        current_time = time.time()
-                        time_since_refresh = current_time - last_refresh_time
-                        
-                        if needs_thorough_refresh and time_since_refresh > 0.1:
-                            # Store the time we did the heavy refresh
-                            last_refresh_time = current_time
-                            
-                            # Ensure content is visible with an aggressive, guaranteed update sequence
-                            # 1. Scroll to ensure visibility
-                            messages_container.scroll_end(animate=False)
-                            
-                            # 2. Force a comprehensive refresh with layout recalculation
-                            self.refresh(layout=True)
-                            
-                            # 3. Small delay for rendering
-                            await asyncio.sleep(0.01)
-                            
-                            # 4. Another scroll to account for any layout changes
-                            messages_container.scroll_end(animate=False)
-                            
                     except Exception as e:
                         debug_log(f"Error updating UI: {str(e)}")
                         log.error(f"Error updating UI: {str(e)}")
+                        print(f"Error updating UI: {str(e)}")
 
             # --- Remove the inner run_generation_worker function ---
 

app/ui/chat_interface.py

@@ -132,9 +132,6 @@ class MessageDisplay(Static): # Inherit from Static instead of RichLog
         # This avoids text reflowing as new tokens arrive
         formatted_content = self._format_content(content)
         
-        # Print debug info to console
-        print(f"MessageDisplay.update_content: Updating with {len(content)} chars")
-        
         # Use a direct update that forces refresh - critical fix for streaming
         # This ensures content is immediately visible
         self.update(formatted_content, refresh=True)
@@ -143,11 +140,7 @@ class MessageDisplay(Static): # Inherit from Static instead of RichLog
         try:
             # Always force app refresh for every update
             if self.app:
-                # First do a quick refresh without layout recalculation
-                self.app.refresh(layout=False)
-                
-                # Then do a full layout refresh to ensure content is visible
-                await asyncio.sleep(0.01)
+                # Force a full layout refresh to ensure content is visible
                 self.app.refresh(layout=True)
                 
                 # Find the messages container and scroll to end
@@ -155,17 +148,10 @@ class MessageDisplay(Static): # Inherit from Static instead of RichLog
                 for container in containers:
                     if hasattr(container, 'scroll_end'):
                         container.scroll_end(animate=False)
-                        
-                # Force another refresh after scrolling
-                await asyncio.sleep(0.01)
-                self.app.refresh(layout=True)
         except Exception as e:
             # Log the error and fallback to local refresh
             print(f"Error refreshing app: {str(e)}")
             self.refresh(layout=True)
-            
-        # Small delay to allow UI to update
-        await asyncio.sleep(0.03)  # Increased delay for better rendering
         
     def _format_content(self, content: str) -> str:
         """Format message content with timestamp and handle markdown links"""

app/utils.py

@@ -20,33 +20,63 @@ logger = logging.getLogger(__name__)
 
