cli.py

"""
命令行交互界面
支持人机协作：AI可以随时暂停询问用户
"""
import asyncio
import sys
import json
from typing import Optional

from prompt_toolkit import PromptSession
from prompt_toolkit.history import FileHistory
from rich.console import Console
from rich.markdown import Markdown
from rich.panel import Panel
from rich.text import Text

from config import Config, APP_NAME, VERSION, SYSTEM_PROMPT
from llm_client import LLMClient
from browser_agent import SmartBrowserAgent


class BrowserAI_CLI:
    """BrowserAI命令行界面 - 支持人机协作"""
    
    def __init__(self):
        self.console = Console()
        self.llm = LLMClient()
        self.agent: Optional[SmartBrowserAgent] = None
        self.session = None
        self.smart_mode = True
        self.interactive_mode = True  # 是否启用人机协作
        self._setup_prompt()
    
    def _setup_prompt(self):
        """设置交互式提示符"""
        try:
            history = FileHistory('.browserai_history')
            self.session = PromptSession(history=history)
        except:
            self.session = PromptSession()
    
    async def ask_user_callback(self, question: str, context: str = "") -> str:
        """AI询问用户时的回调函数"""
        # 显示问题
        panel = Panel(
            f"{question}\n\n[dim]{context}[/]",
            title="🤖 AI 需要你的帮助",
            border_style="yellow",
            padding=(1, 2)
        )
        self.console.print(panel)
        
        # 等待用户输入
        try:
            if self.session:
                answer = await self.session.prompt_async("💬 你的回答: ", multiline=False)
            else:
                answer = input("💬 你的回答: ")
            return answer
        except:
            return "继续"
    
    def print_banner(self):
        """打印启动横幅"""
        cfg = Config()
        browser_mode = "Edge/Chrome" if cfg.BROWSER.use_local_browser else "内置浏览器"
        
        banner = f"""
╭─────────────────────────────────────────────╮
│                                             │
│   🤖 {APP_NAME} v{VERSION}                    │
│   智能浏览器助手 · 自动决策 · 人机协作       │
│                                             │
│   浏览器模式: {browser_mode:<20}                 │
│   人机协作: {'开启' if self.interactive_mode else '关闭':<20}                 │
│                                             │
│   输入 /help 查看命令                       │
│   输入 /quit 退出                          │
│                                             │
╰─────────────────────────────────────────────╯
"""
        self.console.print(banner, style="cyan")
    
    def print_help(self):
        """打印帮助信息"""
        help_text = """
## 可用命令

| 命令 | 说明 |
|------|------|
| `/help` | 显示此帮助信息 |
| `/quit` | 退出程序 |
| `/clear` | 清除对话历史 |
| `/interactive` | 切换人机协作模式 |
| `/smart` | 切换智能模式 |
| `/mode` | 查看当前模式 |
| `/browser` | 查看浏览器配置 |

## 核心特性

🧠 **人机协作模式**（智能询问）
AI在以下情况会暂停询问你：
- 需要登录账号（"请登录后继续"）
- 需要验证码（"请输入验证码"）
- 需要选择（"要点哪个？"）
- 需要确认（"确认提交订单吗？"）

🌐 **本地浏览器支持**
默认使用你电脑上的Edge/Chrome浏览器，可以使用你的：
- 已登录的账号（Cookie）
- 保存的密码
- 浏览器插件

## 使用示例

```
> 打开百度
> 去知乎搜索"AI Agent"，找到相关内容
> 访问GitHub trending
> 截图保存当前页面
> 帮我查看某宝购物车的价格
```

## 配置选项 (.env 文件)

```
USE_LOCAL_BROWSER=true          # 使用本地浏览器
LOCAL_BROWSER_PATH=            # 手动指定路径
HEADLESS=false                 # 是否无头模式
```
"""
        self.console.print(Markdown(help_text))
    
    async def init_browser(self):
        """初始化浏览器"""
        if not self.agent:
            self.agent = SmartBrowserAgent()
            # 设置回调函数
            self.agent.set_ask_callback(self.ask_user_callback)
            await self.agent.start()
    
    async def close_browser(self):
        """关闭浏览器"""
        if self.agent:
            await self.agent.close()
            self.agent = None
    
    async def handle_command(self, user_input: str) -> Optional[bool]:
        """处理命令"""
        cmd = user_input.lower().strip()
        
        if cmd in ["/quit", "/exit", "/q"]:
            return False
        
        elif cmd == "/help":
            self.print_help()
        
        elif cmd == "/clear":
            self.llm.clear_history()
            self.console.print("✅ 对话历史已清除", style="green")
        
