hud-python 0.2.10__py3-none-any.whl → 0.3.0__py3-none-any.whl

hud/tools/helper/server_initialization.py

@@ -0,0 +1,115 @@
+"""Helper for MCP server initialization with progress notifications.
+
+Example:
+    ```python
+    from hud.tools.helper import mcp_intialize_wrapper
+
+
+    @mcp_intialize_wrapper
+    async def initialize_environment(session=None, progress_token=None):
+        # Send progress if available
+        if session and progress_token:
+            await session.send_progress_notification(
+                progress_token=progress_token, progress=0, total=100, message="Starting services..."
+            )
+
+        # Your initialization code works with or without session
+        start_services()
+
+
+    # Create and run server - initialization happens automatically
+    mcp = FastMCP("My Server")
+    mcp.run()
+    ```
+"""
+
+from __future__ import annotations
+
+from typing import TYPE_CHECKING
+
+import mcp.types as types
+from mcp.server.session import ServerSession
+
+if TYPE_CHECKING:
+    from collections.abc import Awaitable, Callable
+
+    from mcp.shared.session import RequestResponder
+
+# Store the original _received_request method
+_original_received_request = ServerSession._received_request
+_init_function: Callable | None = None
+_initialized = False
+
+
+async def _patched_received_request(
+    self: ServerSession, responder: RequestResponder[types.ClientRequest, types.ServerResult]
+) -> types.ServerResult | None:
+    """Intercept initialization to run custom setup with progress notifications."""
+    global _initialized, _init_function
+
+    # Check if this is an initialization request
+    if isinstance(responder.request.root, types.InitializeRequest):
+        params = responder.request.root.params
+        # Extract progress token if present
+        progress_token = None
+        if hasattr(params, "meta") and params.meta and hasattr(params.meta, "progressToken"):
+            progress_token = params.meta.progressToken
+
+        # Run our initialization function if provided and not already done
+        if _init_function and not _initialized:
+            try:
+                await _init_function(session=self, progress_token=progress_token)
+                ServerSession._received_request = _original_received_request
+            except Exception as e:
+                if progress_token:
+                    await self.send_progress_notification(
+                        progress_token=progress_token,
+                        progress=0,
+                        total=100,
+                        message=f"Initialization failed: {e!s}",
+                    )
+                raise
+
+    # Call the original handler to send the InitializeResult
+    result = await _original_received_request(self, responder)
+    _initialized = True
+
+    return result
+
+
+def mcp_intialize_wrapper(
+    init_function: Callable[[ServerSession | None, str | None], Awaitable[None]] | None = None,
+) -> Callable:
+    """Decorator to enable progress notifications during MCP server initialization.
+
+    Your init function receives optional session and progress_token parameters.
+    If provided, use them to send progress updates. If not, the function still works.
+
+    Usage:
+        @mcp_intialize_wrapper
+        async def initialize(session=None, progress_token=None):
+            if session and progress_token:
+                await session.send_progress_notification(...)
+            # Your init code here
+
+    Must be applied before creating FastMCP instance or calling mcp.run().
+    """
+    global _init_function
+
+    def decorator(func: Callable[[ServerSession | None, str | None], Awaitable[None]]) -> Callable:
+        global _init_function
+        # Store the initialization function
+        _init_function = func
+
+        # Apply the monkey patch if not already applied
+        if ServerSession._received_request != _patched_received_request:
+            ServerSession._received_request = _patched_received_request  # type: ignore[assignment]
+
+        return func
+
+    # If called with a function directly
+    if init_function is not None:
+        return decorator(init_function)
+
+    # If used as @decorator
+    return decorator