 async def generate_conversation_title(message: str, model: str, client: Any) -> str:
     """Generate a descriptive title for a conversation based on the first message"""
+    try:
+        from app.main import debug_log
+    except ImportError:
+        debug_log = lambda msg: None
+    
+    debug_log(f"Starting title generation with model: {model}, client type: {type(client).__name__}")
+    
     # --- Choose a specific, reliable model for title generation ---
-    # Prefer Haiku if Anthropic is available, otherwise fallback
+    # First, determine if we have a valid client
+    if client is None:
+        debug_log("Client is None, will use default title")
+        return f"Conversation ({datetime.now().strftime('%Y-%m-%d %H:%M')})"
+    
+    # Determine the best model to use for title generation
     title_model_id = None
-    if client and isinstance(client, anthropic.AsyncAnthropic): # Check if the passed client is Anthropic
-        # Check if Haiku is listed in the client's available models (more robust)
-        available_anthropic_models = client.get_available_models()
-        haiku_id = "claude-3-haiku-20240307"
-        if any(m["id"] == haiku_id for m in available_anthropic_models):
-             title_model_id = haiku_id
-             logger.info(f"Using Anthropic Haiku for title generation: {title_model_id}")
-        else:
-             # If Haiku not found, try Sonnet
-             sonnet_id = "claude-3-sonnet-20240229"
-             if any(m["id"] == sonnet_id for m in available_anthropic_models):
-                  title_model_id = sonnet_id
-                  logger.info(f"Using Anthropic Sonnet for title generation: {title_model_id}")
-             else:
-                  logger.warning(f"Neither Haiku nor Sonnet found in Anthropic client's list. Falling back.")
-
-    # Fallback logic if no specific Anthropic model was found or client is not Anthropic
+    
+    # Check if client is Anthropic
+    is_anthropic = 'anthropic' in str(type(client)).lower()
+    if is_anthropic:
+        debug_log("Using Anthropic client for title generation")
+        # Try to get available models safely
+        try:
+            available_anthropic_models = client.get_available_models()
+            debug_log(f"Found {len(available_anthropic_models)} Anthropic models")
+            
+            # Try Claude 3 Haiku first (fastest)
+            haiku_id = "claude-3-haiku-20240307"
+            if any(m.get("id") == haiku_id for m in available_anthropic_models):
+                title_model_id = haiku_id
+                debug_log(f"Using Anthropic Haiku for title generation: {title_model_id}")
+            else:
+                # If Haiku not found, try Sonnet
+                sonnet_id = "claude-3-sonnet-20240229"
+                if any(m.get("id") == sonnet_id for m in available_anthropic_models):
+                    title_model_id = sonnet_id
+                    debug_log(f"Using Anthropic Sonnet for title generation: {title_model_id}")
+                else:
+                    debug_log("Neither Haiku nor Sonnet found in Anthropic models list")
+        except Exception as e:
+            debug_log(f"Error getting Anthropic models: {str(e)}")
+
+    # Check if client is OpenAI
+    is_openai = 'openai' in str(type(client)).lower()
+    if is_openai and not title_model_id:
+        debug_log("Using OpenAI client for title generation")
+        # Use GPT-3.5 for title generation (fast and cost-effective)
+        title_model_id = "gpt-3.5-turbo"
+        debug_log(f"Using OpenAI model for title generation: {title_model_id}")
+    
+    # Fallback logic if no specific model was found
     if not title_model_id:
-        # Use the originally passed model (user's selected chat model) as the final fallback
+        # Use the originally passed model as the final fallback
         title_model_id = model
-        logger.warning(f"Falling back to originally selected model for title generation: {title_model_id}")
-        # Consider adding fallbacks to OpenAI/Ollama here if needed based on config/availability
-
+        debug_log(f"Falling back to originally selected model for title generation: {title_model_id}")
+    
     logger.info(f"Generating title for conversation using model: {title_model_id}")
+    debug_log(f"Final model selected for title generation: {title_model_id}")
 
     # Create a special prompt for title generation
     title_prompt = [
@@ -65,36 +95,44 @@ async def generate_conversation_title(message: str, model: str, client: Any) ->
     
