Bug fixes. Added missing functionality. ++

.gitignore

@@ -46,3 +46,5 @@ requirements.txt
 system_prompt.txt
 CLAUDE*
 SESSION*_COMPLETE.md
+oai/FILE_ATTACHMENTS_FIX.md
+oai/OLLAMA_ERROR_HANDLING.md

README.md

@@ -2,6 +2,11 @@
 
 A powerful, modern **Textual TUI** chat client with **multi-provider support** (OpenRouter, Anthropic, OpenAI, Ollama) and **MCP (Model Context Protocol)** integration, enabling AI to access local files and query SQLite databases.
 
+>[!WARNING]  
+>  v3.0.0-b4 is beta software. While I strive for stability, beta versions may contain bugs, incomplete features, or unexpected behavior. I actively work on improvements and appreciate your feedback.
+>
+>Beta releases are ideal for testing new features and providing feedback. For production use or maximum stability, consider using the latest stable release. 
+
 ## Features
 
 ### Core Features
@@ -141,6 +146,33 @@ On first run, you'll be prompted for your API key. Configure additional provider
 Ctrl+Q              # Quit
 ```
 
+### File Attachments
+
+Attach files to your messages using `@<file>` or `@file` syntax:
+
+```bash
+# Single file
+Describe this image @<photo.jpg>
+Analyze @~/Documents/report.pdf
+
+# Multiple files
+Compare @image1.png and @image2.png
+
+# With paths
+Review this code @./src/main.py
+What's in @/Users/username/Downloads/screenshot.png
+```
+
+**Supported file types:**
+- **Images**: `.png`, `.jpg`, `.jpeg`, `.gif`, `.webp`, `.bmp` (for vision models)
+- **PDFs**: `.pdf` (for document-capable models)
+- **Code/Text**: `.py`, `.js`, `.md`, `.txt`, `.json`, etc. (all models)
+
+**Notes:**
+- Files are automatically base64-encoded for images/PDFs
+- Maximum file size: 10MB per file
+- Works with vision-capable models (Claude, GPT-4o, Gemini, etc.)
+
 ## Web Search
 
 oAI provides universal web search capabilities for all AI providers with three options:
@@ -341,7 +373,6 @@ oai/
 │   │   └── styles.tcss    # TUI styling
 │   └── utils/             # Logging, export, etc.
 ├── pyproject.toml         # Package configuration
-├── build.sh               # Binary build script
 └── README.md
 ```
 

oai/__init__.py

@@ -9,7 +9,7 @@ Author: Rune
 License: MIT
 """
 
-__version__ = "3.0.0-b3"
+__version__ = "3.0.0-b4"
 __author__ = "Rune Olsen"
 __license__ = "MIT"
 

oai/cli.py

@@ -125,6 +125,23 @@ def _launch_tui(
             provider_api_keys=provider_api_keys,
             ollama_base_url=settings.ollama_base_url,
         )
+    except ConnectionError as e:
+        # Special handling for Ollama connection errors
+        if selected_provider == "ollama":
+            typer.echo(f"Error: Cannot connect to Ollama server", err=True)
+            typer.echo(f"", err=True)
+            typer.echo(f"Details: {e}", err=True)
+            typer.echo(f"", err=True)
+            typer.echo(f"Please make sure Ollama is running:", err=True)
+            typer.echo(f"  ollama serve", err=True)
+            typer.echo(f"", err=True)
+            typer.echo(f"Or configure a different Ollama URL:", err=True)
+            typer.echo(f"  oai config ollama_base_url <url>", err=True)
+            typer.echo(f"", err=True)
+            typer.echo(f"Current Ollama URL: {settings.ollama_base_url}", err=True)
+        else:
+            typer.echo(f"Error: Connection failed: {e}", err=True)
+        raise typer.Exit(1)
     except Exception as e:
         typer.echo(f"Error: Failed to initialize client: {e}", err=True)
         raise typer.Exit(1)

oai/commands/handlers.py

@@ -139,7 +139,6 @@ class HelpCommand(Command):
                 ("/retry", "Resend the last prompt.", "/retry"),
                 ("/memory", "Toggle conversation memory.", "/memory on"),
                 ("/online", "Toggle online mode (web search).", "/online on"),
-                ("/paste", "Paste from clipboard with optional prompt.", "/paste Explain"),
             ]),
             ("[bold cyan]━━━ NAVIGATION ━━━[/]", [
                 ("/prev", "View previous response in history.", "/prev"),
@@ -150,6 +149,7 @@ class HelpCommand(Command):
                 ("/model", "Select AI model.", "/model gpt"),
                 ("/info", "Show model information.", "/info"),
                 ("/config", "View or change settings.", "/config stream on"),
+                ("/config log", "Set log level.", "/config log debug"),
                 ("/maxtoken", "Set session token limit.", "/maxtoken 2000"),
                 ("/system", "Set system prompt.", "/system You are an expert"),
             ]),
@@ -1022,13 +1022,20 @@ class ConfigCommand(Command):
             else:
                 pass
 
-        elif setting == "loglevel":
+        elif setting in ["loglevel", "log"]:
             valid_levels = ["debug", "info", "warning", "error", "critical"]
-            if value and value.lower() in valid_levels:
-                settings.set_log_level(value.lower())
-                print_success(f"Log level set to: {value.lower()}")
+            if value:
+                if value.lower() in valid_levels:
+                    settings.set_log_level(value.lower())
+                    message = f"✓ Log level set to: {value.lower()}"
+                    return CommandResult.success(message=message)
+                else:
+                    message = f"Invalid log level. Valid levels: {', '.join(valid_levels)}"
+                    return CommandResult.error(message)
             else:
-                print_info(f"Valid levels: {', '.join(valid_levels)}")
+                # Show current log level
+                message = f"Current log level: {settings.log_level}\nValid levels: {', '.join(valid_levels)}"
+                return CommandResult.success(message=message)
 
