diff --git a/.gitignore b/.gitignore deleted file mode 100644 index 225a2a7..0000000 --- a/.gitignore +++ /dev/null @@ -1,7 +0,0 @@ -.idea -.vscode -__pycache__ -test - -*.log -.env* \ No newline at end of file diff --git a/.idea/shelf/___.xml b/.idea/shelf/___.xml new file mode 100644 index 0000000..51670f5 --- /dev/null +++ b/.idea/shelf/___.xml @@ -0,0 +1,4 @@ + + \ No newline at end of file diff --git "a/.idea/shelf/\345\234\250\345\217\230\345\237\272\344\271\213\345\211\215\346\234\252\346\217\220\344\272\244\347\232\204\346\233\264\346\224\271_[\346\233\264\346\224\271]/shelved.patch" "b/.idea/shelf/\345\234\250\345\217\230\345\237\272\344\271\213\345\211\215\346\234\252\346\217\220\344\272\244\347\232\204\346\233\264\346\224\271_[\346\233\264\346\224\271]/shelved.patch" new file mode 100644 index 0000000..aa80353 --- /dev/null +++ "b/.idea/shelf/\345\234\250\345\217\230\345\237\272\344\271\213\345\211\215\346\234\252\346\217\220\344\272\244\347\232\204\346\233\264\346\224\271_[\346\233\264\346\224\271]/shelved.patch" @@ -0,0 +1,1902 @@ +Index: core/computer/screen/__init__.py +=================================================================== +diff --git a/core/computer/screen/__init__.py b/core/computer/screen/__init__.py +deleted file mode 100644 +--- a/core/computer/screen/__init__.py (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,6 +0,0 @@ +-from .screen import Screen +- +- +-__all__ = [Screen] +- +-__version__ = '1.0.0' +\ No newline at end of file +Index: core/computer/keyboard/__init__.py +=================================================================== +diff --git a/core/computer/keyboard/__init__.py b/core/computer/keyboard/__init__.py +deleted file mode 100644 +--- a/core/computer/keyboard/__init__.py (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,47 +0,0 @@ +-""" +-键盘操作包 +-""" +-from .keyboard import ( +- Keyboard, +- type_text, +- press, +- hotkey, +- key_down, +- key_up, +- hold_key, +- copy, +- paste, +- cut, +- select_all, +- undo, +- redo, +- save, +- find, +- new_tab, +- close_tab, +- refresh, +- switch_window +-) +- +-__all__ = [ +- # Keyboard类和函数 +- 'Keyboard', +- 'type_text', +- 'press', +- 'hotkey', +- 'key_down', +- 'key_up', +- 'hold_key', +- 'copy', +- 'paste', +- 'cut', +- 'select_all', +- 'undo', +- 'redo', +- 'save', +- 'find', +- 'new_tab', +- 'close_tab', +- 'refresh', +- 'switch_window', +-] +Index: core/computer/code/languages/powershell.py +=================================================================== +diff --git a/core/computer/code/languages/powershell.py b/core/computer/code/languages/powershell.py +deleted file mode 100644 +--- a/core/computer/code/languages/powershell.py (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,64 +0,0 @@ +-import queue +-import shutil +-import subprocess +-import threading +-import time +- +-from ..base_language import BaseLanguage +- +- +-class PowerShellLanguage(BaseLanguage): +- def __init__(self): +- super().__init__() +- self.process = None +- +- def is_available(self): +- return shutil.which("powershell") is not None or shutil.which("pwsh") is not None +- +- def run(self, code: str): +- message = queue.Queue() +- execution_thread = threading.Thread(target=self._execute, args=(code, message)) +- execution_thread.daemon = True +- execution_thread.start() +- return message +- +- def _execute(self, code: str, message: queue.Queue): +- self.is_running = True +- self.start_time = time.time() +- self.should_stop = False +- +- executable = "pwsh" if shutil.which("pwsh") else "powershell" +- +- try: +- # -Command - tells powershell to read command from stdin +- self.process = subprocess.Popen( +- [executable, "-NoProfile", "-Command", code], +- stdout=subprocess.PIPE, +- stderr=subprocess.STDOUT, +- text=True, +- bufsize=1, +- universal_newlines=True +- ) +- +- for line in self.process.stdout: +- if self.should_stop: +- break +- message.put({"type": "text", "content": line}) +- +- return_code = self.process.wait() +- message.put({"type": "text", "content": f"Return code: {return_code}"}) +- +- except Exception as e: +- message.put({"type": "error", "content": f"[PowerShellLanguage]Error: {e}"}) +- finally: +- self.is_running = False +- self.process = None +- +- def interrupt(self): +- super().interrupt() +- if self.process: +- self.process.terminate() +- try: +- self.process.wait(timeout=2) +- except subprocess.TimeoutExpired: +- self.process.kill() +Index: core/computer/mouse/mouse.py +=================================================================== +diff --git a/core/computer/mouse/mouse.py b/core/computer/mouse/mouse.py +deleted file mode 100644 +--- a/core/computer/mouse/mouse.py (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,368 +0,0 @@ +-import pyautogui +-import time +-from typing import Tuple, Optional, Union +- +- +-class Mouse: +- """鼠标操作类，提供各种鼠标控制功能""" +- +- def __init__(self): +- """初始化鼠标模块""" +- # 设置PyAutoGUI的安全特性 +- pyautogui.FAILSAFE = True # 移动鼠标到屏幕左上角会触发异常 +- pyautogui.PAUSE = 0.1 # 每次操作后暂停0.1秒 +- +- def click(self, x: Optional[int] = None, y: Optional[int] = None, +- clicks: int = 1, interval: float = 0.0, +- button: str = 'left', duration: float = 0.0) -> dict: +- """ +- 鼠标点击操作 +- +- Args: +- x: X坐标，None表示当前位置 +- y: Y坐标，None表示当前位置 +- clicks: 点击次数，默认1次 +- interval: 多次点击之间的间隔（秒） +- button: 鼠标按键 'left'/'right'/'middle' +- duration: 移动到目标位置的持续时间（秒） +- +- Returns: +- 包含操作结果的字典 +- +- Examples: +- >>> mouse.click(100, 200) # 点击坐标(100, 200) +- >>> mouse.click() # 点击当前位置 +- >>> mouse.click(button='right') # 右键点击当前位置 +- """ +- try: +- if x is not None and y is not None: +- pyautogui.click(x, y, clicks=clicks, interval=interval, +- button=button, duration=duration) +- else: +- pyautogui.click(clicks=clicks, interval=interval, button=button) +- +- current_pos = pyautogui.position() +- return { +- 'success': True, +- 'action': 'click', +- 'position': (current_pos.x, current_pos.y), +- 'button': button, +- 'clicks': clicks +- } +- except Exception as e: +- return { +- 'success': False, +- 'action': 'click', +- 'error': str(e) +- } +- +- def double_click(self, x: Optional[int] = None, y: Optional[int] = None, +- button: str = 'left', duration: float = 0.0) -> dict: +- """ +- 鼠标双击操作 +- +- Args: +- x: X坐标，None表示当前位置 +- y: Y坐标，None表示当前位置 +- button: 鼠标按键 'left'/'right'/'middle' +- duration: 移动到目标位置的持续时间（秒） +- +- Returns: +- 包含操作结果的字典 +- +- Examples: +- >>> mouse.double_click(300, 400) # 双击坐标(300, 400) +- """ +- return self.click(x, y, clicks=2, button=button, duration=duration) +- +- def right_click(self, x: Optional[int] = None, y: Optional[int] = None, +- duration: float = 0.0) -> dict: +- """ +- 鼠标右键点击 +- +- Args: +- x: X坐标，None表示当前位置 +- y: Y坐标，None表示当前位置 +- duration: 移动到目标位置的持续时间（秒） +- +- Returns: +- 包含操作结果的字典 +- +- Examples: +- >>> mouse.right_click(500, 300) # 右键点击坐标(500, 300) +- """ +- return self.click(x, y, button='right', duration=duration) +- +- def move(self, x: int, y: int, duration: float = 0.0) -> dict: +- """ +- 移动鼠标到指定位置 +- +- Args: +- x: 目标X坐标 +- y: 目标Y坐标 +- duration: 移动持续时间（秒），0表示立即移动 +- +- Returns: +- 包含操作结果的字典 +- +- Examples: +- >>> mouse.move(800, 600) # 立即移动到(800, 600) +- >>> mouse.move(800, 600, duration=1.0) # 用1秒时间移动到(800, 600) +- """ +- try: +- pyautogui.moveTo(x, y, duration=duration) +- current_pos = pyautogui.position() +- return { +- 'success': True, +- 'action': 'move', +- 'position': (current_pos.x, current_pos.y), +- 'duration': duration +- } +- except Exception as e: +- return { +- 'success': False, +- 'action': 'move', +- 'error': str(e) +- } +- +- def move_rel(self, x_offset: int, y_offset: int, duration: float = 0.0) -> dict: +- """ +- 相对移动鼠标（从当前位置偏移） +- +- Args: +- x_offset: X轴偏移量 +- y_offset: Y轴偏移量 +- duration: 移动持续时间（秒） +- +- Returns: +- 包含操作结果的字典 +- +- Examples: +- >>> mouse.move_rel(100, 0) # 向右移动100像素 +- """ +- try: +- pyautogui.moveRel(x_offset, y_offset, duration=duration) +- current_pos = pyautogui.position() +- return { +- 'success': True, +- 'action': 'move_rel', +- 'position': (current_pos.x, current_pos.y), +- 'offset': (x_offset, y_offset), +- 'duration': duration +- } +- except Exception as e: +- return { +- 'success': False, +- 'action': 'move_rel', +- 'error': str(e) +- } +- +- def drag(self, x: int, y: int, duration: float = 0.5, +- button: str = 'left') -> dict: +- """ +- 拖拽鼠标到指定位置 +- +- Args: +- x: 目标X坐标 +- y: 目标Y坐标 +- duration: 拖拽持续时间（秒） +- button: 使用的鼠标按键 +- +- Returns: +- 包含操作结果的字典 +- +- Examples: +- >>> mouse.drag(500, 500, duration=1.0) # 拖拽到(500, 500) +- """ +- try: +- pyautogui.drag(x, y, duration=duration, button=button) +- current_pos = pyautogui.position() +- return { +- 'success': True, +- 'action': 'drag', +- 'position': (current_pos.x, current_pos.y), +- 'button': button, +- 'duration': duration +- } +- except Exception as e: +- return { +- 'success': False, +- 'action': 'drag', +- 'error': str(e) +- } +- +- def drag_rel(self, x_offset: int, y_offset: int, duration: float = 0.5, +- button: str = 