hud/tools/helper/utils.py

@@ -0,0 +1,58 @@
+from __future__ import annotations
+
+import asyncio
+import inspect
+from functools import wraps
+from typing import TYPE_CHECKING, Any
+
+if TYPE_CHECKING:
+    from collections.abc import Callable
+
+    from mcp.server.fastmcp import FastMCP
+
+
+def register_instance_tool(mcp: FastMCP, name: str, instance: Any) -> Callable[..., Any]:
+    """Register ``instance.__call__`` as a FastMCP tool.
+
+    Parameters
+    ----------
+    mcp:
+        A :class:`mcp.server.fastmcp.FastMCP` instance.
+    name:
+        Public tool name.
+    instance:
+        Object with an ``async def __call__`` (or sync) implementing the tool.
+    """
+
+    if inspect.isclass(instance):
+        class_name = instance.__name__
+        raise TypeError(
+            f"register_instance_tool() expects an instance, but got class '{class_name}'. "
+            f"Use: register_instance_tool(mcp, '{name}', {class_name}()) "
+            f"Not: register_instance_tool(mcp, '{name}', {class_name})"
+        )
+
+    call_fn = instance.__call__
+    sig = inspect.signature(call_fn)
+
+    # Remove *args/**kwargs so Pydantic doesn't treat them as required fields
+    from typing import Any as _Any
+
+    filtered = [
+        p.replace(kind=p.POSITIONAL_OR_KEYWORD, annotation=_Any)
+        for p in sig.parameters.values()
+        if p.kind not in (p.VAR_POSITIONAL, p.VAR_KEYWORD)
+    ]
+
+    public_sig = inspect.Signature(parameters=filtered, return_annotation=_Any)
+
+    @wraps(call_fn)
+    async def _wrapper(*args: Any, **kwargs: Any) -> Any:  # type: ignore[override]
+        result = call_fn(*args, **kwargs)
+        if asyncio.iscoroutine(result):
+            result = await result
+        return result
+
+    _wrapper.__signature__ = public_sig  # type: ignore[attr-defined]
+
+    return mcp.tool(name=name)(_wrapper)

