orgo 0.0.9__tar.gz → 0.0.11__tar.gz

{orgo-0.0.9 → orgo-0.0.11}/PKG-INFO

@@ -1,7 +1,7 @@
 Metadata-Version: 2.4
 Name: orgo
-Version: 0.0.9
-Summary: Desktop infrastructure for AI agents
+Version: 0.0.11
+Summary: Computers for AI agents
 Author: Orgo Team
 License: MIT
 Project-URL: Homepage, https://www.orgo.ai

{orgo-0.0.9 → orgo-0.0.11}/pyproject.toml

@@ -4,8 +4,8 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "orgo"
-version = "0.0.9"
-description = "Desktop infrastructure for AI agents"
+version = "0.0.11"
+description = "Computers for AI agents"
 authors = [{name = "Orgo Team"}]
 license = {text = "MIT"}
 readme = "README.md"

{orgo-0.0.9 → orgo-0.0.11}/src/orgo/api/client.py

@@ -1,4 +1,5 @@
 """API client for Orgo service"""
+# src/orgo/api/client.py
 
 import requests
 from typing import Dict, Any, Optional
@@ -6,10 +7,9 @@ from typing import Dict, Any, Optional
 from orgo.utils.auth import get_api_key
 
 class ApiClient:
-    BASE_URL = "https://www.orgo.ai/api"
-    
-    def __init__(self, api_key: Optional[str] = None):
+    def __init__(self, api_key: Optional[str] = None, base_url: Optional[str] = None):
         self.api_key = get_api_key(api_key)
+        self.base_url = base_url or "https://www.orgo.ai/api"
         self.session = requests.Session()
         self.session.headers.update({
             "Authorization": f"Bearer {self.api_key}",
@@ -18,7 +18,7 @@ class ApiClient:
         })
     
     def _request(self, method: str, endpoint: str, data: Optional[Dict[str, Any]] = None) -> Dict[str, Any]:
-        url = f"{self.BASE_URL}/{endpoint}"
+        url = f"{self.base_url}/{endpoint}"
         
         try:
             if method.upper() == "GET":