     while tries > 0:
         try:
-            # Generate a title using the same model but with a separate request
-            # Assuming client has a method like generate_completion or similar
-            # Adjust the method call based on the actual client implementation
+            debug_log(f"Attempt {3-tries} to generate title")
+            # First try generate_completion if available
             if hasattr(client, 'generate_completion'):
-                title = await client.generate_completion(
-                    messages=title_prompt,
-                    model=title_model_id, # Use the chosen title model
-                    temperature=0.7,
-                    max_tokens=60  # Titles should be short
-                )
-            elif hasattr(client, 'generate_stream'): # Fallback or alternative method?
-                 # If generate_completion isn't available, maybe adapt generate_stream?
-                 # This part needs clarification based on the client's capabilities.
-                 # For now, let's assume a hypothetical non-streaming call or adapt stream
-                 # Simplified adaptation: collect stream chunks
-                 title_chunks = []
-                 try:
-                     # Use the chosen title model here too
-                     async for chunk in client.generate_stream(title_prompt, title_model_id, style=""):
-                         if chunk is not None:  # Ensure we only process non-None chunks
-                             title_chunks.append(chunk)
-                     title = "".join(title_chunks)
-                     # If we didn't get any content, use a default
-                     if not title.strip():
-                         title = f"Conversation ({datetime.now().strftime('%Y-%m-%d %H:%M')})"
-                 except Exception as stream_error:
-                     logger.error(f"Error during title stream processing: {str(stream_error)}")
-                     title = f"Conversation ({datetime.now().strftime('%Y-%m-%d %H:%M')})"
+                debug_log("Using generate_completion method")
+                try:
+                    title = await client.generate_completion(
+                        messages=title_prompt,
+                        model=title_model_id,
+                        temperature=0.7,
+                        max_tokens=60  # Titles should be short
+                    )
+                    debug_log(f"Title generated successfully: {title}")
+                except Exception as completion_error:
+                    debug_log(f"Error in generate_completion: {str(completion_error)}")
+                    raise  # Re-raise to be caught by outer try/except
+            # Fall back to generate_stream if completion not available
+            elif hasattr(client, 'generate_stream'):
+                debug_log("Using generate_stream method")
+                title_chunks = []
+                try:
+                    async for chunk in client.generate_stream(title_prompt, title_model_id, style=""):
+                        if chunk is not None:
+                            title_chunks.append(chunk)
+                            debug_log(f"Received chunk of length: {len(chunk)}")
+                    
+                    title = "".join(title_chunks)
+                    debug_log(f"Combined title from chunks: {title}")
+                    
+                    # If we didn't get any content, use a default
+                    if not title.strip():
+                        debug_log("Empty title received, using default")
+                        title = f"Conversation ({datetime.now().strftime('%Y-%m-%d %H:%M')})"
+                except Exception as stream_error:
+                    debug_log(f"Error during title stream processing: {str(stream_error)}")
+                    raise  # Re-raise to be caught by outer try/except
             else:
-                 raise NotImplementedError("Client does not support a suitable method for title generation.")
+                debug_log("Client does not support any title generation method")
+                raise NotImplementedError("Client does not support a suitable method for title generation.")
 
             # Sanitize and limit the title
             title = title.strip().strip('"\'').strip()
@@ -102,20 +140,23 @@ async def generate_conversation_title(message: str, model: str, client: Any) ->
                 title = title[:37] + "..."
                 
             logger.info(f"Generated title: {title}")
-            return title # Return successful title
+            debug_log(f"Final sanitized title: {title}")
+            return title  # Return successful title
             
         except Exception as e:
             last_error = str(e)
-            logger.error(f"Error generating title (tries left: {tries - 1}): {last_error}")
+            debug_log(f"Error generating title (tries left: {tries-1}): {last_error}")
+            logger.error(f"Error generating title (tries left: {tries-1}): {last_error}")
             tries -= 1
-            if tries > 0: # Only sleep if there are more retries
+            if tries > 0:  # Only sleep if there are more retries
                 await asyncio.sleep(1)  # Small delay before retry
     
-    # If all retries fail, log the last error and return a default title
+    # If all retries fail, log the error and return a default title
+    debug_log(f"Failed to generate title after multiple retries. Using default title.")
     logger.error(f"Failed to generate title after multiple retries. Last error: {last_error}")
     return f"Conversation ({datetime.now().strftime('%Y-%m-%d %H:%M')})"
 
-# Make this the worker function directly
+# Worker function for streaming response generation
 async def generate_streaming_response(
     app: 'SimpleChatApp',
     messages: List[Dict],
@@ -136,10 +177,12 @@ async def generate_streaming_response(
     logger.info(f"Starting streaming response with model: {model}")
     debug_log(f"Starting streaming response with model: '{model}', client type: {type(client).__name__}")
 
+    # Validate messages
     if not messages:
         debug_log("Error: messages list is empty")
         raise ValueError("Messages list cannot be empty")
 
+    # Ensure all messages have required fields
     for i, msg in enumerate(messages):
         try:
             debug_log(f"Message {i}: role={msg.get('role', 'missing')}, content_len={len(msg.get('content', ''))}")
@@ -157,14 +200,14 @@ async def generate_streaming_response(
             }
             debug_log(f"Repaired message {i}")
 