hud/tools/playwright_tool.py

@@ -0,0 +1,373 @@
+"""Playwright web automation tool for HUD."""
+
+from __future__ import annotations
+
+import logging
+import os
+from typing import TYPE_CHECKING, Any, Literal
+
+from mcp import ErrorData, McpError
+from mcp.types import INVALID_PARAMS, ImageContent, TextContent
+from pydantic import Field
+
+from hud.tools.base import ToolResult, tool_result_to_content_blocks
+
+if TYPE_CHECKING:
+    from playwright.async_api import Browser, BrowserContext, Page
+
+logger = logging.getLogger(__name__)
+
+
+class PlaywrightTool:
+    """Playwright tool for web automation."""
+
+    def __init__(self) -> None:
+        super().__init__()
+        self._playwright = None
+        self._browser: Browser | None = None
+        self._context: BrowserContext | None = None
+        self._page: Page | None = None
+
+    @property
+    def page(self) -> Page:
+        """Get the current page, raising an error if not initialized."""
+        if self._page is None:
+            raise RuntimeError("Browser page is not initialized. Call ensure_browser_launched().")
+        return self._page
+
+    async def __call__(
+        self,
+        action: str = Field(
+            ...,
+            description="The action to perform (navigate, screenshot, click, type, get_page_info, wait_for_element)",  # noqa: E501
+        ),
+        url: str | None = Field(None, description="URL to navigate to (for navigate action)"),
+        selector: str | None = Field(
+            None, description="CSS selector for element (for click, type, wait_for_element actions)"
+        ),
+        text: str | None = Field(None, description="Text to type (for type action)"),
+        path: str | None = Field(
+            None, description="File path to save screenshot (for screenshot action)"
+        ),
+        wait_for_load_state: Literal["commit", "domcontentloaded", "load", "networkidle"]
+        | None = Field(
+            None,
+            description="State to wait for: commit, domcontentloaded, load, networkidle (default: networkidle)",  # noqa: E501
+        ),
+    ) -> list[ImageContent | TextContent]:
+        """
+        Execute a Playwright web automation action.
+
+        Returns:
+            List of MCP content blocks
+        """
+        logger.info("PlaywrightTool executing action: %s", action)
+
+        try:
+            if action == "navigate":
+                if url is None:
+                    raise McpError(
+                        ErrorData(
+                            code=INVALID_PARAMS, message="url parameter is required for navigate"
+                        )
+                    )
+                result = await self.navigate(url, wait_for_load_state or "networkidle")
+
+            elif action == "screenshot":
+                result = await self.screenshot(path)
+
+            elif action == "click":
+                if selector is None:
+                    raise McpError(
+                        ErrorData(
+                            code=INVALID_PARAMS, message="selector parameter is required for click"
+                        )
+                    )
+                result = await self.click(selector)
+
+            elif action == "type":
+                if selector is None:
+                    raise McpError(
+                        ErrorData(
+                            code=INVALID_PARAMS, message="selector parameter is required for type"
+                        )
+                    )
+                if text is None:
+                    raise McpError(
+                        ErrorData(
+                            code=INVALID_PARAMS, message="text parameter is required for type"
+                        )
+                    )
+                result = await self.type_text(selector, text)
+
+            elif action == "get_page_info":
+                result = await self.get_page_info()
+
+            elif action == "wait_for_element":
+                if selector is None:
+                    raise McpError(
+                        ErrorData(
+                            code=INVALID_PARAMS,
+                            message="selector parameter is required for wait_for_element",
+                        )
+                    )
+                result = await self.wait_for_element(selector)
+
+            else:
+                raise McpError(ErrorData(code=INVALID_PARAMS, message=f"Unknown action: {action}"))
+
+            # Convert dict result to ToolResult
+            if isinstance(result, dict):
+                if result.get("success"):
+                    if "screenshot" in result:
+                        # Return screenshot as image content
+                        tool_result = ToolResult(
+                            output=result.get("message", ""), base64_image=result["screenshot"]
+                        )
+                    else:
+                        tool_result = ToolResult(output=result.get("message", ""))
+                else:
+                    tool_result = ToolResult(error=result.get("error", "Unknown error"))
+            else:
+                tool_result = result
+
+            # Convert result to content blocks
+            return tool_result_to_content_blocks(tool_result)
+
+        except McpError:
+            raise
+        except Exception as e:
+            logger.error("PlaywrightTool error: %s", e)
+            raise McpError(ErrorData(code=INVALID_PARAMS, message=f"Playwright error: {e}")) from e
+
+    async def _ensure_browser(self) -> None:
+        """Ensure browser is launched and ready."""
+        if self._browser is None or not self._browser.is_connected():
+            logger.info("Launching Playwright browser...")
+
+            # Ensure DISPLAY is set
+            os.environ["DISPLAY"] = os.environ.get("DISPLAY", ":1")
+
+            if self._playwright is None:
+                try:
+                    from playwright.async_api import async_playwright
+
+                    self._playwright = await async_playwright().start()
+                except ImportError:
+                    raise ImportError(
+                        "Playwright is not installed. Please install with: pip install playwright"
+                    ) from None
+
+            self._browser = await self._playwright.chromium.launch(
+                headless=False,
+                args=[
+                    "--no-sandbox",
+                    "--disable-dev-shm-usage",
+                    "--disable-gpu",
+                    "--disable-web-security",
+                    "--disable-features=IsolateOrigins,site-per-process",
+                    "--disable-blink-features=AutomationControlled",
+                    "--window-size=1920,1080",
+                    "--window-position=0,0",
+                    "--start-maximized",