         else:
             pass
@@ -1582,10 +1589,30 @@ class ProviderCommand(Command):
                 else:
                     # No models available, clear selection
                     context.session.selected_model = None
-                    message = f"Switched to {provider_name} provider (no models available)"
-                    return CommandResult.success(message=message)
+                    if provider_name == "ollama":
+                        # Special message for Ollama with helpful instructions
+                        message = f"⚠️  Switched to Ollama, but no models are installed.\n\n"
+                        message += "To install models, run:\n"
+                        message += "  ollama pull llama3.2\n"
+                        message += "  ollama pull mistral\n\n"
+                        message += "See available models at: https://ollama.com/library"
+                        return CommandResult.warning(message=message)
+                    else:
+                        message = f"Switched to {provider_name} provider (no models available)"
+                        return CommandResult.warning(message=message)
 
             return CommandResult.success(message=f"Switched to {provider_name} provider")
+        except ConnectionError as e:
+            # Connection error - likely Ollama server not running
+            if provider_name == "ollama":
+                error_msg = f"❌ Cannot connect to Ollama server.\n\n"
+                error_msg += f"Error: {str(e)}\n\n"
+                error_msg += "Please make sure Ollama is running:\n"
+                error_msg += "  ollama serve\n\n"
+                error_msg += f"Or check your Ollama URL in config (currently: {context.settings.ollama_base_url})"
+                return CommandResult.error(error_msg)
+            else:
+                return CommandResult.error(f"Connection error: {e}")
         except Exception as e:
             return CommandResult.error(f"Failed to switch provider: {e}")
 
@@ -1600,32 +1627,15 @@ class PasteCommand(Command):
     @property
     def help(self) -> CommandHelp:
         return CommandHelp(
-            description="Paste from clipboard and send to AI.",
-            usage="/paste [prompt]",
+            description="[Disabled] Use Cmd+V (macOS) or Ctrl+V (Linux/Windows) to paste directly.",
+            usage="/paste",
+            notes="This command is disabled. Use keyboard shortcuts to paste instead.",
         )
 
     def execute(self, args: str, context: CommandContext) -> CommandResult:
-        try:
-            import pyperclip
-            content = pyperclip.paste()
-        except ImportError:
-            message = "pyperclip not installed"
-            return CommandResult.error(message)
-        except Exception as e:
-            message = f"Failed to access clipboard: {e}"
-            return CommandResult.error(str(e))
-
-        if not content:
-            message = "Clipboard is empty"
-            return CommandResult.error(message)
-
-        # Build the prompt
-        if args:
-            full_prompt = f"{args}\n\n```\n{content}\n```"
-        else:
-            full_prompt = content
-
-        return CommandResult.success(data={"paste_prompt": full_prompt})
+        # Disabled - use Cmd+V (macOS) or Ctrl+V (Linux/Windows) instead
+        message = "💡 Tip: Use Cmd+V (macOS) or Ctrl+V (Linux/Windows) to paste directly"
+        return CommandResult.success(message=message)
 
 class ModelCommand(Command):
     """Select AI model."""

oai/commands/registry.py

@@ -23,6 +23,7 @@ class CommandStatus(str, Enum):
     """Status of command execution."""
 
     SUCCESS = "success"
+    WARNING = "warning"
     ERROR = "error"
     CONTINUE = "continue"  # Continue to next handler
     EXIT = "exit"  # Exit the application
@@ -50,6 +51,11 @@ class CommandResult:
         """Create a success result."""
         return cls(status=CommandStatus.SUCCESS, message=message, data=data)
 
+    @classmethod
+    def warning(cls, message: str, data: Any = None) -> "CommandResult":
+        """Create a warning result."""
+        return cls(status=CommandStatus.WARNING, message=message, data=data)
+
     @classmethod
     def error(cls, message: str) -> "CommandResult":
         """Create an error result."""

oai/core/client.py

@@ -221,18 +221,24 @@ class AIClient:
             f"messages={len(chat_messages)}, stream={stream}"
         )
 
-        return self.provider.chat(
-            model=model_id,
-            messages=chat_messages,
-            stream=stream,
-            max_tokens=max_tokens,
-            temperature=temperature,
-            tools=tools,
-            tool_choice=tool_choice,
-            transforms=transforms,
-            enable_web_search=enable_web_search,
-            web_search_config=web_search_config or {},
-        )
+        # Build provider chat parameters
+        chat_params = {
+            "model": model_id,
+            "messages": chat_messages,
+            "stream": stream,
+            "max_tokens": max_tokens,
+            "temperature": temperature,
+            "tools": tools,
+            "tool_choice": tool_choice,
+            "transforms": transforms,
+        }
+
+        # Only pass web search params to Anthropic provider
+        if self.provider_name == "anthropic":
+            chat_params["enable_web_search"] = enable_web_search
+            chat_params["web_search_config"] = web_search_config or {}
+
+        return self.provider.chat(**chat_params)
 
     def chat_with_tools(
         self,

oai/core/session.py

@@ -24,6 +24,7 @@ from oai.mcp.manager import MCPManager
 from oai.providers.base import ChatResponse, StreamChunk, UsageStats
 from oai.utils.logging import get_logger
 from oai.utils.web_search import perform_web_search, format_search_results
+from oai.utils.files import parse_file_attachments, prepare_file_attachment
 
 logger = get_logger()
 
@@ -218,8 +219,40 @@ class ChatSession:
                 messages.append({"role": "user", "content": entry.prompt})
                 messages.append({"role": "assistant", "content": entry.response})
 