-    import time
-
+    # Initialize variables for response tracking
     full_response = ""
     buffer = []
     last_update = time.time()
     update_interval = 0.05  # Reduced interval for more frequent updates
 
     try:
+        # Validate client
         if client is None:
             debug_log("Error: client is None, cannot proceed with streaming")
             raise ValueError("Model client is None, cannot proceed with streaming")
@@ -173,9 +216,15 @@ async def generate_streaming_response(
             debug_log(f"Error: client {type(client).__name__} does not have generate_stream method")
             raise ValueError(f"Client {type(client).__name__} does not support streaming")
 
+        # Determine client type
         is_ollama = 'ollama' in str(type(client)).lower()
-        debug_log(f"Is Ollama client: {is_ollama}")
+        is_openai = 'openai' in str(type(client)).lower()
+        is_anthropic = 'anthropic' in str(type(client)).lower()
+        
+        debug_log(f"Client types - Ollama: {is_ollama}, OpenAI: {is_openai}, Anthropic: {is_anthropic}")
 
+        # Only show loading indicator for Ollama (which may need to load models)
+        # This prevents Ollama-specific UI elements from showing when using other providers
         if is_ollama and hasattr(app, 'query_one'):
             try:
                 debug_log("Showing initial model loading indicator for Ollama")
@@ -190,6 +239,7 @@ async def generate_streaming_response(
         debug_log(f"Starting stream generation with messages length: {len(messages)}")
         logger.info(f"Starting stream generation for model: {model}")
 
+        # Initialize stream generator
         try:
             debug_log("Calling client.generate_stream()")
             stream_generator = client.generate_stream(messages, model, style)
@@ -199,10 +249,12 @@ async def generate_streaming_response(
             logger.error(f"Error initializing stream generator: {str(stream_init_error)}")
             raise
 
-        if hasattr(client, 'is_loading_model') and not client.is_loading_model() and hasattr(app, 'query_one'):
+        # Update UI if model is ready (Ollama specific)
+        # Only check is_loading_model for Ollama clients to prevent errors with other providers
+        if is_ollama and hasattr(client, 'is_loading_model') and not client.is_loading_model() and hasattr(app, 'query_one'):
             try:
-                debug_log("Model is ready for generation, updating UI")
-                logger.info("Model is ready for generation, updating UI")
+                debug_log("Ollama model is ready for generation, updating UI")
+                logger.info("Ollama model is ready for generation, updating UI")
                 loading = app.query_one("#loading-indicator")
                 loading.remove_class("model-loading")
                 loading.update("▪▪▪ Generating response...")
@@ -210,9 +262,11 @@ async def generate_streaming_response(
                 debug_log(f"Error updating UI after stream init: {str(e)}")
                 logger.error(f"Error updating UI after stream init: {str(e)}")
 
+        # Process stream chunks
         debug_log("Beginning to process stream chunks")
         try:
             async for chunk in stream_generator:
+                # Check for task cancellation
                 if asyncio.current_task().cancelled():
                     debug_log("Task cancellation detected during chunk processing")
                     logger.info("Task cancellation detected during chunk processing")
@@ -221,30 +275,32 @@ async def generate_streaming_response(
                         await client.cancel_stream()
                     raise asyncio.CancelledError()
 