+                    "--disable-background-timer-throttling",
+                    "--disable-backgrounding-occluded-windows",
+                    "--disable-renderer-backgrounding",
+                    "--disable-features=TranslateUI",
+                    "--disable-ipc-flooding-protection",
+                    "--disable-default-apps",
+                    "--no-first-run",
+                    "--disable-sync",
+                    "--no-default-browser-check",
+                ],
+            )
+
+            if self._browser is None:
+                raise RuntimeError("Browser failed to initialize")
+
+            self._context = await self._browser.new_context(
+                viewport={"width": 1920, "height": 1080},
+                ignore_https_errors=True,
+            )
+
+            if self._context is None:
+                raise RuntimeError("Browser context failed to initialize")
+
+            self._page = await self._context.new_page()
+            logger.info("Playwright browser launched successfully")
+
+    async def navigate(
+        self,
+        url: str,
+        wait_for_load_state: Literal[
+            "commit", "domcontentloaded", "load", "networkidle"
+        ] = "networkidle",
+    ) -> dict[str, Any]:
+        """Navigate to a URL.
+
+        Args:
+            url: URL to navigate to
+            wait_for_load_state: Load state to wait for (load, domcontentloaded, networkidle)
+
+        Returns:
+            Dict with navigation result
+        """
+        await self._ensure_browser()
+
+        logger.info("Navigating to %s", url)
+        try:
+            await self.page.goto(url, wait_until=wait_for_load_state)
+            current_url = self.page.url
+            title = await self.page.title()
+
+            return {
+                "success": True,
+                "url": current_url,
+                "title": title,
+                "message": f"Successfully navigated to {url}",
+            }
+        except Exception as e:
+            logger.error("Navigation failed: %s", e)
+            return {
+                "success": False,
+                "error": str(e),
+                "message": f"Failed to navigate to {url}: {e}",
+            }
+
+    async def screenshot(self, path: str | None = None) -> dict[str, Any]:
+        """Take a screenshot of the current page.
+
+        Args:
+            path: Optional path to save screenshot
+
+        Returns:
+            Dict with screenshot result
+        """
+        await self._ensure_browser()
+
+        try:
+            if path:
+                await self.page.screenshot(path=path, full_page=True)
+                return {"success": True, "path": path, "message": f"Screenshot saved to {path}"}
+            else:
+                # Return base64 encoded screenshot
+                screenshot_bytes = await self.page.screenshot(full_page=True)
+                import base64
+
+                screenshot_b64 = base64.b64encode(screenshot_bytes).decode()
+                return {
+                    "success": True,
+                    "screenshot": screenshot_b64,
+                    "message": "Screenshot captured",
+                }
+        except Exception as e:
+            logger.error("Screenshot failed: %s", e)
+            return {"success": False, "error": str(e), "message": f"Failed to take screenshot: {e}"}
+
+    async def click(self, selector: str) -> dict[str, Any]:
+        """Click an element by selector.
+
+        Args:
+            selector: CSS selector for element to click
+
+        Returns:
+            Dict with click result
+        """
+        await self._ensure_browser()
+
+        try:
+            await self.page.click(selector)
+            return {"success": True, "message": f"Clicked element: {selector}"}
+        except Exception as e:
+            logger.error("Click failed: %s", e)
+            return {
+                "success": False,
+                "error": str(e),
+                "message": f"Failed to click {selector}: {e}",
+            }
+
+    async def type_text(self, selector: str, text: str) -> dict[str, Any]:
+        """Type text into an element.
+
+        Args:
+            selector: CSS selector for input element
+            text: Text to type
+
+        Returns:
+            Dict with type result
+        """
+        await self._ensure_browser()
+
+        try:
+            await self.page.fill(selector, text)
+            return {"success": True, "message": f"Typed '{text}' into {selector}"}
+        except Exception as e:
+            logger.error("Type failed: %s", e)
+            return {
+                "success": False,
+                "error": str(e),
+                "message": f"Failed to type into {selector}: {e}",
+            }
+
+    async def get_page_info(self) -> dict[str, Any]:
+        """Get current page information.
+
+        Returns:
+            Dict with page info
+        """
+        await self._ensure_browser()
+
+        try:
+            url = self.page.url
+            title = await self.page.title()
+            return {
+                "success": True,
+                "url": url,
+                "title": title,
+                "message": f"Current page: {title} ({url})",
+            }
+        except Exception as e:
+            logger.error("Get page info failed: %s", e)
+            return {"success": False, "error": str(e), "message": f"Failed to get page info: {e}"}
+
+    async def wait_for_element(self, selector: str) -> dict[str, Any]:
+        """Wait for an element to appear.
+
+        Args:
+            selector: CSS selector for element
+
+        Returns:
+            Dict with wait result
+        """
+        await self._ensure_browser()
+
+        try:
+            await self.page.wait_for_selector(selector, timeout=30000)
+            return {"success": True, "message": f"Element {selector} appeared"}
+        except Exception as e:
+            logger.error("Wait for element failed: %s", e)
+            return {
+                "success": False,
+                "error": str(e),
+                "message": f"Element {selector} did not appear within 30000ms: {e}",
+            }
+
+    async def close(self) -> None:
+        """Close browser and cleanup."""
+        if self._browser:
+            try:
+                await self._browser.close()
+                logger.info("Browser closed")
+            except Exception as e:
+                logger.error("Error closing browser: %s", e)
+
+        if self._playwright:
+            try:
+                await self._playwright.stop()
+            except Exception as e:
+                logger.error("Error stopping playwright: %s", e)
+
+        self._browser = None
+        self._context = None
+        self._page = None
+        self._playwright = None