-        # Add current message
-        messages.append({"role": "user", "content": user_input})
+        # Parse file attachments from user input
+        cleaned_text, file_paths = parse_file_attachments(user_input)
+
+        # Build content for current message
+        if file_paths:
+            # Multi-modal message with attachments
+            content_parts = []
+
+            # Add text part if there's any text
+            if cleaned_text.strip():
+                content_parts.append({
+                    "type": "text",
+                    "text": cleaned_text.strip()
+                })
+
+            # Add file attachments
+            for file_path in file_paths:
+                attachment = prepare_file_attachment(
+                    file_path,
+                    self.selected_model or {}
+                )
+                if attachment:
+                    content_parts.append(attachment)
+                else:
+                    logger.warning(f"Could not attach file: {file_path}")
+
+            # If we have content parts, use them; otherwise fall back to text
+            if content_parts:
+                messages.append({"role": "user", "content": content_parts})
+            else:
+                messages.append({"role": "user", "content": user_input})
+        else:
+            # Simple text message
+            messages.append({"role": "user", "content": user_input})
 
         return messages
 
@@ -517,15 +550,21 @@ class ChatSession:
         Returns:
             Tuple of (full_text, usage)
         """
-        response = self.client.chat(
-            messages=messages,
-            model=model_id,
-            stream=True,
-            max_tokens=max_tokens,
-            transforms=transforms,
-            enable_web_search=enable_web_search,
-            web_search_config=web_search_config or {},
-        )
+        # Build chat parameters
+        chat_params = {
+            "messages": messages,
+            "model": model_id,
+            "stream": True,
+            "max_tokens": max_tokens,
+            "transforms": transforms,
+        }
+
+        # Only pass web search params to Anthropic provider
+        if self.client.provider_name == "anthropic":
+            chat_params["enable_web_search"] = enable_web_search
+            chat_params["web_search_config"] = web_search_config or {}
+
+        response = self.client.chat(**chat_params)
 
         if isinstance(response, ChatResponse):
             return response.content or "", response.usage
@@ -647,8 +686,9 @@ class ChatSession:
             ):
                 yield chunk
         else:
-            # Non-streaming request
-            response = self.client.chat(
+            # Non-streaming request - run in thread to avoid blocking event loop
+            response = await asyncio.to_thread(
+                self.client.chat,
                 messages=messages,
                 model=model_id,
                 stream=False,
@@ -691,7 +731,9 @@ class ChatSession:
         api_messages = list(messages)
 
         while loop_count < max_loops:
-            response = self.client.chat(
+            # Run in thread to avoid blocking event loop
+            response = await asyncio.to_thread(
+                self.client.chat,
                 messages=api_messages,
                 model=model_id,
                 stream=False,
@@ -837,33 +879,55 @@ class ChatSession:
         Yields:
             StreamChunk objects
         """
-        response = self.client.chat(
-            messages=messages,
-            model=model_id,
-            stream=True,
-            max_tokens=max_tokens,
-            transforms=transforms,
-            enable_web_search=enable_web_search,
-            web_search_config=web_search_config or {},
-        )
+        # Build chat parameters
+        chat_params = {
+            "messages": messages,
+            "model": model_id,
+            "stream": True,
+            "max_tokens": max_tokens,
+            "transforms": transforms,
+        }
 
-        if isinstance(response, ChatResponse):
-            # Non-streaming response
-            chunk = StreamChunk(
-                id="",
-                delta_content=response.content,
-                usage=response.usage,
-                error=None,
-            )
-            yield chunk
-            return
+        # Only pass web search params to Anthropic provider
+        if self.client.provider_name == "anthropic":
+            chat_params["enable_web_search"] = enable_web_search
+            chat_params["web_search_config"] = web_search_config or {}
 
-        # Stream the response
-        for chunk in response:
-            if chunk.error:
-                yield StreamChunk(id="", delta_content=None, usage=None, error=chunk.error)
-                break
-            yield chunk
+        # For streaming, call directly (generator yields control naturally)
+        # For non-streaming, we'll detect it and run in thread
+        if chat_params.get("stream", True):
+            # Streaming - call directly, iteration will yield control
+            response = self.client.chat(**chat_params)
+
+            if isinstance(response, ChatResponse):
+                # Provider returned non-streaming despite stream=True
+                chunk = StreamChunk(
+                    id="",
+                    delta_content=response.content,
+                    usage=response.usage,
+                    error=None,
+                )
+                yield chunk
+                return
+
+            # Stream the response - yield control between chunks
+            for chunk in response:
+                await asyncio.sleep(0)  # Yield control to event loop
+                if chunk.error:
+                    yield StreamChunk(id="", delta_content=None, usage=None, error=chunk.error)
+                    break
+                yield chunk
+        else:
+            # Non-streaming - run in thread to avoid blocking
+            response = await asyncio.to_thread(self.client.chat, **chat_params)
+            if isinstance(response, ChatResponse):
+                chunk = StreamChunk(
+                    id="",
+                    delta_content=response.content,
+                    usage=response.usage,
+                    error=None,
+                )
+                yield chunk
 
     # ========== END ASYNC METHODS ==========
 

oai/providers/anthropic.py

@@ -647,6 +647,9 @@ class AnthropicProvider(AIProvider):
                 "description": model.description,
                 "context_length": model.context_length,
                 "pricing": model.pricing,
+                "supported_parameters": model.supported_parameters,
+                "input_modalities": model.input_modalities,
+                "output_modalities": model.output_modalities,
             }
             for model in models
         ]
@@ -669,5 +672,8 @@ class AnthropicProvider(AIProvider):
                 "description": model.description,
                 "context_length": model.context_length,
                 "pricing": model.pricing,
+                "supported_parameters": model.supported_parameters,
+                "input_modalities": model.input_modalities,
+                "output_modalities": model.output_modalities,
             }
         return None

oai/providers/ollama.py

@@ -59,6 +59,9 @@ class OllamaProvider(AIProvider):
 