orgo-0.0.11/src/orgo/computer.py

@@ -0,0 +1,176 @@
+"""Computer class for interacting with Orgo virtual environments"""
+# src/orgo/computer.py
+
+import os
+import io
+import base64
+from typing import Dict, List, Any, Optional, Callable, Union
+from PIL import Image
+
+from .api.client import ApiClient
+from .prompt import get_provider
+
+
+class Computer:
+    def __init__(self, project_id=None, api_key=None, config=None, base_api_url=None):
+        """
+        Initialize an Orgo virtual computer.
+        
+        Args:
+            project_id: Existing project ID to connect to (optional)
+            api_key: Orgo API key (defaults to ORGO_API_KEY env var)
+            config: Configuration for new computer (optional)
+            base_api_url: Custom API URL (optional)
+        """
+        self.api_key = api_key or os.environ.get("ORGO_API_KEY")
+        self.base_api_url = base_api_url
+        self.api = ApiClient(self.api_key, self.base_api_url)
+        
+        if project_id:
+            self.project_id = project_id
+            self._info = self.api.connect_computer(project_id)
+        else:
+            response = self.api.create_computer(config)
+            self.project_id = response.get("name")
+            self._info = response
+            
+        if not self.project_id:
+            raise ValueError("Failed to initialize computer: No project ID returned")
+    
+    def status(self) -> Dict[str, Any]:
+        """Get current computer status"""
+        return self.api.get_status(self.project_id)
+    
+    def restart(self) -> Dict[str, Any]:
+        """Restart the computer"""
+        return self.api.restart_computer(self.project_id)
+    
+    def shutdown(self) -> Dict[str, Any]:
+        """Terminate the computer instance"""
+        return self.api.shutdown_computer(self.project_id)
+    
+    # Navigation methods
+    def left_click(self, x: int, y: int) -> Dict[str, Any]:
+        """Perform left mouse click at specified coordinates"""
+        return self.api.left_click(self.project_id, x, y)
+    
+    def right_click(self, x: int, y: int) -> Dict[str, Any]:
+        """Perform right mouse click at specified coordinates"""
+        return self.api.right_click(self.project_id, x, y)
+    
+    def double_click(self, x: int, y: int) -> Dict[str, Any]:
+        """Perform double click at specified coordinates"""
+        return self.api.double_click(self.project_id, x, y)
+    
+    def scroll(self, direction: str = "down", amount: int = 1) -> Dict[str, Any]:
+        """Scroll in specified direction and amount"""
+        return self.api.scroll(self.project_id, direction, amount)
+    
+    # Input methods
+    def type(self, text: str) -> Dict[str, Any]:
+        """Type the specified text"""
+        return self.api.type_text(self.project_id, text)
+    
+    def key(self, key: str) -> Dict[str, Any]:
+        """Press a key or key combination (e.g., "Enter", "ctrl+c")"""
+        return self.api.key_press(self.project_id, key)
+    
+    # View methods
+    def screenshot(self) -> Image.Image:
+        """Capture screenshot and return as PIL Image"""
+        response = self.api.get_screenshot(self.project_id)
+        img_data = base64.b64decode(response.get("image", ""))
+        return Image.open(io.BytesIO(img_data))
+    
+    def screenshot_base64(self) -> str:
+        """Capture screenshot and return as base64 string"""
+        response = self.api.get_screenshot(self.project_id)
+        return response.get("image", "")
+    
+    # Execution methods
+    def bash(self, command: str) -> str:
+        """Execute a bash command and return output"""
+        response = self.api.execute_bash(self.project_id, command)
+        return response.get("output", "")
+    
+    def wait(self, seconds: float) -> Dict[str, Any]:
+        """Wait for specified number of seconds"""
+        return self.api.wait(self.project_id, seconds)
+    
+    # AI control method
+    def prompt(self, 
+               instruction: str,
+               provider: str = "anthropic",
+               model: str = "claude-3-7-sonnet-20250219",
+               display_width: int = 1024,
+               display_height: int = 768,
+               callback: Optional[Callable[[str, Any], None]] = None,
+               thinking_enabled: bool = False,
+               thinking_budget: int = 1024,
+               max_tokens: int = 4096,
+               max_iterations: int = 20,
+               max_saved_screenshots: int = 5,
+               api_key: Optional[str] = None) -> List[Dict[str, Any]]:
+        """
+        Control the computer with natural language instructions using an AI assistant.
+        
+        Args:
+            instruction: What you want the AI to do with the computer
+            provider: AI provider to use (default: "anthropic")
+            model: Model to use (default: "claude-3-7-sonnet-20250219")
+            display_width: Screen width in pixels
+            display_height: Screen height in pixels
+            callback: Optional callback function for progress updates
+            thinking_enabled: Enable Claude's thinking capability (default: False)
+            thinking_budget: Token budget for thinking (default: 1024)
+            max_tokens: Maximum tokens for model response
+            max_iterations: Maximum number of agent loop iterations
+            max_saved_screenshots: Maximum number of screenshots to keep in history (default: 5)
+            api_key: API key for the AI provider (defaults to env var)
+            
+        Returns:
+            List of messages from the conversation
+            
+        Examples:
+            # Simple usage with environment variables
+            computer.prompt("Open Firefox and search for Python tutorials")
+            
+            # With explicit API key
+            computer.prompt("Open Terminal and list files", api_key="your-anthropic-key")
+            
+            # With callback for progress updates
+            computer.prompt("Create a new text file", callback=my_callback_function)
+            
+            # With thinking enabled (Claude 3.7 Sonnet)
+            computer.prompt(
+                "Analyze a complex webpage", 
+                thinking_enabled=True
+            )
+            
+            # With custom screenshot management
+            computer.prompt(
+                "Perform a complex multi-step task",
+                max_saved_screenshots=10  # Keep more screenshots for complex tasks
+            )
+        """
+        # Get the provider instance
+        provider_instance = get_provider(provider)
+        
+        # Execute the prompt
+        return provider_instance.execute(
+            computer_id=self.project_id,
+            instruction=instruction,
+            callback=callback,
+            api_key=api_key,
+            model=model,
+            display_width=display_width,
+            display_height=display_height,
+            thinking_enabled=thinking_enabled,
+            thinking_budget=thinking_budget,
+            max_tokens=max_tokens,
+            max_iterations=max_iterations,
+            max_saved_screenshots=max_saved_screenshots,
+            # Pass through the Orgo API client configuration
+            orgo_api_key=self.api_key,
+            orgo_base_url=self.base_api_url
+        )