hud/tools/tests/__init__.py

@@ -0,0 +1,3 @@
+from __future__ import annotations
+
+__all__ = []

hud/tools/tests/test_bash.py

@@ -0,0 +1,152 @@
+"""Tests for bash tool."""
+
+from __future__ import annotations
+
+from unittest.mock import AsyncMock, MagicMock, patch
+
+import pytest
+
+from hud.tools.base import ToolResult
+from hud.tools.bash import BashTool, ToolError, _BashSession
+
+
+class TestBashSession:
+    """Tests for _BashSession."""
+
+    @pytest.mark.asyncio
+    async def test_session_start(self):
+        """Test starting a bash session."""
+        session = _BashSession()
+        assert session._started is False
+
+        with patch("asyncio.create_subprocess_shell") as mock_create:
+            mock_process = MagicMock()
+            mock_create.return_value = mock_process
+
+            await session.start()
+
+            assert session._started is True
+            assert session._process == mock_process
+            mock_create.assert_called_once()
+
+    def test_session_stop_not_started(self):
+        """Test stopping a session that hasn't started."""
+        session = _BashSession()
+
+        with pytest.raises(ToolError) as exc_info:
+            session.stop()
+
+        assert "Session has not started" in str(exc_info.value)
+
+    @pytest.mark.asyncio
+    async def test_session_run_not_started(self):
+        """Test running command on a session that hasn't started."""
+        session = _BashSession()
+
+        with pytest.raises(ToolError) as exc_info:
+            await session.run("echo test")
+
+        assert "Session has not started" in str(exc_info.value)
+
+    @pytest.mark.asyncio
+    async def test_session_run_success(self):
+        """Test successful command execution."""
+        session = _BashSession()
+        session._started = True
+
+        # Mock process
+        mock_process = MagicMock()
+        mock_process.returncode = None
+        mock_process.stdin = MagicMock()
+        mock_process.stdin.write = MagicMock()
+        mock_process.stdin.drain = AsyncMock()
+        mock_process.stdout = MagicMock()
+        mock_process.stdout.readuntil = AsyncMock(return_value=b"Hello World\n<<exit>>\n")
+        mock_process.stderr = MagicMock()
+        mock_process.stderr.read = AsyncMock(return_value=b"")
+
+        session._process = mock_process
+
+        result = await session.run("echo Hello World")
+
+        assert result.output == "Hello World\n"
+        assert result.error == ""
+
+
+class TestBashTool:
+    """Tests for BashTool."""
+
+    def test_bash_tool_init(self):
+        """Test BashTool initialization."""
+        tool = BashTool()
+        assert tool._session is None
+
+    @pytest.mark.asyncio
+    async def test_call_with_command(self):
+        """Test calling tool with a command."""
+        tool = BashTool()
+
+        # Mock session
+        mock_session = MagicMock()
+        mock_session.run = AsyncMock(return_value=ToolResult(output="test output"))
+
+        # Mock _BashSession creation
+        with patch("hud.tools.bash._BashSession") as mock_session_class:
+            mock_session_class.return_value = mock_session
+            mock_session.start = AsyncMock()
+
+            result = await tool(command="echo test")
+
+            assert isinstance(result, ToolResult)
+            assert result.output == "test output"
+            mock_session.start.assert_called_once()
+            mock_session.run.assert_called_once_with("echo test")
+
+    @pytest.mark.asyncio
+    async def test_call_restart(self):
+        """Test restarting the tool."""
+        tool = BashTool()
+
+        # Set up existing session
+        old_session = MagicMock()
+        old_session.stop = MagicMock()
+        tool._session = old_session
+
+        # Mock new session
+        new_session = MagicMock()
+        new_session.start = AsyncMock()
+
+        with patch("hud.tools.bash._BashSession", return_value=new_session):
+            result = await tool(restart=True)
+
+            assert isinstance(result, ToolResult)
+            assert result.system == "tool has been restarted."
+            old_session.stop.assert_called_once()
+            new_session.start.assert_called_once()
+            assert tool._session == new_session
+
+    @pytest.mark.asyncio
+    async def test_call_no_command_error(self):
+        """Test calling without command raises error."""
+        tool = BashTool()
+
+        with pytest.raises(ToolError) as exc_info:
+            await tool()
+
+        assert "no command provided" in str(exc_info.value)
+
+    @pytest.mark.asyncio
+    async def test_call_with_existing_session(self):
+        """Test calling with an existing session."""
+        tool = BashTool()
+
+        # Set up existing session
+        existing_session = MagicMock()
+        existing_session.run = AsyncMock(return_value=ToolResult(output="result"))
+        tool._session = existing_session
+
+        result = await tool(command="ls")
+
+        assert isinstance(result, ToolResult)
+        assert result.output == "result"
+        existing_session.run.assert_called_once_with("ls")