Browser use refactoring

.gitignore

@@ -168,3 +168,10 @@ benchmark/
 
 # node modules
 node_modules/
+
+tmp/
+
+# Key files
+*.pem
+*.key
+*.gpg

patchwork/app.py

@@ -144,20 +144,22 @@ def sigint_handler(signum, frame):
 @click.option("patched_api_key", "--patched_api_key", help="API key to use with the patched.codes service.")
 @click.option("disable_telemetry", "--disable_telemetry", is_flag=True, help="Disable telemetry.", default=False)
 @click.option("debug", "--debug", is_flag=True, help="Enable debug mode.", default=False)
+@click.option("plain", "--plain", is_flag=True, help="Enable plain mode (no panel).", default=False)
 def cli(
     log: str,
     patchflow: str,
     opts: list[str],
     config: str | None,
     output: str | None,
+    plain: bool,
     data_format: str,
     patched_api_key: str | None,
     disable_telemetry: bool,
     debug: bool,
 ):
     setup_cli()
-
-    init_cli_logger(log)
+    is_plain_console = plain or debug
+    init_cli_logger(log, is_plain_console)
 
     if "::" in patchflow:
         module_path, _, patchflow_name = patchflow.partition("::")
@@ -167,7 +169,7 @@ def cli(
 
     possbile_module_paths = deque((module_path,))
 
-    panel = logger.panel("Initializing Patchwork CLI") if debug else nullcontext()
+    panel = nullcontext() if is_plain_console else logger.panel("Initializing Patchwork CLI")
 
     with panel:
         inputs = {}
@@ -227,7 +229,7 @@ def cli(
             # treat --key=value as a key-value pair
             inputs[key] = value
 
-    patchflow_panel = nullcontext() if debug else logger.panel(f"Patchflow {patchflow} inputs")
+    patchflow_panel = nullcontext() if is_plain_console else logger.panel(f"Patchflow {patchflow} inputs")
 
     with patchflow_panel as _:
         if debug is True:

patchwork/common/utils/browser_initializer.py

@@ -0,0 +1,167 @@
+import logging
+import os
+from typing import List, Optional
+
+from browser_use import Browser, BrowserConfig, BrowserContextConfig, Controller
+from browser_use.agent.views import ActionResult
+from browser_use.browser.context import BrowserContext
+
+logger = logging.getLogger(__name__)
+
+
+async def set_file_input(index: int, paths: str | List[str], browser: BrowserContext):
+    """
+    Set the file input value to the given path or list of paths.
+
+    Args:
+        index: The DOM element index to target
+        paths: Local file path or list of local file paths to upload
+        browser: Browser context for interaction
+
+    Returns:
+        ActionResult: Result of the upload operation
+    """
+    if isinstance(paths, str):
+        paths = [paths]
+
+    for path in paths:
+        if not os.path.exists(path):
+            return ActionResult(error=f"File {path} does not exist")
+
+    dom_el = await browser.get_dom_element_by_index(index)
+    file_upload_dom_el = dom_el.get_file_upload_element()
+
+    if file_upload_dom_el is None:
+        msg = f"No file upload element found at index {index}. The element may be hidden or not an input type file"
+        logger.info(msg)
+        return ActionResult(error=msg)
+
+    file_upload_el = await browser.get_locate_element(file_upload_dom_el)
+
+    if file_upload_el is None:
+        msg = f"No file upload element found at index {index}. The element may be hidden or not an input type file"
+        logger.info(msg)
+        return ActionResult(error=msg)
+
+    try:
+        await file_upload_el.set_input_files(paths)
+        msg = f"Successfully set file input value to {paths}"
+        logger.info(msg)
+        return ActionResult(extracted_content=msg, include_in_memory=True)
+    except Exception as e:
+        msg = f"Failed to upload file to index {index}: {str(e)}"
+        logger.info(msg)
+        return ActionResult(error=msg)
+
+
+async def close_current_tab(browser: BrowserContext):
+    await browser.close_current_tab()
+    msg = "🔄  Closed current tab"
+    logger.info(msg)
+    return ActionResult(extracted_content=msg, include_in_memory=True)
+
+
+class BrowserInitializer:
+    """
+    Initialize and cache browser and controller instances.
+
+    This class uses a singleton pattern to ensure we only create one browser
+    instance throughout the application lifecycle, which saves resources.
+    """
+
+    _browser = None
+    _controller = None
+    _browser_context = None
+
+    @classmethod
+    def init_browser(cls, config=BrowserConfig()):
+        """
+        Initialize and cache the Browser instance.
+
+        Returns:
+            Browser: Browser instance for web automation
+        """
+        if cls._browser is not None:
+            return cls._browser
+
+        cls._browser = Browser(config=config)
+        return cls._browser
+
+    @classmethod
+    def init_browser_context(cls, config: Optional[BrowserConfig], downloads_path: Optional[str] = None):
+        """
+        Initialize and cache the BrowserContext instance.
+
+        Returns:
+            BrowserContext: BrowserContext instance for managing browser context
+        """
+        if cls._browser_context is not None:
+            return cls._browser_context
+
+        if downloads_path and not os.path.exists(downloads_path):
+            os.makedirs(downloads_path)
+
+        context_config = BrowserContextConfig(
+            # cookies_file=cookies_file,
+            browser_window_size={"width": 1920, "height": 1080},
+        )
+        browser = cls.init_browser(config=config)
+
+        class BrowserContextWithDownloadHandling(BrowserContext):
+            async def handle_download(self, download):
+                suggested_filename = download.suggested_filename
+                unique_filename = await self._get_unique_filename(downloads_path, suggested_filename)
+                download_path = os.path.join(downloads_path, unique_filename)
+                await download.save_as(download_path)
+                logger.info(f"Downloaded file saved to {download_path}")
+
+            async def _initialize_session(self):
+                async def _download_listener(download):
+                    logger.info("[BUD] Download event triggered")
+                    await self.handle_download(download)
+                    return download
+
+                def _new_page_listener(page):
+                    logger.info("[BUD] Adding download event listener to page")
+                    page.on("download", _download_listener)
+                    return page
+
+                await super()._initialize_session()
+
+                logger.info("[BUD] Adding page event listener to context")
+                self.session.context.on("page", _new_page_listener)
+
+                logger.info(f"[BUD] Adding download event listener to {len(self.session.context.pages)} existing pages")
+                for page in self.session.context.pages:
+                    page.on("download", _download_listener)
+
+        cls._browser_context = (
+            BrowserContextWithDownloadHandling(browser=browser, config=context_config)
+            if downloads_path
+            else BrowserContext(browser=browser, config=context_config)
+        )
+        return cls._browser_context
+
+    @classmethod
+    def init_controller(cls):
+        """
+        Initialize and cache the Controller instance.
+
+        Returns:
+            Controller: Controller instance for managing browser actions
+        """
+        if cls._controller is not None:
+            return cls._controller
+
+        controller = Controller()
+
+        controller.action(
+            "Set the value of a file input to the given path or list of paths",
+        )(set_file_input)
+
+        controller.action(
+            description="Close the tab that is currently active",
+        )(close_current_tab)
+
+        cls._controller = controller
+        return cls._controller