orgo-0.0.11/src/orgo/prompt.py

@@ -0,0 +1,405 @@
+"""
+Prompt module for interacting with virtual computers using AI models.
+"""
+
+import os
+from typing import Dict, List, Any, Optional, Callable, Union, Protocol
+
+
+class PromptProvider(Protocol):
+    """Protocol defining the interface for prompt providers."""
+    
+    def execute(self, 
+                computer_id: str,
+                instruction: str,
+                callback: Optional[Callable[[str, Any], None]] = None,
+                **kwargs) -> List[Dict[str, Any]]:
+        """
+        Execute a prompt to control the computer.
+        
+        Args:
+            computer_id: ID of the computer to control
+            instruction: User instruction
+            callback: Optional progress callback function
+            **kwargs: Additional provider-specific parameters
+            
+        Returns:
+            List of messages from the conversation
+        """
+        ...
+
+
+class AnthropicProvider:
+    """Anthropic Claude-based prompt provider."""
+    
+    def __init__(self):
+        """Initialize the Anthropic provider."""
+        try:
+            import anthropic
+            self.anthropic = anthropic
+        except ImportError:
+            raise ImportError(
+                "Anthropic SDK not installed. Please install with 'pip install anthropic'"
+            )
+    
+    def execute(self,
+                computer_id: str,
+                instruction: str,
+                callback: Optional[Callable[[str, Any], None]] = None,
+                api_key: Optional[str] = None,
+                model: str = "claude-3-7-sonnet-20250219",
+                display_width: int = 1024,
+                display_height: int = 768,
+                orgo_api_key: Optional[str] = None,
+                orgo_base_url: Optional[str] = None,
+                max_saved_screenshots: int = 2,
+                **kwargs) -> List[Dict[str, Any]]:
+        """
+        Execute a prompt using Anthropic's Claude.
+        
+        Args:
+            computer_id: ID of the computer to control
+            instruction: User instruction
+            callback: Optional progress callback
+            api_key: Anthropic API key
+            model: Model to use
+            display_width: Display width in pixels
+            display_height: Display height in pixels
+            orgo_api_key: API key for Orgo (passed to ApiClient)
+            orgo_base_url: Base URL for Orgo API (passed to ApiClient)
+            max_saved_screenshots: Maximum number of screenshots to maintain in conversation history
+            **kwargs: Additional parameters to pass to the Anthropic API
+            
+        Returns:
+            List of messages from the conversation
+        """
+        # Get API key from kwargs, env var, or raise error
+        api_key = api_key or os.environ.get("ANTHROPIC_API_KEY")
+        if not api_key:
+            raise ValueError("No Anthropic API key provided. Set ANTHROPIC_API_KEY environment variable or pass api_key.")
+        
+        # Initialize the client
+        client = self.anthropic.Anthropic(api_key=api_key)
+        
+        # Prepare the messages
+        messages = [{"role": "user", "content": instruction}]
+        
+        # Set up the system prompt
+        system_prompt = f"""You are Claude, an AI assistant that controls a virtual Ubuntu computer with internet access.
+
+<SYSTEM_CAPABILITY>
+* You are utilising an Ubuntu virtual machine with a display resolution of {display_width}x{display_height}.
+* You can take screenshots to see the current state and control the computer by clicking, typing, pressing keys, and scrolling.
+* The virtual environment is an Ubuntu system with standard applications.
+* Always start by taking a screenshot to see the current state before performing any actions.
+</SYSTEM_CAPABILITY>
+
+<UBUNTU_DESKTOP_GUIDELINES>
+* CRITICAL INSTRUCTION: When opening applications or files on the Ubuntu desktop, you MUST USE DOUBLE-CLICK rather than single-click.
+* Single-click only selects desktop icons but DOES NOT open them. To open desktop icons, you MUST use double-click.
+* Common desktop interactions:
+  - Desktop icons: DOUBLE-CLICK to open applications and folders
+  - Menu items: SINGLE-CLICK to select options
+  - Taskbar icons: SINGLE-CLICK to open applications
+  - Window buttons: SINGLE-CLICK to use close, minimize, maximize buttons
+  - File browser items: DOUBLE-CLICK to open folders and files
+  - When submitting, use the 'Enter' key, not the 'Return' key.
+* If you see an icon on the desktop that you need to open, ALWAYS use the double_click action, never use left_click.
+</UBUNTU_DESKTOP_GUIDELINES>"""
+        
+        try:
+            # Define the computer tool per Anthropic's documentation
+            tools = [
+                {
+                    "type": "computer_20250124",
+                    "name": "computer",
+                    "display_width_px": display_width,
+                    "display_height_px": display_height,
+                    "display_number": 1
+                }
+            ]
+            
+            # Start the conversation with Claude
+            if callback:
+                callback("status", "Starting conversation with Claude")
+            
+            # Track whether we're in the agent loop
+            iteration = 0
+            max_iterations = kwargs.get("max_iterations", 20)  # Default to 20 iterations max
+            
+            # Create an API client with the proper settings
+            from .api.client import ApiClient
+            api_client = ApiClient(orgo_api_key, orgo_base_url)
+            
+            # Track how many screenshots we've seen so we can prune when needed
+            screenshot_count = 0
+            
+            # Start the agent loop
+            while iteration < max_iterations:
+                iteration += 1
+                