'left') -> dict: +- """ +- 相对拖拽（从当前位置偏移） +- +- Args: +- x_offset: X轴偏移量 +- y_offset: Y轴偏移量 +- duration: 拖拽持续时间（秒） +- button: 使用的鼠标按键 +- +- Returns: +- 包含操作结果的字典 +- +- Examples: +- >>> mouse.drag_rel(100, 50) # 向右下拖拽 +- """ +- try: +- pyautogui.dragRel(x_offset, y_offset, duration=duration, button=button) +- current_pos = pyautogui.position() +- return { +- 'success': True, +- 'action': 'drag_rel', +- 'position': (current_pos.x, current_pos.y), +- 'offset': (x_offset, y_offset), +- 'button': button, +- 'duration': duration +- } +- except Exception as e: +- return { +- 'success': False, +- 'action': 'drag_rel', +- 'error': str(e) +- } +- +- def scroll(self, clicks: int, x: Optional[int] = None, +- y: Optional[int] = None) -> dict: +- """ +- 滚动鼠标滚轮 +- +- Args: +- clicks: 滚动单位数，正数向上滚动，负数向下滚动 +- x: 在指定位置滚动的X坐标（可选） +- y: 在指定位置滚动的Y坐标（可选） +- +- Returns: +- 包含操作结果的字典 +- +- Examples: +- >>> mouse.scroll(5) # 向上滚动5个单位 +- >>> mouse.scroll(-3) # 向下滚动3个单位 +- >>> mouse.scroll(5, 500, 500) # 在(500, 500)位置向上滚动 +- """ +- try: +- if x is not None and y is not None: +- pyautogui.scroll(clicks, x, y) +- else: +- pyautogui.scroll(clicks) +- +- return { +- 'success': True, +- 'action': 'scroll', +- 'clicks': clicks, +- 'position': (x, y) if x and y else 'current' +- } +- except Exception as e: +- return { +- 'success': False, +- 'action': 'scroll', +- 'error': str(e) +- } +- +- def get_position(self) -> dict: +- """ +- 获取当前鼠标位置 +- +- Returns: +- 包含鼠标位置的字典 +- +- Examples: +- >>> pos = mouse.get_position() +- >>> print(f"鼠标位置: ({pos['x']}, {pos['y']})") +- """ +- try: +- pos = pyautogui.position() +- return { +- 'success': True, +- 'x': pos.x, +- 'y': pos.y, +- 'position': (pos.x, pos.y) +- } +- except Exception as e: +- return { +- 'success': False, +- 'error': str(e) +- } +- +- def mouse_down(self, x: Optional[int] = None, y: Optional[int] = None, +- button: str = 'left') -> dict: +- """ +- 按下鼠标按键（不释放） +- +- Args: +- x: X坐标（可选） +- y: Y坐标（可选） +- button: 鼠标按键 +- +- Returns: +- 包含操作结果的字典 +- """ +- try: +- if x is not None and y is not None: +- pyautogui.mouseDown(x, y, button=button) +- else: +- pyautogui.mouseDown(button=button) +- +- return { +- 'success': True, +- 'action': 'mouse_down', +- 'button': button +- } +- except Exception as e: +- return { +- 'success': False, +- 'action': 'mouse_down', +- 'error': str(e) +- } +- +- def mouse_up(self, x: Optional[int] = None, y: Optional[int] = None, +- button: str = 'left') -> dict: +- """ +- 释放鼠标按键 +- +- Args: +- x: X坐标（可选） +- y: Y坐标（可选） +- button: 鼠标按键 +- +- Returns: +- 包含操作结果的字典 +- """ +- try: +- if x is not None and y is not None: +- pyautogui.mouseUp(x, y, button=button) +- else: +- pyautogui.mouseUp(button=button) +- +- return { +- 'success': True, +- 'action': 'mouse_up', +- 'button': button +- } +- except Exception as e: +- return { +- 'success': False, +- 'action': 'mouse_up', +- 'error': str(e) +- } +- +- +-# 创建全局实例，支持直接调用 +-_mouse_instance = Mouse() +- +-# 导出便捷函数 +-click = _mouse_instance.click +-double_click = _mouse_instance.double_click +-right_click = _mouse_instance.right_click +-move = _mouse_instance.move +-move_rel = _mouse_instance.move_rel +-drag = _mouse_instance.drag +-drag_rel = _mouse_instance.drag_rel +-scroll = _mouse_instance.scroll +-get_position = _mouse_instance.get_position +-mouse_down = _mouse_instance.mouse_down +-mouse_up = _mouse_instance.mouse_up +\ No newline at end of file +Index: core/llm/code/agent.py +=================================================================== +diff --git a/core/llm/code/agent.py b/core/llm/code/agent.py +deleted file mode 100644 +--- a/core/llm/code/agent.py (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,202 +0,0 @@ +-import logging +-import os +-import time +-import queue +-import threading +-from queue import Queue +-from dotenv import load_dotenv +-from litellm import completion +-from core.computer.code import Code +-from .default_prompt import default_prompt, default_prompt_end +-from .code_parser import CodeParser +- +-dotenv_path = os.path.join(os.path.dirname(__file__), ".env") +-load_dotenv(dotenv_path) +- +-class CodeAgent: +- def __init__(self): +- self.code_executer = Code() +- self.SYSTEM_PROMPT = default_prompt.format(language=str(self.code_executer.language_list)) +- logging.info("[CodeAgent]"+self.SYSTEM_PROMPT) +- self.SYSTEM_PROMPT_END = default_prompt_end +- self.loop_breakers = ["The task is done.", "The task is impossible."] +- self.model = os.getenv("CodeAgent_MODEL") +- self.api_base = os.getenv("CodeAgent_API_BASE") +- self.api_key = os.getenv("CodeAgent_API_KEY") +- self.stop_code = False +- self.stop_agent = False +- self.permission = None +- +- def _listener(self, message_from_client: Queue): +- """ +- 监听来自客户端的信息 +- :param message_from_client: +- :return: +- """ +- while True: +- try: +- message = message_from_client.get(timeout=0.5) +- except queue.Empty: +- continue +- print("get client message") +- # 不是发给CodeAgent的信息，放回 +- if message["name"] != "CodeAgent": +- message_from_client.put(message) +- continue +- # 允许执行代码 +- if message["type"] == "request": +- if message["content"] == "deny": +- self.permission = False +- logging.info("[CodeAgent]User denied execution") +- elif message["content"] == "approve": +- self.permission = True +- logging.info("[CodeAgent]User approved execution") +- # 停止相关 +- elif message["content"] == "stop_agent": +- self.stop_agent = True +- logging.info("[CodeAgent]User stopped agent") +- elif message["content"] == "stop_code": +- self.stop_code = True +- logging.info("[CodeAgent]User stopped code execution") +- +- def _should_stop(self, ai_content: str = None): +- if ai_content: +- for loop_breaker in self.loop_breakers: +- if loop_breaker in ai_content: +- return True +- return False +- +- def task(self, description: str, message_from_client: Queue, message_to_client: Queue): +- self.stop_agent = False +- messages = [ +- {"role": "system", "content": self.SYSTEM_PROMPT}, +- {"role": "user", "content": description}] +- listener_thread = threading.Thread(target=self._listener, args=(message_from_client,)) +- listener_thread.daemon = True +- listener_thread.start() +- logging.info("[CodeAgent][START]") +- message_to_client.put({"name": "CodeAgent", "type": "status", "content": "[START]"}) +- +- while not self.stop_agent: +- # 请求ai对话 +- response = completion( +- model=self.model, +- api_base=self.api_base, +- api_key=self.api_key, +- messages=messages, +- stream=True, +- ) +- # 流式返回 +- message_to_client.put({"name": "CodeAgent", "type": "ai_content", "content": "[BEGIN]"}) +- ai_content = "" +- for chunk in response: +- # 检查是否需要停止 +- if self.stop_agent: +- logging.info("[CodeAgent][STOP]: User stop") +- message_to_client.put({"name": "CodeAgent", "type": "status", "content": "[STOP]"}) +- return +- if chunk.choices[0].delta.content: +- delta = chunk.choices[0].delta.content +- message_to_client.put({"name": "CodeAgent", "type": "ai_content", "content": delta}) +- ai_content += delta +- message_to_client.put({"name": "CodeAgent", "type": "ai_content", "content": "[END]"}) +- messages.append({"role": "assistant", "content": ai_content}) +- +- # 检查ai是否想要停止 +- if self._should_stop(ai_content): +- break +- +- # TODO: If the context is too long, use RAG or other methods to handle it +- # if len(messages) > 10: +- # messages = [messages[0]] + messages[-5:] +- +- # 获取模型返回的代码 +- codes = CodeParser(ai_content) +- if not codes: +- messages.append({"role": "user", "content": "No code blocks found. Skip execution. " + self.SYSTEM_PROMPT_END}) +- continue +- +- # 依次执行代码 +- logging.info(f"Find {len(codes)} code blocks. Executing...") +- for i in range(len(codes)): +- +- logging.info(f"Executing code block {i}..., code: {codes[i]}") +- code = codes[i] +- self.permission = None +- message_to_client.put({"name": "CodeAgent", "type": "status", "content": f"[BLOCK{i}]"}) +- message_to_client.put({"name": "CodeAgent", "type": "request", "content": f"[BLOCK{i}]need_permission"}) +- +- # 阻塞等待客户端返回允许结果 +- print("wtf???") +- while self.permission is None: +- logging.info("waiting permission") +- print("waiting permission") +- # 检查是否需要停止 +- time.sleep(0.5) +- if self.stop_agent: +- logging.info("[CodeAgent][STOP]: User stop") +- message_to_client.put({"name": "CodeAgent", "type": "status", "content": "[STOP]"}) +- return +- if not self.permission: +- logging.info(f"User rejected to excecute code block{i}") +- messages.append({"role": "user", "content": f"User rejected to excecute code block{i}"}) +- continue +- +- logging.info(f"User approved to excecute code block {i}, start executing") +- results = self.code_executer.run(lang=code["lang"], code=code["code"]) # 正确解包 +- code_output_content = [{"type": "text", "text": f"Code block{i} is executed, and the output is as follows:"}] +- +- self.stop_code = False +- while self.code_executer.is_running() or not results.empty(): +- logging.info(f"[CodeAgent]: Code block{i} is