patchwork/logger.py

@@ -18,7 +18,7 @@
 from patchwork.managed_files import HOME_FOLDER, LOG_FILE
 
 # Create a global console object
-console = Console()
+console = Console(force_terminal=True, no_color=False)
 
 # Add TRACE level to logging
 logging.TRACE = logging.DEBUG - 1
@@ -39,10 +39,11 @@ def evict_null_handler():
 
 
 class TerminalHandler(RichHandler):
-    def __init__(self, log_level: str):
+    def __init__(self, log_level: str, plain: bool):
+        self.plain = plain
         super().__init__(
             console=console,
-            rich_tracebacks=True,
+            rich_tracebacks=not plain,
             tracebacks_suppress=[click],
             show_time=False,
             show_path=False,
@@ -95,22 +96,25 @@ def deregister_progress_bar(self):
     @contextlib.contextmanager
     def panel(self, title: str):
         global console
-        self.__panel_lines = []
-        self.__panel_title = title
-        self.__panel = Panel("", title=title)
-        renderables = [self.__panel]
-        if self.__progress_bar is not None:
-            renderables.append(self.__progress_bar)
-
-        self.__live = Live(Group(*renderables), console=console, vertical_overflow="visible")
-        try:
-            self.__live.start()
+        if self.plain:
             yield
-        except Exception as e:
-            raise e
-        finally:
-            self.__reset_live()
-            self.console.print("\n")
+        else:
+            self.__panel_lines = []
+            self.__panel_title = title
+            self.__panel = Panel("", title=title)
+            renderables = [self.__panel]
+            if self.__progress_bar is not None:
+                renderables.append(self.__progress_bar)
+
+            self.__live = Live(Group(*renderables), console=console, vertical_overflow="visible")
+            try:
+                self.__live.start()
+                yield
+            except Exception as e:
+                raise e
+            finally:
+                self.__reset_live()
+                self.console.print("\n")
 
     def emit(self, record: logging.LogRecord) -> None:
         markup = getattr(record, "markup", None)
@@ -126,7 +130,8 @@ def emit(self, record: logging.LogRecord) -> None:
         if self.__panel is not None:
             self.__emit_panel(record)
         else:
-            setattr(record, "markup", True)
+            if not self.plain:
+                setattr(record, "markup", True)
             super().emit(record)
 
     def __emit_panel(self, record: logging.LogRecord) -> None:
@@ -143,7 +148,7 @@ def inner(record: logging.LogRecord) -> bool:
         return inner
 
 
-def init_cli_logger(log_level: str) -> logging.Logger:
+def init_cli_logger(log_level: str, plain: bool) -> logging.Logger:
     global logger
 
     evict_null_handler()
@@ -158,7 +163,7 @@ def init_cli_logger(log_level: str) -> logging.Logger:
     except FileNotFoundError:
         logger.error(f"Unable to create log file: {LOG_FILE}")
 
-    th = TerminalHandler(log_level.upper())
+    th = TerminalHandler(log_level.upper(), plain)
     logger.addHandler(th)
     setattr(logger, "panel", th.panel)
     setattr(logger, "register_progress_bar", th.register_progress_bar)