+                # Prune old screenshots if we've exceeded our limit
+                if screenshot_count > max_saved_screenshots:
+                    self._prune_old_screenshots(messages, screenshot_count - max_saved_screenshots)
+                    screenshot_count = max_saved_screenshots
+                
+                # Create the request parameters
+                request_params = {
+                    "model": model,
+                    "max_tokens": kwargs.get("max_tokens", 4096),
+                    "system": system_prompt,
+                    "messages": messages,
+                    "tools": tools,
+                    "betas": ["computer-use-2025-01-24"],
+                }
+                
+                # Add thinking parameter only if explicitly enabled
+                if kwargs.get("thinking_enabled"):
+                    request_params["thinking"] = {
+                        "type": "enabled", 
+                        "budget_tokens": kwargs.get("thinking_budget", 1024)
+                    }
+                
+                # Create message request to Claude
+                response = client.beta.messages.create(**request_params)
+                
+                # Extract the content from the response
+                response_content = response.content
+                
+                # Add Claude's response to the conversation history
+                assistant_message = {"role": "assistant", "content": response_content}
+                messages.append(assistant_message)
+                
+                # Notify callback of any text content
+                for block in response_content:
+                    if block.type == "text" and callback:
+                        callback("text", block.text)
+                    elif block.type == "thinking" and callback:
+                        callback("thinking", block.thinking)
+                    elif block.type == "tool_use" and callback:
+                        tool_params = {
+                            "action": block.name.split(".")[-1],
+                            **block.input
+                        }
+                        callback("tool_use", tool_params)
+                
+                # Check if Claude requested any tool actions
+                tool_results = []
+                for block in response_content:
+                    if block.type == "tool_use":
+                        # Execute the tool action
+                        result = self._execute_tool(computer_id, block.input, callback, api_client)
+                        
+                        # Format the result for Claude
+                        tool_result = {
+                            "type": "tool_result",
+                            "tool_use_id": block.id
+                        }
+                        
+                        # Handle image vs text results
+                        if isinstance(result, dict) and "type" in result and result["type"] == "image":
+                            tool_result["content"] = [result]
+                            # Increment screenshot count when we add a new screenshot
+                            if block.input.get("action") == "screenshot":
+                                screenshot_count += 1
+                        else:
+                            tool_result["content"] = [{"type": "text", "text": str(result)}]
+                        
+                        tool_results.append(tool_result)
+                
+                # If no tools were used, Claude is done - return the messages
+                if not tool_results:
+                    if callback:
+                        callback("status", "Task completed")
+                    return messages
+                
+                # Add tool results to messages for the next iteration
+                messages.append({"role": "user", "content": tool_results})
+            
+            # We've reached the maximum iteration limit
+            if callback:
+                callback("status", f"Reached maximum iterations ({max_iterations})")
+            
+            return messages
+            
+        except Exception as e:
+            if callback:
+                callback("error", str(e))
+            raise
+            
+    def _prune_old_screenshots(self, messages: List[Dict[str, Any]], num_to_prune: int):
+        """
+        Remove old screenshots from the conversation history.
+        
+        Args:
+            messages: The conversation history
+            num_to_prune: Number of screenshots to remove
+        """
+        screenshots_pruned = 0
+        
+        # Start from the beginning of the messages (excluding the first user message)
+        for i in range(1, len(messages)):
+            if messages[i]["role"] != "user":
+                continue
+                
+            content = messages[i]["content"]
+            if not isinstance(content, list):
+                continue
+                
+            # Look for tool_result blocks in the content
+            for j, block in enumerate(content):
+                if not isinstance(block, dict):
+                    continue
+                    
+                if block.get("type") != "tool_result":
+                    continue
+                    
+                # Check if this tool_result contains an image
+                block_content = block.get("content", [])
+                for k, content_item in enumerate(block_content):
+                    if not isinstance(content_item, dict):
+                        continue
+                        
+                    if content_item.get("type") == "image":
+                        # This is a screenshot, remove it
+                        if "source" in content_item and "data" in content_item["source"]:
+                            # Replace the base64 data with a placeholder to save space
+                            content_item["source"]["data"] = "[IMAGE DATA REMOVED]"