running, code_executer:{self.code_executer.is_running()}, results.empty():{results.empty()}") +- # 检查是否需要停止 +- if self.stop_code or self.stop_agent: +- self.code_executer.interrupt() +- break +- +- try: +- delta = results.get(timeout=0.2) +- except queue.Empty: +- continue +- +- # TODO: 有bug，图片先不返回 +- if delta["type"] == "image/png" or delta["type"] == "image/jpeg": +- # pass +- message_to_client.put({"name": "CodeAgent", **delta}) +- code_output_content.append( +- { +- "type": "image_url", +- "image_url": { +- "url": f"data:{delta["type"]};base64,{delta["content"]}" +- } +- } +- ) +- else: # 暂时把其他所有内容转为text +- message_to_client.put({"name": "CodeAgent", "type": "text", "content": delta["content"]}) +- code_output_content.append( +- { +- "type": "text", +- "text": delta["content"] +- } +- ) +- +- logging.info(f"Finish executing code block {i}") +- messages.append({"role": "user", "content": code_output_content}) +- +- messages.append({"role": "user", "content": "All code blocks are executed. " + self.SYSTEM_PROMPT_END}) +- +- logging.info("[CodeAgent][STOP]: task finished") +- message_to_client.put({"name": "CodeAgent", "type": "status", "content": "[STOP]"}) +- return messages[-1]["content"] +- +- +- +- +- +- +- +- +- +- +- +- +Index: core/computer/code/code.py +=================================================================== +diff --git a/core/computer/code/code.py b/core/computer/code/code.py +deleted file mode 100644 +--- a/core/computer/code/code.py (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,74 +0,0 @@ +-import queue +- +-from .languages import PythonLanguage, BashLanguage, PowerShellLanguage +- +- +-class Code: +- def __init__(self): +- self.python = PythonLanguage() +- # self.cmd = CMDLanguage() +- self.bash = BashLanguage() +- self.powershell = PowerShellLanguage() +- self.current_language = None +- self.language_map = { +- "python": self.python, +- # "cmd": self.cmd, +- "powershell": self.powershell, +- "pwsh": self.powershell, # PowerShell Core别名 +- "bash": self.bash, +- "sh": self.bash, # Shell别名 +- } +- self.language_list = [] +- for lang, lang_obj in self.language_map.items(): +- if hasattr(lang_obj, 'is_available'): +- if lang_obj.is_available(): +- self.language_list.append(lang) +- else: +- # Python总是可用的 +- self.language_list.append(lang) +- +- # def language_list(self): +- # available = [] +- +- def run(self, lang: str, code: str): +- """ +- 运行代码 +- +- :param lang: 要运行的代码语言 +- :type lang: str +- :param code: 要运行的代码 +- :type code: str +- """ +- lang = lang.lower() +- if lang in self.language_list: +- self.current_language = self.language_map[lang] +- return self.current_language.run(code) +- else: +- message = queue.Queue() +- message.put({"type": "error", "content": f"[Code]Unsupported language:{lang}"}) +- return message +- +- def interrupt(self): +- """ +- 中断执行中的代码 +- """ +- message = queue.Queue() +- if self.current_language and self.current_language.is_running: +- self.current_language.interrupt() +- message.put({"type": "text", "content": "Code execution interrupted"}) +- else: +- message.put({"type": "text", "content": "No code is running"}) +- return message +- +- def get_elapsed_time(self): +- """ +- 获取代码已运行时间 +- """ +- if self.current_language: +- return self.current_language.get_elapsed_time() +- +- def is_running(self): +- """ +- 代码是否正在运行 +- """ +- return self.current_language and self.current_language.is_running +\ No newline at end of file +Index: core/computer/code/languages/__init__.py +=================================================================== +diff --git a/core/computer/code/languages/__init__.py b/core/computer/code/languages/__init__.py +deleted file mode 100644 +--- a/core/computer/code/languages/__init__.py (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,12 +0,0 @@ +-""" +- +-""" +- +- +-from .bash import BashLanguage +-from .powershell import PowerShellLanguage +-from .python import PythonLanguage +- +-__all__ = ['PythonLanguage', 'BashLanguage', 'PowerShellLanguage'] +- +-__version__ = '1.0.0' +\ No newline at end of file +Index: core/llm/code/消息设计.md +=================================================================== +diff --git a/core/llm/code/消息设计.md b/core/llm/code/消息设计.md +deleted file mode 100644 +--- a/core/llm/code/消息设计.md (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,83 +0,0 @@ +-# 消息设计 (Message Design) +- +-Agent 与客户端（UI）之间通过 `Queue` 进行通信，消息格式为 JSON 字典。 +- +-## 1. 客户端 -> Agent (Client to Agent) +-客户端发送的消息主要用于控制 Agent 的行为和响应权限请求。 +- +-**基本格式:** +-```json +-{ +- "name": "CodeAgent", // 给谁发送就填谁 +- "type": "request", +- "content": "指令内容" +-} +-``` +- +-**指令内容 (content):** +-- `approve`: 允许执行当前代码块。 +-- `deny`: 拒绝执行当前代码块。 +-- `stop_agent`: 停止整个 Agent 任务。 +-- `stop_code`: 停止当前正在执行的代码（中断执行）。 +- +-## 3. Agent -> 客户端 (Agent to Client) +-Agent 发送的消息用于更新状态、流式传输 AI 回复、请求权限以及展示代码执行结果。 +- +-**基本格式:** +-```json +-{ +- "name": "CodeAgent", +- "type": "消息类型", +- "content": "消息内容" +-} +-``` +- +-### 3.1 状态消息 (type: "status") +-用于通知 UI 当前 Agent 的工作阶段。 +-- `[START]`: 任务开始。 +-- `[STOP]`: 任务结束（完成或被用户停止）。 +-- `[BLOCK{i}]`: 正在处理第 `i` 个代码块（准备执行）。 +- +-### 3.2 AI 内容流 (type: "ai_content") +-用于流式展示 LLM 的思考和回复过程。 +-- `[BEGIN]`: 开始生成内容。 +-- `[END]`: 内容生成结束。 +-- ``: AI 生成的文本片段（增量更新）。 +- +-### 3.3 权限请求 (type: "request") +-当 Agent 解析出代码块并准备执行时，向 UI 请求权限。 +-- `[BLOCK{i}]need_permission`: 请求执行第 `i` 个代码块的权限。 +- - UI 收到此消息后应显示“批准/拒绝”按钮。 +- +-### 3.4 代码执行输出 (type: "text" / "image/png" 等) +-展示代码执行的实时输出。 +- +-- **文本输出 (type: "text"):** +- - `content`: 标准输出 (stdout) 或标准错误 (stderr) 的文本内容。 +- +-- **图片输出 (type: "image/png" / "image/jpeg"):** +- - *注意: 目前 `agent.py` 中图片处理逻辑被注释/跳过，需要修复才能正常接收图片。* +- - `content`: Base64 编码的图片数据。 +- +-- **其他类型:** +- - `html`: HTML 内容。 +- - `javascript`: JS 代码。 +- - `error`: 执行错误信息。 +- +-## 4. 交互流程示例 (Interaction Flow) +- +-调用`CodeAgent.task("任务描述")`之后： +- +-1. **任务开始**: Agent 发送 `status: [START]`。 +-2. **AI 思考**: Agent 发送 `ai_content: [BEGIN]`, 随后发送多条内容片段, 最后发送 `ai_content: [END]`。 +-3. **代码解析**: Agent 发现代码块。(Agent发现的代码块并不会传给客户端，客户端需要自己按顺序切分代码片段) +-4. **请求权限**: Agent 发送 `status: [BLOCK0]`, `request: [BLOCK0]need_permission`。 +-5. **用户确认**: 客户端发送 `content: approve`。 +-6. **代码执行**: Agent 执行代码，持续发送 `type: text` (日志/结果)。 +-7. **执行完成**: Agent 继续下一轮或结束。 +-8. **任务结束**: Agent 发送 `status: [STOP]`。 +- +-## 5. TODO (Suggestions) +-1. **修复图片传输**: base64太长，之后可以采取压缩方法 +-2. **结构化状态**: 目前的 `[START]` 等状态码混在 `content` 字符串中，建议后续优化为独立的枚举字段，方便前端解析。 +-3. **上下文未处理** +Index: core/computer/mouse/__init__.py +=================================================================== +diff --git a/core/computer/mouse/__init__.py b/core/computer/mouse/__init__.py +deleted file mode 100644 +--- a/core/computer/mouse/__init__.py (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,33 +0,0 @@ +-""" +-鼠标操作包 +-""" +-from .mouse import ( +- Mouse, +- click, +- double_click, +- right_click, +- move, +- move_rel, +- drag, +- drag_rel, +- scroll, +- get_position, +- mouse_down, +- mouse_up +-) +- +-__all__ = [ +- # Mouse类和函数 +- 'Mouse', +- 'click', +- 'double_click', +- 'right_click', +- 'move', +- 'move_rel', +- 'drag', +- 'drag_rel', +- 'scroll', +- 'get_position', +- 'mouse_down', +- 'mouse_up', +-] +\ No newline at end of file +Index: core/computer/code/__init__.py +=================================================================== +diff --git a/core/computer/code/__init__.py b/core/computer/code/__init__.py +deleted file mode 100644 +--- a/core/computer/code/__init__.py (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,11 +0,0 @@ +-""" +-代码运行包 +-""" +- +- +-from .base_language import BaseLanguage +-from .code import Code +- +-__all__ = ['Code', 'BaseLanguage'] +- +-__version__ = '1.0.0' +\ No newline at end of file +Index: core/computer/screen/screen.py +=================================================================== +diff --git a/core/computer/screen/screen.py b/core/computer/screen/screen.py +deleted file mode 100644 +--- a/core/computer/screen/screen.py (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,35 +0,0 @@ +-import base64 +-import io +- +-from PIL import Image, ImageGrab +- +- +-class Screen: +- def __init__(self): +- pass +- def screenshot(self, resize_factor: float = 0.5, format: str= "png", quality: int = 100): +- """ +- 获取截屏 +- :param resize_factor: 缩放因子 +- :param format: 格式，可选png/jpeg +- :param quality: 图片质量，jpeg有效 +- :return: [{"type": f"image/{format}", "content": base64_encoded}] +- """ +- image = ImageGrab.grab() +- new_width = int(image.size[0] * resize_factor) +- new_height = int(image.size[1] * resize_factor) +- image = image.resize((new_width, new_height), Image.LANCZOS) +- +- img_byte_arr= io.BytesIO() +- format = format.lower() +- if format == "png": +- image.save(img_byte_arr, format=format) +- elif format == "jpeg": +- image.save(img_byte_arr, format=format, quality=quality) +- else: +- raise ValueError("[Screen]Unsupported format: ", format) +- +- img_bytes = img_byte_arr.getvalue() +- base64_encoded = base64.b64encode(img_bytes).decode('utf-8') +- +- return [{"type": f"image/{format}", "content": base64_encoded}] +\ No newline at end of file +Index: core/computer/code/languages/python.py +=================================================================== +diff --git a/core/computer/code/languages/python.py b/core/computer/code/languages/python.py +deleted file mode 100644 +--- a/core/computer/code/languages/python.py (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,131 +0,0 @@ +-import logging +-import queue +-import re +-import threading +-import time +- +-from jupyter_client.manager import KernelManager +- +-from ..base_language import BaseLanguage +- +- +-class PythonLanguage(BaseLanguage): +- def __init__(self): +- super().__init__() +- self.km = None +- self.kc = None +- self.current_msg_id = None +- +- def start(self): +- self.should_stop = False +- self.start_time = time.time() +- self.elapsed_time = 0 +- +- if self.km: +- self.stop() +- try: +- self.km = KernelManager(kernel_name='python3') +- self.is_running = True +- self.km.start_kernel() +- self.kc = self.km.client() +- self.kc.start_channels() +- self.kc.wait_for_ready() +- logging.info("[PythonLanguage]Started kernel client&manager") +- except Exception as e: +- self.stop() +- logging.error("[PythonLanguage]Error starting kernel: %s", e) +- +- def stop(self): +- self.is_running = False +- if self.start_time: +- self.elapsed_time = time.time() - self.start_time +- self.start_time = None +- +- try: +- if self.km: +- self.km.shutdown_kernel() +- self.is_running = False +- self.km = None +- if self.kc: +- self.kc.stop_channels() +- self.current_msg_id = None +- self.kc = None +- logging.info("[PythonLanguage]Stopped kernel client&manager") +- except Exception as e: +- logging.error("[PythonLanguage]Error during cleanup kernel: %s", e) +- +- def wait_for_shutdown(self): +- while self.is_running or self.kc: +- time.sleep(0.1) +- +- def run(self, code: str): +- message = queue.Queue() +- +- excecution_thread = threading.Thread(target=self._execute_jupyter, args=(code, message)) +- excecution_thread.daemon = True +- excecution_thread.start() +- +- return message +- +- def _execute_jupyter(self, code: str, message: queue.Queue): +- self.start() +- +- if self.km.is_alive() is False: +- message.put({"type": "error", "content": "[PythonLanguage]Faild to start Jupyter kernel"}) +- return +- +- # 只在需要时添加matplotlib魔法命令 +- if "matplotlib" in code and "%matplotlib" not in code: +- code = "%matplotlib inline\n" + code +- +- try: +- self.current_msg_id=self.kc.execute(code) +- except Exception as e: +- message.put({"type": "error", "content": f"[PythonLanguage]Error while executing code: {e}"}) +- logging.error("[PythonLanguage]Error while executing code: %s", e) +- print("[PythonLanguage]start runing...") +- while True: +- if self.should_stop: +- self.km.interrupt_kernel() +- +- try: +- msg = self.kc.get_iopub_msg(timeout=1) +- except queue.Empty: +- continue +- +- if msg['parent_header'].get('msg_id') != self.current_msg_id: +- continue +- +- msg_type = msg['msg_type'] +- content = msg['content'] +- +- print("get_one_msg") +- +- if msg_type == "stream": +- message.put({"type": "text", "content": content['text']}) +- +- elif msg_type == "error": +- content = "\n".join(content["traceback"]) +- # 移除颜色标识 +- ansi_escape = re.compile(r"\x1B\[[0-?]*[ -/]*[@-~]") +- content = ansi_escape.sub("", content) +- message.put({"type": "text", "content": content}) +- +- elif msg_type in ["display_data", "execute_result"]: +- data = content["data"] +- if "image/png" in data: +- message.put({"type": "image/png", "content": data["image/png"]}) +- elif "image/jpeg" in data: +- message.put({"type": "image/jpeg", "content": data["image/jpeg"]}) +- elif "text/html" in data: +- message.put({"type": "html", "content": data["text/html"]}) +- elif "text/plain" in data: +- message.put({"type": "text", "content": data["text/plain"]}) +- elif "application/javascript" in data: +- message.put({"type": "javascript", "content": data["application/javascript"]}) +- +- elif msg_type == 'status': +- if content['execution_state'] == 'idle': +- break +- print("[PythonLanguage]finished.") +- self.stop() +Index: core/llm/code/code_parser.py +=================================================================== +diff --git a/core/llm/code/code_parser.py b/core/llm/code/code_parser.py +deleted file mode 100644 +--- a/core/llm/code/code_parser.py (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,10 +0,0 @@ +-import re +- +-def CodeParser(text: str): +- pattern = r"```([\w\+\-\.]