         Returns:
             True if server is accessible
+
+        Raises:
+            ConnectionError: If server is not accessible
         """
         try:
             response = requests.get(f"{self.base_url}/api/tags", timeout=2)
@@ -66,11 +69,59 @@ class OllamaProvider(AIProvider):
                 logger.info(f"Ollama server accessible at {self.base_url}")
                 return True
             else:
-                logger.warning(f"Ollama server returned status {response.status_code}")
-                return False
+                error_msg = f"Ollama server returned status {response.status_code} at {self.base_url}"
+                logger.error(error_msg)
+                raise ConnectionError(error_msg)
         except requests.RequestException as e:
-            logger.warning(f"Ollama server not accessible at {self.base_url}: {e}")
-            return False
+            error_msg = f"Cannot connect to Ollama server at {self.base_url}. Is Ollama running? Error: {e}"
+            logger.error(error_msg)
+            raise ConnectionError(error_msg) from e
+
+    @property
+    def name(self) -> str:
+        """Get provider name."""
+        return "Ollama"
+
+    @property
+    def capabilities(self) -> ProviderCapabilities:
+        """Get provider capabilities."""
+        return ProviderCapabilities(
+            streaming=True,
+            tools=False,  # Tool support varies by model
+            images=False,  # Image support varies by model
+            online=True,  # Web search via DuckDuckGo/Google
+            max_context=8192,  # Varies by model
+        )
+
+    def list_models(self, filter_text_only: bool = True) -> List[ModelInfo]:
+        """
+        List models from local Ollama installation.
+
+        Args:
+            filter_text_only: Ignored for Ollama
+
+        Returns:
+            List of available models
+        """
+        try:
+            response = requests.get(f"{self.base_url}/api/tags", timeout=5)
+            response.raise_for_status()
+            data = response.json()
+
+            models = []
+            for model_data in data.get("models", []):
+                models.append(self._parse_model(model_data))
+
+            if models:
+                logger.info(f"Found {len(models)} Ollama models")
+            else:
+                logger.warning(f"Ollama server at {self.base_url} has no models installed. Install models with: ollama pull <model_name>")
+            return models
+
+        except requests.RequestException as e:
+            logger.error(f"Failed to connect to Ollama server at {self.base_url}: {e}")
+            logger.error(f"Make sure Ollama is running. Start it with: ollama serve")
+            return []
 
     @property
     def name(self) -> str:
@@ -397,6 +448,9 @@ class OllamaProvider(AIProvider):
                 "description": model.description,
                 "context_length": model.context_length,
                 "pricing": model.pricing,
+                "supported_parameters": model.supported_parameters,
+                "input_modalities": model.input_modalities,
+                "output_modalities": model.output_modalities,
             }
             for model in models
         ]
@@ -419,5 +473,8 @@ class OllamaProvider(AIProvider):
                 "description": model.description,
                 "context_length": model.context_length,
                 "pricing": model.pricing,
+                "supported_parameters": model.supported_parameters,
+                "input_modalities": model.input_modalities,
+                "output_modalities": model.output_modalities,
             }
         return None

oai/providers/openai.py

@@ -604,6 +604,9 @@ class OpenAIProvider(AIProvider):
                 "description": model.description,
                 "context_length": model.context_length,
                 "pricing": model.pricing,
+                "supported_parameters": model.supported_parameters,
+                "input_modalities": model.input_modalities,
+                "output_modalities": model.output_modalities,
             }
             for model in models
         ]
@@ -626,5 +629,8 @@ class OpenAIProvider(AIProvider):
                 "description": model.description,
                 "context_length": model.context_length,
                 "pricing": model.pricing,
+                "supported_parameters": model.supported_parameters,
+                "input_modalities": model.input_modalities,
+                "output_modalities": model.output_modalities,
             }
         return None

oai/tui/app.py

@@ -7,7 +7,7 @@ from typing import Optional
 
 import pyperclip
 from textual.app import App, ComposeResult
-from textual.widgets import Input
+from textual.widgets import TextArea
 
 from oai import __version__
 from oai.commands.registry import CommandStatus, registry
@@ -36,6 +36,7 @@ from oai.tui.widgets import (
     SystemMessageWidget,
     UserMessageWidget,
 )
+from oai.tui.widgets.input_bar import ChatTextArea
 from oai.tui.widgets.command_dropdown import CommandDropdown
 
 
@@ -66,6 +67,8 @@ class oAIChatApp(App):
         self.input_history: list[str] = []
         self.history_index: int = -1
         self._navigating_history: bool = False
+        self._cancel_generation: bool = False
+        self._is_generating: bool = False
 
     def compose(self) -> ComposeResult:
         """Compose the TUI layout."""
@@ -80,6 +83,9 @@ class oAIChatApp(App):
 
     def on_mount(self) -> None:
         """Handle app mount."""
+        # Load input history from file
+        self._load_input_history()
+
         # Focus the input
         input_bar = self.query_one(InputBar)
         chat_input = input_bar.get_input()
@@ -97,11 +103,47 @@ class oAIChatApp(App):
         if self.session.online_enabled:
             input_bar.update_online_mode(True)
 
+    def _load_input_history(self) -> None:
+        """Load input history from history.txt file."""
+        from oai.constants import HISTORY_FILE
+
+        try:
+            if HISTORY_FILE.exists():
+                with open(HISTORY_FILE, "r", encoding="utf-8") as f:
+                    # Load all non-empty lines and unescape newlines
+                    self.input_history = [
+                        line.strip().replace("\\n", "\n")
+                        for line in f
+                        if line.strip()
+                    ]
+        except Exception as e:
+            self.logger.error(f"Failed to load input history: {e}")
+
+    def _save_input_to_history(self, user_input: str) -> None:
+        """Append input to history.txt file."""
+        from oai.constants import HISTORY_FILE
+
+        try:
+            # Escape newlines so multiline inputs stay as one history entry
+            escaped_input = user_input.replace("\n", "\\n")
+            with open(HISTORY_FILE, "a", encoding="utf-8") as f:
+                f.write(f"{escaped_input}\n")
+        except Exception as e:
+            self.logger.error(f"Failed to save input to history: {e}")
+
     def on_key(self, event) -> None:
         """Handle global keyboard shortcuts."""
         # Debug: Show what key was pressed
         # self.notify(f"Key pressed: {event.key}", severity="information")
 
+        # Handle Escape to cancel generation
+        if event.key == "escape" and self._is_generating:
+            self._cancel_generation = True
+            self.notify("⏹️  Stopping generation...", severity="warning")
+            event.prevent_default()
+            event.stop()
+            return
+
         # Don't handle keys if a modal screen is open (let the modal handle them)
         if len(self.screen_stack) > 1:
             return
@@ -119,25 +161,29 @@ class oAIChatApp(App):
             if dropdown_visible:
                 if event.key == "up":
                     event.prevent_default()
+                    event.stop()
                     dropdown.move_selection_up()
                     return
                 elif event.key == "down":
                     event.prevent_default()
+                    event.stop()
                     dropdown.move_selection_down()
                     return
                 elif event.key == "tab":
                     # Tab accepts the selected command and adds space for arguments
                     event.prevent_default()
+                    event.stop()
                     selected = dropdown.get_selected_command()
                     if selected:
-                        chat_input.value = selected + " "
-                        chat_input.cursor_position = len(chat_input.value)
+                        chat_input.text = selected + " "
+                        chat_input.move_cursor_relative(rows=0, columns=len(selected) + 1)
                         dropdown.hide()
                     return
                 elif event.key == "enter":
                     # Enter accepts the selected command
                     # If command needs more input, add space; otherwise submit
                     event.prevent_default()
+                    event.stop()  # Stop propagation to prevent TextArea from processing
                     selected = dropdown.get_selected_command()
                     if selected:
                         # Commands that require additional arguments
@@ -155,13 +201,13 @@ class oAIChatApp(App):
 