-                if hasattr(client, 'is_loading_model'):
+                # Handle Ollama model loading state changes - only for Ollama clients
+                if is_ollama and hasattr(client, 'is_loading_model'):
                     try:
                         model_loading = client.is_loading_model()
-                        debug_log(f"Model loading state: {model_loading}")
+                        debug_log(f"Ollama model loading state: {model_loading}")
                         if hasattr(app, 'query_one'):
                             try:
                                 loading = app.query_one("#loading-indicator")
                                 if model_loading and hasattr(loading, 'has_class') and not loading.has_class("model-loading"):
-                                    debug_log("Model loading started during streaming")
-                                    logger.info("Model loading started during streaming")
+                                    debug_log("Ollama model loading started during streaming")
+                                    logger.info("Ollama model loading started during streaming")
                                     loading.add_class("model-loading")
                                     loading.update("⚙️ Loading Ollama model...")
                                 elif not model_loading and hasattr(loading, 'has_class') and loading.has_class("model-loading"):
-                                    debug_log("Model loading finished during streaming")
-                                    logger.info("Model loading finished during streaming")
+                                    debug_log("Ollama model loading finished during streaming")
+                                    logger.info("Ollama model loading finished during streaming")
                                     loading.remove_class("model-loading")
                                     loading.update("▪▪▪ Generating response...")
                             except Exception as ui_e:
                                 debug_log(f"Error updating UI elements: {str(ui_e)}")
                                 logger.error(f"Error updating UI elements: {str(ui_e)}")
                     except Exception as e:
-                        debug_log(f"Error checking model loading state: {str(e)}")
-                        logger.error(f"Error checking model loading state: {str(e)}")
+                        debug_log(f"Error checking Ollama model loading state: {str(e)}")
+                        logger.error(f"Error checking Ollama model loading state: {str(e)}")
 
+                # Process chunk content
                 if chunk:
                     if not isinstance(chunk, str):
                         debug_log(f"WARNING: Received non-string chunk of type: {type(chunk).__name__}")
@@ -259,7 +315,8 @@ async def generate_streaming_response(
                     buffer.append(chunk)
                     current_time = time.time()
 
-                    # Always update immediately for the first few chunks
+                    # Update UI with new content
+                    # Always update immediately for the first few chunks for better responsiveness
                     if (current_time - last_update >= update_interval or
                         len(''.join(buffer)) > 5 or  # Reduced buffer size threshold
                         len(full_response) < 50):    # More aggressive updates for early content
@@ -268,34 +325,25 @@ async def generate_streaming_response(
                         full_response += new_content
                         debug_log(f"Updating UI with content length: {len(full_response)}")
                         
-                        # Print to console for debugging
-                        print(f"Streaming update: +{len(new_content)} chars, total: {len(full_response)}")
+                        # Only print to console for debugging if not OpenAI
+                        # This prevents Ollama debug output from appearing in OpenAI responses
+                        if not is_openai:
+                            print(f"Streaming update: +{len(new_content)} chars, total: {len(full_response)}")
                         
                         try:
                             # Call the UI callback with the full response so far
                             await callback(full_response)
                             debug_log("UI callback completed successfully")
-                            print("UI callback completed successfully")
                             
-                            # Force app refresh after each update - CRITICAL for visibility
+                            # Force app refresh after each update
                             if hasattr(app, 'refresh'):
-                                # First do a layout=False refresh which is faster
-                                app.refresh(layout=False)
-                                # Then do a full layout refresh to ensure content is visible
-                                await asyncio.sleep(0.01)
-                                app.refresh(layout=True)
-                                
-                                # Try to force scroll to end
-                                try:
-                                    messages_container = app.query_one("#messages-container")
-                                    if messages_container and hasattr(messages_container, 'scroll_end'):
-                                        messages_container.scroll_end(animate=False)
-                                except Exception as scroll_err:
-                                    debug_log(f"Error scrolling: {str(scroll_err)}")
+                                app.refresh(layout=True)  # Force layout refresh
                         except Exception as callback_err:
                             debug_log(f"Error in UI callback: {str(callback_err)}")
                             logger.error(f"Error in UI callback: {str(callback_err)}")
-                            print(f"Error updating UI: {str(callback_err)}")
+                            # Only print error to console if not OpenAI
+                            if not is_openai:
+                                print(f"Error updating UI: {str(callback_err)}")
                             