        elif cmd == "/interactive":
            self.interactive_mode = not self.interactive_mode
            if self.agent:
                self.agent.interactive_mode = self.interactive_mode
            status = "开启" if self.interactive_mode else "关闭"
            self.console.print(f"人机协作模式已{status}", style="green")
        
        elif cmd == "/smart":
            self.smart_mode = not self.smart_mode
            status = "开启" if self.smart_mode else "关闭"
            self.console.print(f"智能模式已{status}", style="green")
        
        elif cmd == "/mode":
            cfg = Config()
            self.console.print(f"智能决策: {'开启' if self.smart_mode else '关闭'}", style="blue")
            self.console.print(f"人机协作: {'开启' if self.interactive_mode else '关闭'}", style="blue")
            self.console.print(f"本地浏览器: {'Edge/Chrome' if cfg.BROWSER.use_local_browser else '内置'}", style="blue")
        
        elif cmd == "/browser":
            cfg = Config()
            self.console.print(f"使用本地浏览器: {cfg.BROWSER.use_local_browser}", style="blue")
            if cfg.BROWSER.local_browser_path:
                self.console.print(f"指定路径: {cfg.BROWSER.local_browser_path}", style="blue")
            self.console.print(f"无头模式: {cfg.BROWSER.headless}", style="blue")
        
        else:
            return None
        
        return True
    
    def is_browser_task(self, user_input: str) -> bool:
        """判断是否为浏览器任务"""
        browser_keywords = [
            "打开", "访问", "搜索", "查找", "浏览", "截图",
            "点击", "输入", "填写", "登录", "注册", "查看", "看看", "瞧瞧",
            "去", "帮我", "查", "找", "搜", "浏览", "翻",
            "baidu", "google", "bing", "github", "知乎", "bilibili", "淘宝",
            "twitter", "推特", "微博", "facebook", "instagram", "youtube",
            "https:", "http:", "www.", ".com", ".cn", ".org",
            "网页", "网站", "页面", "url", "链接", "浏览器",
            "账号", "社交媒体", "信息", "资料", "主页", "博客"
        ]
        text_lower = user_input.lower()
        return any(kw in text_lower or kw in user_input for kw in browser_keywords)
    
    async def process_message(self, user_input: str):
        """处理用户消息"""
        needs_browser = self.is_browser_task(user_input)
        
        if needs_browser:
            await self.init_browser()
            
            if self.smart_mode:
                print("\n🤖 正在执行任务，AI会自动决策...")
                if self.interactive_mode:
                    print("（如有疑问，AI会暂停询问你）\n")
                
                result = await self.agent.run_task(user_input, self.llm)
                self.console.print(f"\n{result}")
            else:
                result = await self._execute_simple_task(user_input)
                self.console.print(f"\n{result}")
        else:
            response = self.llm.chat(user_input)
            self.console.print(f"\n🤖 {response}")
    
    async def _execute_simple_task(self, user_input: str) -> str:
        """简单模式执行任务"""
        if "搜索" in user_input:
            query = user_input.split("搜索")[-1].strip()
            if query:
                engine = "bing"
                if "百度" in user_input:
                    engine = "baidu"
                elif "google" in user_input.lower():
                    engine = "google"
                
                await self.agent.execute_action({
                    "action": "search",
                    "query": query,
                    "engine": engine
                })
                return f"✅ 已在{engine}搜索: {query}"
        
        elif "打开" in user_input or "访问" in user_input:
            url = user_input.replace("打开", "").replace("访问", "").strip()
            await self.agent.execute_action({"action": "navigate", "url": url})
            return f"✅ 已打开: {url}"
        
        elif "截图" in user_input:
            await self.agent.execute_action({"action": "screenshot", "path": "screenshot.png"})
            return "✅ 已保存截图到 screenshot.png"
        
        info = await self.agent.get_page_info()
        return f"当前页面: {info.get('title', 'Unknown')}"
    
    async def run(self):
        """主循环"""
        if not Config.validate():
            self.console.print("❌ 配置验证失败", style="red")
            return
        
        self.llm.add_system_message()
        self.print_banner()
        
        try:
            while True:
                try:
                    if self.session:
                        user_input = await self.session.prompt_async("> ", multiline=False)
                    else:
                        user_input = input("> ")
                    
                    if not user_input.strip():
                        continue
                    
                    cmd_result = await self.handle_command(user_input)
                    
                    if cmd_result is False:
                        break
                    elif cmd_result is True:
                        continue
                    
                    await self.process_message(user_input)
                    
                except KeyboardInterrupt:
                    print("\n")
                    continue
                except EOFError:
                    break
        
        finally:
            await self.close_browser()
            self.console.print(f"\n👋 感谢使用 {APP_NAME}！", style="cyan")


def main():
    """入口函数"""
    cli = BrowserAI_CLI()
    asyncio.run(cli.run())


if __name__ == "__main__":
    main()