                         if needs_input:
                             # Add space and wait for user to type more
-                            chat_input.value = selected + " "
-                            chat_input.cursor_position = len(chat_input.value)
+                            chat_input.text = selected + " "
+                            chat_input.move_cursor_relative(rows=0, columns=len(selected) + 1)
                             dropdown.hide()
                         else:
                             # Command is complete, submit it directly
                             dropdown.hide()
-                            chat_input.value = ""  # Clear immediately
+                            chat_input.clear()  # Clear immediately
                             # Process the command directly
                             async def submit_command():
                                 await self._process_submitted_input(selected)
@@ -170,8 +216,14 @@ class oAIChatApp(App):
                 elif event.key == "escape":
                     # Escape closes dropdown
                     event.prevent_default()
+                    event.stop()
                     dropdown.hide()
                     return
+            # Escape to clear input (when dropdown not visible)
+            elif event.key == "escape":
+                event.prevent_default()
+                chat_input.clear()
+                return
             # Otherwise, arrow keys navigate history
             elif event.key == "up":
                 event.prevent_default()
@@ -211,9 +263,9 @@ class oAIChatApp(App):
             event.prevent_default()
             self.action_copy_last_response()
 
-    def on_input_changed(self, event: Input.Changed) -> None:
-        """Handle input value changes to show/hide command dropdown."""
-        if event.input.id != "chat-input":
+    def on_text_area_changed(self, event: TextArea.Changed) -> None:
+        """Handle text area value changes to show/hide command dropdown."""
+        if event.text_area.id != "chat-input":
             return
 
         # Don't show dropdown when navigating history
@@ -221,7 +273,7 @@ class oAIChatApp(App):
             return
 
         dropdown = self.query_one(CommandDropdown)
-        value = event.value
+        value = event.text_area.text
 
         # Show dropdown if input starts with /
         if value.startswith("/") and not value.startswith("//"):
@@ -229,18 +281,22 @@ class oAIChatApp(App):
         else:
             dropdown.hide()
 
-    async def on_input_submitted(self, event: Input.Submitted) -> None:
-        """Handle input submission."""
+    async def on_chat_text_area_submit(self, event: ChatTextArea.Submit) -> None:
+        """Handle Enter key submission from ChatTextArea."""
         user_input = event.value.strip()
         if not user_input:
             return
 
-        # Clear input field immediately
-        event.input.value = ""
+        # Clear input field and refocus for next message
+        input_bar = self.query_one(InputBar)
+        chat_input = input_bar.get_input()
+        chat_input.clear()
 
-        # Process the input (async, will wait for AI response)
         await self._process_submitted_input(user_input)
 
+        # Keep input focused so user can type while AI responds
+        chat_input.focus()
+
     async def _process_submitted_input(self, user_input: str) -> None:
         """Process submitted input (command or message).
 
@@ -254,20 +310,30 @@ class oAIChatApp(App):
         dropdown = self.query_one(CommandDropdown)
         dropdown.hide()
 
-        # Add to history
+        # Add to in-memory history and save to file
         self.input_history.append(user_input)
         self.history_index = -1
+        self._save_input_to_history(user_input)
 
-        # Always show what the user typed
+        # Show user message
         chat_display = self.query_one(ChatDisplay)
         user_widget = UserMessageWidget(user_input)
         await chat_display.add_message(user_widget)
 
-        # Check if it's a command
+        # Check if it's a command or message
         if user_input.startswith("/"):
-            await self.handle_command(user_input)
+            # Defer command processing until after the UI renders
+            self.call_after_refresh(lambda: self.handle_command(user_input))
         else:
-            await self.handle_message(user_input)
+            # Defer assistant widget and AI call until after UI renders
+            async def setup_and_call_ai():
+                model_name = self.session.selected_model.get("name", "Assistant") if self.session.selected_model else "Assistant"
+                assistant_widget = AssistantMessageWidget(model_name, chat_display=chat_display)
+                await chat_display.add_message(assistant_widget)
+                assistant_widget.set_content("_Thinking... (Press Esc to stop)_")
+                await self.handle_message(user_input, assistant_widget)
+
+            self.call_after_refresh(setup_and_call_ai)
 
     async def handle_command(self, command_text: str) -> None:
         """Handle a slash command."""
@@ -449,6 +515,11 @@ class oAIChatApp(App):
             chat_display = self.query_one(ChatDisplay)
             error_widget = SystemMessageWidget(f"❌ {result.message}")
             await chat_display.add_message(error_widget)
+        elif result.status == CommandStatus.WARNING:
+            # Display warning in chat
+            chat_display = self.query_one(ChatDisplay)
+            warning_widget = SystemMessageWidget(f"⚠️  {result.message}")
+            await chat_display.add_message(warning_widget)
         elif result.message:
             # Display success message
             chat_display = self.query_one(ChatDisplay)
@@ -486,61 +557,78 @@ class oAIChatApp(App):
         # Update online mode indicator
         input_bar.update_online_mode(self.session.online_enabled)
 
-    async def handle_message(self, user_input: str) -> None:
-        """Handle a chat message (user message already added by caller)."""
+    async def handle_message(self, user_input: str, assistant_widget: AssistantMessageWidget = None) -> None:
+        """Handle a chat message (user message and assistant widget already added by caller)."""
         chat_display = self.query_one(ChatDisplay)
 