+                            screenshots_pruned += 1
+                            
+                            if screenshots_pruned >= num_to_prune:
+                                return
+    
+    def _execute_tool(self, 
+                      computer_id: str,
+                      params: Dict[str, Any], 
+                      callback: Optional[Callable[[str, Any], None]] = None,
+                      api_client = None) -> Union[str, Dict[str, Any]]:
+        """Execute a tool action via the API client."""
+        action = params.get("action")
+        
+        if callback:
+            callback("tool_executing", {"action": action, "params": params})
+        
+        try:
+            # Use the provided API client or create a new one
+            if api_client is None:
+                # Import here to avoid circular imports
+                from .api.client import ApiClient
+                api_client = ApiClient()
+            
+            # Map actions to API methods
+            if action == "screenshot":
+                response = api_client.get_screenshot(computer_id)
+                if callback:
+                    callback("tool_result", {"type": "image", "action": "screenshot"})
+                return {
+                    "type": "image",
+                    "source": {
+                        "type": "base64",
+                        "media_type": "image/jpeg",
+                        "data": response.get("image", "")
+                    }
+                }
+                
+            elif action == "left_click":
+                if not params.get("coordinate"):
+                    raise ValueError("Coordinates required for left click")
+                x, y = params["coordinate"]
+                api_client.left_click(computer_id, x, y)
+                if callback:
+                    callback("tool_result", {"action": "left_click", "x": x, "y": y})
+                return f"Left-clicked at ({x}, {y})"
+                
+            elif action == "right_click":
+                if not params.get("coordinate"):
+                    raise ValueError("Coordinates required for right click")
+                x, y = params["coordinate"]
+                api_client.right_click(computer_id, x, y)
+                if callback:
+                    callback("tool_result", {"action": "right_click", "x": x, "y": y})
+                return f"Right-clicked at ({x}, {y})"
+                
+            elif action == "double_click":
+                if not params.get("coordinate"):
+                    raise ValueError("Coordinates required for double click")
+                x, y = params["coordinate"]
+                api_client.double_click(computer_id, x, y)
+                if callback:
+                    callback("tool_result", {"action": "double_click", "x": x, "y": y})
+                return f"Double-clicked at ({x}, {y})"
+                
+            elif action == "type":
+                if not params.get("text"):
+                    raise ValueError("Text required for typing")
+                text = params["text"]
+                api_client.type_text(computer_id, text)
+                if callback:
+                    callback("tool_result", {"action": "type", "text": text})
+                return f"Typed: \"{text}\""
+                
+            elif action == "key":
+                if not params.get("text"):
+                    raise ValueError("Key required for key press")
+                key = params["text"]
+                # Handle the 'return' key as 'enter' when needed
+                if key.lower() == "return":
+                    key = "enter"
+                api_client.key_press(computer_id, key)
+                if callback:
+                    callback("tool_result", {"action": "key", "key": key})
+                return f"Pressed key: {key}"
+                
+            elif action == "scroll":
+                if not params.get("scroll_direction") or params.get("scroll_amount") is None:
+                    raise ValueError("Direction and amount required for scrolling")
+                direction = params["scroll_direction"]
+                amount = params["scroll_amount"]
+                api_client.scroll(computer_id, direction, amount)
+                if callback:
+                    callback("tool_result", {"action": "scroll", "direction": direction, "amount": amount})
+                return f"Scrolled {direction} by {amount}"
+                
+            elif action == "wait":
+                duration = params.get("duration", 1)
+                api_client.wait(computer_id, duration)
+                if callback:
+                    callback("tool_result", {"action": "wait", "duration": duration})
+                return f"Waited for {duration} second(s)"
+                
+            else:
+                error_msg = f"Unsupported action: {action}"
+                if callback:
+                    callback("error", error_msg)
+                raise ValueError(error_msg)
+                
+        except Exception as e:
+            error_msg = f"Error executing {action}: {str(e)}"
+            if callback:
+                callback("error", error_msg)
+            return f"Error: {error_msg}"
+
+
+# Default provider mapping
+PROVIDER_MAPPING = {
+    "anthropic": AnthropicProvider,
+    # Add more providers here as needed, e.g.:
+    # "openai": OpenAIProvider,
+    # "fireworks": FireworksProvider,
+}
+
+
+def get_provider(provider_name: str = "anthropic") -> PromptProvider:
+    """
+    Get a prompt provider by name.
+    
+    Args:
+        provider_name: Name of the provider
+        
+    Returns:
+        Provider instance
+    """
+    if provider_name not in PROVIDER_MAPPING:
+        raise ValueError(f"Unknown provider: {provider_name}. Available providers: {', '.join(PROVIDER_MAPPING.keys())}")
+    
+    return PROVIDER_MAPPING[provider_name]()

