win_monitor/backend/app/automation/nodes/browser_control.py

from __future__ import annotations

import time
from typing import Any

from ... import windows_automation
from ..context import WorkflowContext
from ..registry import control_ports, field_def, register_node


def _number(value: Any, default: float = 0) -> float:
    try:
        return float(value)
    except (TypeError, ValueError):
        return default


def _boolean(value: Any, default: bool = False) -> bool:
    if value in (None, ""):
        return default
    if isinstance(value, str):
        return value.strip().lower() in {"1", "true", "yes", "y", "on"}
    return bool(value)


def ensure_browser_node(node: dict[str, Any], inputs: dict[str, Any], context: WorkflowContext) -> dict[str, Any]:
    params = node.get("params", {})
    browser = str(inputs.get("browser", params.get("browser")) or "edge")
    url = str(inputs.get("url", params.get("url")) or "").strip()
    maximize = _boolean(inputs.get("maximize", params.get("maximize")), True)
    wait_seconds = max(0, min(_number(inputs.get("wait_seconds", params.get("wait_seconds")), 2), 30))

    if url:
        opened = windows_automation.open_url(url=url, browser=browser, new_window=bool(params.get("new_window", True)))
    elif browser.lower() in {"edge", "msedge"}:
        opened = windows_automation.start_program("cmd.exe /c start msedge", shell=True)
    else:
        opened = windows_automation.start_program("cmd.exe /c start chrome", shell=True)
    context.remember_pid(opened.get("pid"))
    if wait_seconds:
        time.sleep(wait_seconds)
    if maximize:
        windows_automation.keyboard_action("hotkey", keys=["win", "up"])
    return {"action": "ensure_browser", "browser": browser, "url": url, "opened": opened, "maximized": maximize}


def browser_key_node(node: dict[str, Any], inputs: dict[str, Any], context: WorkflowContext) -> dict[str, Any]:
    params = node.get("params", {})
    action = str(inputs.get("action", params.get("action")) or "fullscreen")
    if action == "fullscreen":
        result = windows_automation.keyboard_action("press", key="f11")
    elif action == "back":
        result = windows_automation.keyboard_action("hotkey", keys=["alt", "left"])
    elif action == "forward":
        result = windows_automation.keyboard_action("hotkey", keys=["alt", "right"])
    elif action == "refresh":
        result = windows_automation.keyboard_action("press", key="f5")
    elif action == "escape":
        result = windows_automation.keyboard_action("press", key="escape")
    elif action == "close_tab":
        result = windows_automation.keyboard_action("hotkey", keys=["ctrl", "w"])
    elif action == "close_window":
        result = windows_automation.keyboard_action("hotkey", keys=["alt", "f4"])
    else:
        result = {"action": "browser_noop", "key": None}
    return {"action": action, "keyboard": result}


register_node(
    {
        "type": "browser.ensure_foreground",
        "category": "browser",
        "label": "确保浏览器前台",
        "description": "打开或唤起浏览器，可选打开指定网址，并尝试最大化窗口。",
        "params": {
            "url": field_def("text", "网址", "", description="为空时仅尝试启动浏览器。"),
            "browser": field_def("select", "浏览器", "edge", options=["default", "edge"]),
            "new_window": field_def("boolean", "新窗口", True),
            "maximize": field_def("boolean", "最大化", True),
            "wait_seconds": field_def("number", "等待秒数", 2, minimum=0, maximum=30),
        },
        "inputs": {
            "url": field_def("string", "网址"),
            "browser": field_def("string", "浏览器"),
            "maximize": field_def("boolean", "最大化"),
            "wait_seconds": field_def("number", "等待秒数"),
        },
        "outputs": {
            "action": {"type": "string", "label": "动作"},
            "browser": {"type": "string", "label": "浏览器"},
            "url": {"type": "string", "label": "网址"},
            "opened": {"type": "object", "label": "启动结果"},
            "maximized": {"type": "boolean", "label": "是否最大化"},
        },
        "control_ports": control_ports(),
    },
    ensure_browser_node,
)

register_node(
    {
        "type": "browser.control",
        "category": "browser",
        "label": "浏览器快捷控制",
        "description": "对当前浏览器窗口执行全屏、返回、刷新、关闭标签页等快捷操作。",
        "params": {
            "action": field_def(
                "select",
                "动作",
                "fullscreen",
                options=["fullscreen", "back", "forward", "refresh", "escape", "close_tab", "close_window"],
            ),
        },
        "inputs": {"action": field_def("string", "动作")},
        "outputs": {
            "action": {"type": "string", "label": "动作"},
            "keyboard": {"type": "object", "label": "键盘结果"},
        },
        "control_ports": control_ports(),
    },
    browser_key_node,
)