-        # Create assistant message widget with loading indicator
-        model_name = self.session.selected_model.get("name", "Assistant") if self.session.selected_model else "Assistant"
-        assistant_widget = AssistantMessageWidget(model_name, chat_display=chat_display)
-        await chat_display.add_message(assistant_widget)
+        # If no assistant widget provided (legacy), create one
+        if assistant_widget is None:
+            model_name = self.session.selected_model.get("name", "Assistant") if self.session.selected_model else "Assistant"
+            assistant_widget = AssistantMessageWidget(model_name, chat_display=chat_display)
+            await chat_display.add_message(assistant_widget)
+            assistant_widget.set_content("_Thinking... (Press Esc to stop)_")
 
-        # Show loading indicator immediately
-        assistant_widget.set_content("_Thinking..._")
+        self._is_generating = True
+        self._cancel_generation = False
 
-        try:
-            # Stream response
-            response_iterator = self.session.send_message_async(
-                user_input,
-                stream=self.settings.stream_enabled,
-            )
-
-            # Stream and collect response
-            full_text, usage = await assistant_widget.stream_response(response_iterator)
-
-            # Add to history if we got a response
-            if full_text:
-                # Extract cost from usage or calculate from pricing
-                cost = 0.0
-                if usage and hasattr(usage, 'total_cost_usd') and usage.total_cost_usd:
-                    cost = usage.total_cost_usd
-                    self.notify(f"Cost from API: ${cost:.6f}", severity="information")
-                elif usage and self.session.selected_model:
-                    # Calculate cost from model pricing
-                    pricing = self.session.selected_model.get("pricing", {})
-                    prompt_cost = float(pricing.get("prompt", 0))
-                    completion_cost = float(pricing.get("completion", 0))
-
-                    # Prices are per token, convert to dollars
-                    prompt_total = usage.prompt_tokens * prompt_cost
-                    completion_total = usage.completion_tokens * completion_cost
-                    cost = prompt_total + completion_total
-
-                    if cost > 0:
-                        self.notify(f"Cost calculated: ${cost:.6f}", severity="information")
-
-                self.session.add_to_history(
-                    prompt=user_input,
-                    response=full_text,
-                    usage=usage,
-                    cost=cost,
+        # Run streaming in background to keep UI responsive
+        async def stream_task():
+            try:
+                # Stream response
+                response_iterator = self.session.send_message_async(
+                    user_input,
+                    stream=self.settings.stream_enabled,
                 )
 
-            # Update footer
-            self._update_footer()
+                # Stream and collect response with cancellation support
+                full_text, usage = await assistant_widget.stream_response(
+                    response_iterator,
+                    cancel_check=lambda: self._cancel_generation
+                )
 
-        except Exception as e:
-            assistant_widget.set_content(f"❌ Error: {str(e)}")
+                # Add to history if we got a response
+                if full_text:
+                    # Extract cost from usage or calculate from pricing
+                    cost = 0.0
+                    if usage and hasattr(usage, 'total_cost_usd') and usage.total_cost_usd:
+                        cost = usage.total_cost_usd
+                        self.notify(f"Cost from API: ${cost:.6f}", severity="information")
+                    elif usage and self.session.selected_model:
+                        # Calculate cost from model pricing
+                        pricing = self.session.selected_model.get("pricing", {})
+                        prompt_cost = float(pricing.get("prompt", 0))
+                        completion_cost = float(pricing.get("completion", 0))
+
+                        # Prices are per token, convert to dollars
+                        prompt_total = usage.prompt_tokens * prompt_cost
+                        completion_total = usage.completion_tokens * completion_cost
+                        cost = prompt_total + completion_total
+
+                        if cost > 0:
+                            self.notify(f"Cost calculated: ${cost:.6f}", severity="information")
+
+                    self.session.add_to_history(
+                        prompt=user_input,
+                        response=full_text,
+                        usage=usage,
+                        cost=cost,
+                    )
+
+                # Update footer
+                self._update_footer()
+
+                # Check if generation was cancelled
+                if self._cancel_generation and full_text:
+                    assistant_widget.set_content(full_text + "\n\n_[Generation stopped by user]_")
+
+            except Exception as e:
+                assistant_widget.set_content(f"❌ Error: {str(e)}")
+            finally:
+                self._is_generating = False
+                self._cancel_generation = False
+
+        # Create background task - don't await it!
+        asyncio.create_task(stream_task())
 
     def _update_footer(self) -> None:
         """Update footer statistics."""
@@ -860,9 +948,8 @@ class oAIChatApp(App):
         elif "retry_prompt" in data:
             await self.handle_message(data["retry_prompt"])
 
-        # Paste prompt
-        elif "paste_prompt" in data:
-            await self.handle_message(data["paste_prompt"])
+        # Paste command is disabled - users should use Cmd+V/Ctrl+V instead
+        # No special handling needed
 
     def _show_model_selector(self, search: str = "", set_as_default: bool = False) -> None:
         """Show the model selector screen."""
@@ -994,7 +1081,7 @@ class oAIChatApp(App):
             callback=handle_confirmation
         )
 
-    def _navigate_history_backward(self, input_widget: Input) -> None:
+    def _navigate_history_backward(self, input_widget: TextArea) -> None:
         """Navigate backward through input history (Up arrow)."""
         if not self.input_history:
             return
@@ -1011,14 +1098,14 @@ class oAIChatApp(App):
 
         # Update input with history item
         if 0 <= self.history_index < len(self.input_history):
-            input_widget.value = self.input_history[self.history_index]
+            input_widget.text = self.input_history[self.history_index]
             # Move cursor to end
-            input_widget.cursor_position = len(input_widget.value)
+            input_widget.move_cursor((999, 999))  # Move to end
 