{orgo-0.0.9 → orgo-0.0.11}/src/orgo.egg-info/PKG-INFO

@@ -1,7 +1,7 @@
 Metadata-Version: 2.4
 Name: orgo
-Version: 0.0.9
-Summary: Desktop infrastructure for AI agents
+Version: 0.0.11
+Summary: Computers for AI agents
 Author: Orgo Team
 License: MIT
 Project-URL: Homepage, https://www.orgo.ai

{orgo-0.0.9 → orgo-0.0.11}/src/orgo.egg-info/SOURCES.txt

@@ -2,6 +2,7 @@ README.md
 pyproject.toml
 src/orgo/__init__.py
 src/orgo/computer.py
+src/orgo/prompt.py
 src/orgo.egg-info/PKG-INFO
 src/orgo.egg-info/SOURCES.txt
 src/orgo.egg-info/dependency_links.txt

orgo-0.0.9/src/orgo/computer.py

@@ -1,84 +0,0 @@
-"""Computer class for interacting with Orgo virtual environments"""
-
-import os
-import io
-import base64
-from typing import Dict, Any
-from PIL import Image
-
-from .api.client import ApiClient
-
-class Computer:
-    def __init__(self, project_id=None, api_key=None, config=None):
-        self.api = ApiClient(api_key or os.environ.get("ORGO_API_KEY"))
-        
-        if project_id:
-            self.project_id = project_id
-            self._info = self.api.connect_computer(project_id)
-        else:
-            response = self.api.create_computer(config)
-            self.project_id = response.get("name")
-            self._info = response
-            
-        if not self.project_id:
-            raise ValueError("Failed to initialize computer: No project ID returned")
-    
-    def status(self) -> Dict[str, Any]:
-        """Get current computer status"""
-        return self.api.get_status(self.project_id)
-    
-    def restart(self) -> Dict[str, Any]:
-        """Restart the computer"""
-        return self.api.restart_computer(self.project_id)
-    
-    def shutdown(self) -> Dict[str, Any]:
-        """Terminate the computer instance"""
-        return self.api.shutdown_computer(self.project_id)
-    
-    # Navigation methods
-    def left_click(self, x: int, y: int) -> Dict[str, Any]:
-        """Perform left mouse click at specified coordinates"""
-        return self.api.left_click(self.project_id, x, y)
-    
-    def right_click(self, x: int, y: int) -> Dict[str, Any]:
-        """Perform right mouse click at specified coordinates"""
-        return self.api.right_click(self.project_id, x, y)
-    
-    def double_click(self, x: int, y: int) -> Dict[str, Any]:
-        """Perform double click at specified coordinates"""
-        return self.api.double_click(self.project_id, x, y)
-    
-    def scroll(self, direction: str = "down", amount: int = 1) -> Dict[str, Any]:
-        """Scroll in specified direction and amount"""
-        return self.api.scroll(self.project_id, direction, amount)
-    
-    # Input methods
-    def type(self, text: str) -> Dict[str, Any]:
-        """Type the specified text"""
-        return self.api.type_text(self.project_id, text)
-    
-    def key(self, key: str) -> Dict[str, Any]:
-        """Press a key or key combination (e.g., "Enter", "ctrl+c")"""
-        return self.api.key_press(self.project_id, key)
-    
-    # View methods
-    def screenshot(self) -> Image.Image:
-        """Capture screenshot and return as PIL Image"""
-        response = self.api.get_screenshot(self.project_id)
-        img_data = base64.b64decode(response.get("image", ""))
-        return Image.open(io.BytesIO(img_data))
-    
-    def screenshot_base64(self) -> str:
-        """Capture screenshot and return as base64 string"""
-        response = self.api.get_screenshot(self.project_id)
-        return response.get("image", "")
-    
-    # Execution methods
-    def bash(self, command: str) -> str:
-        """Execute a bash command and return output"""
-        response = self.api.execute_bash(self.project_id, command)
-        return response.get("output", "")
-    
-    def wait(self, seconds: float) -> Dict[str, Any]:
-        """Wait for specified number of seconds"""
-        return self.api.wait(self.project_id, seconds)