                         buffer = []
                         last_update = current_time
@@ -451,8 +499,8 @@ def resolve_model_id(model_id_or_name: str) -> str:
     """
     Resolves a potentially short model ID or display name to the full model ID
     stored in the configuration. Tries multiple matching strategies.
-
-    Fix: Only apply dot-to-colon conversion for Ollama models, not for OpenAI/Anthropic/custom.
+    
+    This function is critical for ensuring models are correctly identified by provider.
     """
     if not model_id_or_name:
         logger.warning("resolve_model_id called with empty input, returning empty string.")
@@ -460,6 +508,16 @@ def resolve_model_id(model_id_or_name: str) -> str:
 
     input_lower = model_id_or_name.lower().strip()
     logger.info(f"Attempting to resolve model identifier: '{input_lower}'")
+    
+    # First, check if this is an OpenAI model - if so, return as-is to ensure correct provider
+    if any(name in input_lower for name in ["gpt", "text-", "davinci"]):
+        logger.info(f"Input '{input_lower}' appears to be an OpenAI model, returning as-is")
+        return model_id_or_name
+        
+    # Next, check if this is an Anthropic model - if so, return as-is to ensure correct provider
+    if any(name in input_lower for name in ["claude", "anthropic"]):
+        logger.info(f"Input '{input_lower}' appears to be an Anthropic model, returning as-is")
+        return model_id_or_name
 
     available_models = CONFIG.get("available_models", {})
     if not available_models:
@@ -470,20 +528,22 @@ def resolve_model_id(model_id_or_name: str) -> str:
     provider = None
     if input_lower in available_models:
         provider = available_models[input_lower].get("provider")
+        logger.info(f"Found model in available_models with provider: {provider}")
     else:
         # Try to find by display name
         for model_info in available_models.values():
             if model_info.get("display_name", "").lower() == input_lower:
                 provider = model_info.get("provider")
+                logger.info(f"Found model by display name with provider: {provider}")
                 break
 
     # Special case for Ollama models with version format (model:version)
-    if provider == "ollama" and ":" in input_lower and not input_lower.startswith("claude-"):
+    if (provider == "ollama" or any(name in input_lower for name in ["llama", "mistral", "codellama", "gemma"])) and ":" in input_lower and not input_lower.startswith("claude-"):
         logger.info(f"Input '{input_lower}' appears to be an Ollama model with version, returning as-is")
         return model_id_or_name
 
     # Only apply dot-to-colon for Ollama models
-    if provider == "ollama" and "." in input_lower and not input_lower.startswith("claude-"):
+    if (provider == "ollama" or any(name in input_lower for name in ["llama", "mistral", "codellama", "gemma"])) and "." in input_lower and not input_lower.startswith("claude-"):
         logger.info(f"Input '{input_lower}' appears to be an Ollama model with dot notation")
         if ":" not in input_lower:
             parts = input_lower.split(".")

{chat_console-0.3.5.dist-info → chat_console-0.3.7.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: chat-console
-Version: 0.3.5
+Version: 0.3.7
 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-0.3.5.dist-info → chat_console-0.3.7.dist-info}/RECORD