         # Clear flag after a short delay
         self.set_timer(0.1, lambda: setattr(self, "_navigating_history", False))
 
-    def _navigate_history_forward(self, input_widget: Input) -> None:
+    def _navigate_history_forward(self, input_widget: TextArea) -> None:
         """Navigate forward through input history (Down arrow)."""
         if not self.input_history or self.history_index == -1:
             return
@@ -1029,12 +1116,12 @@ class oAIChatApp(App):
         # Move forward in history
         if self.history_index < len(self.input_history) - 1:
             self.history_index += 1
-            input_widget.value = self.input_history[self.history_index]
-            input_widget.cursor_position = len(input_widget.value)
+            input_widget.text = self.input_history[self.history_index]
+            input_widget.move_cursor((999, 999))  # Move to end
         else:
             # At the newest item, clear the input
             self.history_index = -1
-            input_widget.value = ""
+            input_widget.clear()
 
         # Clear flag after a short delay
         self.set_timer(0.1, lambda: setattr(self, "_navigating_history", False))

oai/tui/screens/commands_screen.py

@@ -111,6 +111,7 @@ class CommandsScreen(ModalScreen[None]):
 [green]/config model <id>[/] - Set default model
 [green]/config system <prompt>[/] - Set system prompt
 [green]/config maxtoken <num>[/] - Set token limit
+[green]/config log <level>[/] - Set log level (debug/info/warning/error/critical)
 
 [bold cyan]Memory & Context[/]
 

oai/tui/screens/help_screen.py

@@ -64,6 +64,9 @@ class HelpScreen(ModalScreen[None]):
 [bold]F1[/]             Show this help (Ctrl+H may not work)
 [bold]F2[/]             Open model selector (Ctrl+M may not work)
 [bold]F3[/]             Copy last AI response to clipboard
+[bold]Enter[/]          Submit message
+[bold]Ctrl+Enter[/]     Insert newline (for multiline messages)
+[bold]Esc[/]            Stop/cancel AI response generation
 [bold]Ctrl+S[/]         Show session statistics
 [bold]Ctrl+L[/]         Clear chat display
 [bold]Ctrl+P[/]         Show previous message
@@ -71,7 +74,7 @@ class HelpScreen(ModalScreen[None]):
 [bold]Ctrl+Y[/]         Copy last AI response (alternative to F3)
 [bold]Ctrl+Q[/]         Quit application
 [bold]Up/Down[/]        Navigate input history
-[bold]ESC[/]            Close dialogs
+[bold]ESC[/]            Clear input / Close dialogs
 [dim]Note: Some Ctrl keys may be captured by your terminal[/]
 
 [bold cyan]═══ SLASH COMMANDS ═══[/]
@@ -89,6 +92,7 @@ class HelpScreen(ModalScreen[None]):
 /config stream on         Enable streaming responses
 /system [prompt]          Set session system prompt
 /maxtoken [n]             Set session token limit
+/config log [level]       Set log level (debug/info/warning/error/critical)
 
 [bold yellow]Conversation Management:[/]
 /save [name]              Save current conversation

oai/tui/widgets/chat_display.py

@@ -13,6 +13,7 @@ class ChatDisplay(ScrollableContainer):
     async def add_message(self, widget: Static) -> None:
         """Add a message widget to the display."""
         await self.mount(widget)
+        self.refresh(layout=True)
         self.scroll_end(animate=False)
 
     def clear_messages(self) -> None:

oai/tui/widgets/command_dropdown.py

@@ -98,6 +98,7 @@ class CommandDropdown(VerticalScroll):
             ("/config model", "Set default model"),
             ("/config system", "Set system prompt"),
             ("/config maxtoken", "Set token limit"),
+            ("/config log", "Set log level (debug/info/warning/error)"),
             ("/system", "Set system prompt"),
             ("/maxtoken", "Set token limit"),
             ("/retry", "Retry last prompt"),

oai/tui/widgets/input_bar.py

@@ -2,7 +2,53 @@
 
 from textual.app import ComposeResult
 from textual.containers import Horizontal
-from textual.widgets import Input, Static
+from textual.message import Message
+from textual.widgets import Static, TextArea
+
+
+class ChatTextArea(TextArea):
+    """Custom TextArea that sends submit message on Enter (unless dropdown is open)."""
+
+    class Submit(Message):
+        """Message sent when Enter is pressed."""
+        def __init__(self, value: str) -> None:
+            super().__init__()
+            self.value = value
+
+    def _on_key(self, event) -> None:
+        """Handle key events BEFORE TextArea processes them."""
+        # Check if command dropdown is visible
+        dropdown_visible = False
+        try:
+            from oai.tui.widgets.command_dropdown import CommandDropdown
+            dropdown = self.app.query_one(CommandDropdown)
+            dropdown_visible = dropdown.has_class("visible")
+        except:
+            pass
+
+        if event.key == "enter":
+            if dropdown_visible:
+                # Dropdown is open - prevent TextArea from inserting newline
+                # but let event bubble up to app for dropdown handling
+                event.prevent_default()
+                # Don't call stop() - let it bubble to app's on_key
+                # Don't call super - we don't want newline
+                return
+            else:
+                # Dropdown not visible - submit the message
+                event.prevent_default()
+                event.stop()
+                self.post_message(self.Submit(self.text))
+                return
+        elif event.key in ("ctrl+j", "ctrl+enter"):
+            # Insert newline on Ctrl+Enter
+            event.prevent_default()
+            event.stop()
+            self.insert("\n")
+            return
+
+        # For all other keys, let TextArea handle them normally
+        super()._on_key(event)
 
 
 class InputBar(Horizontal):
@@ -16,10 +62,9 @@ class InputBar(Horizontal):
     def compose(self) -> ComposeResult:
         """Compose the input bar."""
         yield Static(self._format_prefix(), id="input-prefix", classes="prefix-hidden" if not (self.mcp_status or self.online_mode) else "")
-        yield Input(
-            placeholder="Type a message or /command...",
-            id="chat-input"
-        )
+        text_area = ChatTextArea(id="chat-input")
+        text_area.show_line_numbers = False
+        yield text_area
 
     def _format_prefix(self) -> str:
         """Format the input prefix with status indicators."""
@@ -44,6 +89,6 @@ class InputBar(Horizontal):
         prefix = self.query_one("#input-prefix", Static)
         prefix.update(self._format_prefix())
 
-    def get_input(self) -> Input:
+    def get_input(self) -> ChatTextArea:
         """Get the input widget."""
-        return self.query_one("#chat-input", Input)
+        return self.query_one("#chat-input", ChatTextArea)

oai/tui/widgets/message.py

@@ -1,5 +1,6 @@
 """Message widgets for oAI TUI."""
 
+import asyncio
 from typing import Any, AsyncIterator, Tuple
 
 from rich.console import Console
@@ -35,7 +36,13 @@ class UserMessageWidget(Static):
 
     def compose(self) -> ComposeResult:
         """Compose the user message."""
-        yield Static(f"[bold green]You:[/] {self.content}")
+        yield Static(f"[bold green]You:[/] {self.content}", id="user-message-content")
+
+    def update_content(self, new_content: str) -> None:
+        """Update the message content."""
+        self.content = new_content
+        content_widget = self.query_one("#user-message-content", Static)
+        content_widget.update(f"[bold green]You:[/] {new_content}")
 
 
 class SystemMessageWidget(Static):
@@ -64,13 +71,22 @@ class AssistantMessageWidget(Static):
         yield Static(f"[bold]{self.model_name}:[/]", id="assistant-label")
         yield RichLog(id="assistant-content", highlight=True, markup=True, wrap=True)
 
-    async def stream_response(self, response_iterator: AsyncIterator) -> Tuple[str, Any]:
-        """Stream tokens progressively and return final text and usage."""
+    async def stream_response(self, response_iterator: AsyncIterator, cancel_check=None) -> Tuple[str, Any]:
+        """Stream tokens progressively and return final text and usage.
+
+        Args:
+            response_iterator: Async iterator of response chunks
+            cancel_check: Optional callable that returns True if generation should be cancelled
+        """
         log = self.query_one("#assistant-content", RichLog)
         self.full_text = ""
         usage = None
 
         async for chunk in response_iterator:
+            # Check for cancellation
+            if cancel_check and cancel_check():
+                break
+
             if hasattr(chunk, "delta_content") and chunk.delta_content:
                 self.full_text += chunk.delta_content
                 log.clear()
@@ -86,6 +102,9 @@ class AssistantMessageWidget(Static):
             if hasattr(chunk, "usage") and chunk.usage:
                 usage = chunk.usage
 
+            # Yield control to event loop so UI stays responsive
+            await asyncio.sleep(0)
+
         return self.full_text, usage
 
     def set_content(self, content: str) -> None:

oai/utils/files.py

@@ -278,11 +278,16 @@ def prepare_file_attachment(
             file_data = f.read()
 
         if category == "image":
-            # Check if model supports images
-            input_modalities = model_capabilities.get("architecture", {}).get("input_modalities", [])
-            if "image" not in input_modalities:
-                logger.warning(f"Model does not support images")
-                return None
+            # Check if model supports images - try multiple possible locations
+            input_modalities = (
+                model_capabilities.get("input_modalities", []) or
+                model_capabilities.get("architecture", {}).get("input_modalities", [])
+            )
+
+            # If no input_modalities found or image not in list, try to attach anyway
+            # Some models support images but don't advertise it properly
+            if input_modalities and "image" not in input_modalities:
+                logger.warning(f"Model may not support images, attempting anyway...")
 
             b64_data = base64.b64encode(file_data).decode("utf-8")
             return {
@@ -291,12 +296,16 @@ def prepare_file_attachment(
             }
 
         elif category == "pdf":
-            # Check if model supports PDFs
-            input_modalities = model_capabilities.get("architecture", {}).get("input_modalities", [])
+            # Check if model supports PDFs - try multiple possible locations
+            input_modalities = (
+                model_capabilities.get("input_modalities", []) or
+                model_capabilities.get("architecture", {}).get("input_modalities", [])
+            )
+
             supports_pdf = any(mod in input_modalities for mod in ["document", "pdf", "file"])
-            if not supports_pdf:
-                logger.warning(f"Model does not support PDFs")
-                return None
+
+            if input_modalities and not supports_pdf:
+                logger.warning(f"Model may not support PDFs, attempting anyway...")
 
             b64_data = base64.b64encode(file_data).decode("utf-8")
             return {
@@ -321,3 +330,49 @@ def prepare_file_attachment(
     except Exception as e:
         logger.error(f"Error preparing file attachment {path}: {e}")
         return None
+
+
+def parse_file_attachments(user_input: str) -> Tuple[str, list[Path]]:
+    """
+    Parse user input for @<file> or @file syntax and extract file paths.
+
+    Args:
+        user_input: User's message that may contain @<file> or @file references
+
+    Returns:
+        Tuple of (cleaned_text, list_of_file_paths)
+
+    Example:
+        >>> parse_file_attachments("Look at @<image.png> and @/path/to/doc.pdf")
+        ("Look at  and ", [Path("image.png"), Path("/path/to/doc.pdf")])
+    """
+    import re
+    logger = get_logger()
+
+    # Pattern to match both @<filepath> and @filepath
+    # Matches @<...> or @followed by path-like strings
+    pattern = r'@<([^>]+)>|@([/~.][\S]+|[a-zA-Z]:[/\\][\S]+)'
+
+    # Find all file references
+    file_paths = []
+    matches_to_remove = []
+
+    for match in re.finditer(pattern, user_input):
+        # Group 1 is for @<filepath>, Group 2 is for @filepath
+        file_path_str = (match.group(1) or match.group(2)).strip()
+        file_path = Path(file_path_str).expanduser().resolve()
+
+        if file_path.exists():
+            file_paths.append(file_path)
+            logger.info(f"Found file attachment: {file_path}")
+            matches_to_remove.append(match.group(0))
+        else:
+            logger.warning(f"File not found: {file_path_str}")
+            matches_to_remove.append(match.group(0))
+
+    # Remove @<file> and @file references from the text
+    cleaned_text = user_input
+    for match_str in matches_to_remove:
+        cleaned_text = cleaned_text.replace(match_str, '', 1)
+
+    return cleaned_text, file_paths

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "oai"
-version = "3.0.0-b3"  # MUST match oai/__init__.py __version__
+version = "3.0.0-b4"  # MUST match oai/__init__.py __version__
 description = "Open AI Chat Client - Multi-provider terminal chat with MCP support"
 readme = "README.md"
 license = {text = "MIT"}