@@ -1,24 +1,24 @@
-app/__init__.py,sha256=zlFCSAkBSbWCOmcYBIlrvLwRHY4lir8TILoawj6_qK0,130
+app/__init__.py,sha256=ZSZR6xIuPhvv1zB4p63eSeGQX8bTkhxBWk2Gn0peFaw,130
 app/config.py,sha256=KawltE7cK2bR9wbe1NSlepwWIjkiFw2bg3vbLmUnP38,7626
 app/database.py,sha256=nt8CVuDpy6zw8mOYqDcfUmNw611t7Ln7pz22M0b6-MI,9967
-app/main.py,sha256=Fp7NWAuAdZ1TEWsBYUZlwZbTLD7kBHmHvAztjzQwxpw,73836
+app/main.py,sha256=clcRjXwySxVjrPtqvPOIfl7r8KbHVLZ1woxyEnvl3JI,72829
 app/models.py,sha256=4-y9Lytay2exWPFi0FDlVeRL3K2-I7E-jBqNzTfokqY,2644
-app/utils.py,sha256=rTxUKndels-cB8P8qCge9AFEpriI9PC-HJ6rE88bS0k,29548
+app/utils.py,sha256=htktBl1JucYEHo1WBrWkfdip4yzRtvyVl24Aaj445xA,32421
 app/api/__init__.py,sha256=A8UL84ldYlv8l7O-yKzraVFcfww86SgWfpl4p7R03-w,62
 app/api/anthropic.py,sha256=UpIP3CgAOUimdVyif41MhBOCAgOyFO8mX9SFQMKRAmc,12483
-app/api/base.py,sha256=bqBT4jne_W6Cvj_GoWWclV4Uk95fQvt-kkYqqZFJd8M,5769
+app/api/base.py,sha256=eShCiZIcW3yeZLONt1xnkP0vU6v5MEaDj3YZ3xcPle8,7294
 app/api/ollama.py,sha256=EBEEKXbgAYWEg_zF5PO_UKO5l_aoU3J_7tfCj9e-fqs,61699
 app/api/openai.py,sha256=6ORruzuuZtIjME3WK-g7kXf7cBmM4td5Njv9JLaWh7E,9557
 app/ui/__init__.py,sha256=RndfbQ1Tv47qdSiuQzvWP96lPS547SDaGE-BgOtiP_w,55
-app/ui/chat_interface.py,sha256=0hvklvcVe2jjIhXfkaG5SrtO1iY9rAPt9xqpQ3G4B14,17873
+app/ui/chat_interface.py,sha256=TJlMzVmrKzr3t0JIhto0vKBvyik7gJ7UEyW3Vqbn3cE,17262
 app/ui/chat_list.py,sha256=WQTYVNSSXlx_gQal3YqILZZKL9UiTjmNMIDX2I9pAMM,11205
 app/ui/model_browser.py,sha256=pdblLVkdyVF0_Bo02bqbErGAtieyH-y6IfhMOPEqIso,71124
 app/ui/model_selector.py,sha256=ue3rbZfjVsjli-rJN5mfSqq23Ci7NshmTb4xWS-uG5k,18685
 app/ui/search.py,sha256=b-m14kG3ovqW1-i0qDQ8KnAqFJbi5b1FLM9dOnbTyIs,9763
 app/ui/styles.py,sha256=04AhPuLrOd2yenfRySFRestPeuTPeMLzhmMB67NdGvw,5615
-chat_console-0.3.5.dist-info/licenses/LICENSE,sha256=srHZ3fvcAuZY1LHxE7P6XWju2njRCHyK6h_ftEbzxSE,1057
-chat_console-0.3.5.dist-info/METADATA,sha256=vWdFjFiYAA3swyvkbg-EynSmhj_yhSlyhwTdaiXTke4,2921
-chat_console-0.3.5.dist-info/WHEEL,sha256=pxyMxgL8-pra_rKaQ4drOZAegBVuX-G_4nRHjjgWbmo,91
-chat_console-0.3.5.dist-info/entry_points.txt,sha256=kkVdEc22U9PAi2AeruoKklfkng_a_aHAP6VRVwrAD7c,67
-chat_console-0.3.5.dist-info/top_level.txt,sha256=io9g7LCbfmTG1SFKgEOGXmCFB9uMP2H5lerm0HiHWQE,4
-chat_console-0.3.5.dist-info/RECORD,,
+chat_console-0.3.7.dist-info/licenses/LICENSE,sha256=srHZ3fvcAuZY1LHxE7P6XWju2njRCHyK6h_ftEbzxSE,1057
+chat_console-0.3.7.dist-info/METADATA,sha256=eDQRUghh8Ihp8z38oAlI0___RBBDJHpLmhBGF0VgZ1w,2921
+chat_console-0.3.7.dist-info/WHEEL,sha256=SmOxYU7pzNKBqASvQJ7DjX3XGUF92lrGhMb3R6_iiqI,91
+chat_console-0.3.7.dist-info/entry_points.txt,sha256=kkVdEc22U9PAi2AeruoKklfkng_a_aHAP6VRVwrAD7c,67
+chat_console-0.3.7.dist-info/top_level.txt,sha256=io9g7LCbfmTG1SFKgEOGXmCFB9uMP2H5lerm0HiHWQE,4
+chat_console-0.3.7.dist-info/RECORD,,

{chat_console-0.3.5.dist-info → chat_console-0.3.7.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (79.0.0)
+Generator: setuptools (79.0.1)
 Root-Is-Purelib: true
 Tag: py3-none-any