*)\n([\s\S]*?)\n```" +- matches = re.findall(pattern, text) +- +- results = [] +- for lang, code in matches: +- results.append({"lang": lang.strip() if lang else "text", "code": code}) +- return results +\ No newline at end of file +Index: core/computer/code/README.md +=================================================================== +diff --git a/core/computer/code/README.md b/core/computer/code/README.md +deleted file mode 100644 +--- a/core/computer/code/README.md (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,214 +0,0 @@ +-# Code执行模块 +- +-这是一个多语言代码执行模块，支持Python、Bash和PowerShell代码的异步执行。 +- +-## 功能概述 +- +-该模块提供了一个统一的接口来执行不同编程语言的代码，并通过队列返回执行结果。支持代码中断、执行时间监控等功能。 +- +-## 架构设计 +- +-### 核心组件 +- +-- **Code类**: 主要的代码执行管理器，负责语言选择和执行调度 +-- **BaseLanguage类**: 所有语言实现的抽象基类 +-- **PythonLanguage类**: Python代码执行器，基于Jupyter内核 +-- **BashLanguage类**: Bash/Shell代码执行器，基于subprocess +-- **PowerShellLanguage类**: PowerShell代码执行器，基于subprocess +- +-### 消息系统 +- +-所有执行结果通过`queue.Queue`对象返回，消息格式为字典： +- +-```python +-{ +- "type": "消息类型", +- "content": "消息内容" +-} +-``` +- +-## run()返回的消息格式 +- +-- **`text`**: 纯文本输出，包括标准输出、错误输出、执行结果等 +-- **`image/png`**: PNG格式的图片数据（base64编码） +-- **`image/jpeg`**: JPEG格式的图片数据（base64编码） +-- **`html`**: HTML格式的内容 +-- **`text/plain`**: 纯文本内容 +-- **`application/javascript`**: JavaScript代码 +-- **`error`**: 模块的错误信息，不包含代码运行的错误 +- +-## 使用方法 +- +-### 基本用法 +- +-```python +-from core.computer.code import Code +- +-# 创建代码执行器 +-code_executor = Code() +- +-# 执行Python代码 +-result_queue = code_executor.run("python", "print('Hello, World!')") +- +-# 执行Bash代码 +-result_queue = code_executor.run("bash", "ls -la") +- +-# 执行PowerShell代码 +-result_queue = code_executor.run("powershell", "Get-ChildItem") +- +-# 获取执行结果 +-while not result_queue.empty(): +- message = result_queue.get() +- print(f"{message['type']}: {message['content']}") +-``` +- +-### 支持的语言 +- +-- Python: `python` +-- Bash: `bash`, `sh` +-- PowerShell: `powershell`, `pwsh` +- +-### 中断执行 +- +-```python +-# 中断当前正在执行的代码 +-result_queue = code_executor.interrupt() +-``` +- +-### 监控执行状态 +- +-```python +-# 检查是否有代码正在运行 +-is_running = code_executor.is_running() +- +-# 获取已运行时间 +-elapsed_time = code_executor.get_elapsed_time() +-``` +- +-## 各语言实现详情 +- +-### Python实现 (PythonLanguage) +- +-- **技术**: 基于Jupyter内核和jupyter_client +-- **特性**: +- - 支持matplotlib内联绘图 +- - 支持图片、HTML、JavaScript等多种输出格式 +- - 完整的Python语法支持 +- - 自动添加`%matplotlib inline`以支持可视化 +-- **依赖**: jupyter_client, ipython kernel +- +-### Bash实现 (BashLanguage) +- +-- **技术**: 基于subprocess.Popen +-- **特性**: +- - 检测bash是否可用 +- - 实时输出流 +- - 支持信号中断 +-- **可用性检查**: `shutil.which("bash")` +- +-### PowerShell实现 (PowerShellLanguage) +- +-- **技术**: 基于subprocess.Popen +-- **特性**: +- - 优先使用pwsh (PowerShell Core)，回退到powershell +- - 跨平台支持 +- - 无配置文件执行 (`-NoProfile`) +-- **可用性检查**: `shutil.which("powershell") or shutil.which("pwsh")` +- +-## 错误处理 +- +-### 模块级错误 +- +-- 不支持的语言 +-- 系统环境问题 +-- 内核启动失败 +- +-### 代码执行错误 +- +-- 语法错误 +-- 运行时错误 +-- 依赖缺失 +- +-## 线程安全 +- +-- 每次代码执行都在独立的线程中运行 +-- 使用队列进行线程间通信 +-- 支持优雅的中断机制 +- +-## 性能考虑 +- +-- Python内核会重复启动和停止，适合短时间执行 +-- Bash和PowerShell为每次执行创建新进程 +-- 所有语言都支持实时输出，不会阻塞主线程 +- +-## 依赖要求 +- +-### Python环境 +-- Python 3.7+ +-- jupyter_client +-- IPython kernel +- +-### 系统环境 +-- Bash: Git Bash, WSL, 或Linux/macOS环境 +-- PowerShell: Windows PowerShell 5.1+ 或 PowerShell Core 6.0+ +- +-## 示例代码 +- +-```python +-# 完整示例 +-from core.computer.code import Code +-import time +- +-def execute_code_example(): +- executor = Code() +- +- # Python可视化示例 +- print("执行Python代码...") +- result = executor.run("python", """ +-import matplotlib.pyplot as plt +-import numpy as np +- +-x = np.linspace(0, 10, 100) +-y = np.sin(x) +- +-plt.figure(figsize=(8, 4)) +-plt.plot(x, y) +-plt.title('正弦波形') +-plt.xlabel('x') +-plt.ylabel('sin(x)') +-plt.grid(True) +-plt.show() +-""") +- +- # 读取结果 +- while not result.empty() or executor.is_running(): +- if not result.empty(): +- msg = result.get() +- print(f"收到消息: {msg['type']}") +- if msg['type'].startswith('image/'): +- print(f"图片数据: {len(msg['content'])} 字节") +- else: +- print(f"内容: {msg['content'][:100]}...") +- time.sleep(0.1) +- +- print("Python执行完成") +- +-if __name__ == "__main__": +- execute_code_example() +-``` +- +-## 注意事项 +- +-1. **Python内核**: 每次执行都会重新启动内核，状态不会保留 +-2. **安全性**: 该模块设计用于受控环境，避免执行不受信任的代码 +-3. **资源管理**: 长时间运行可能产生大量进程，建议监控资源使用 +-4. **平台兼容性**: 某些功能在不同操作系统上表现可能不同 +- +-## 未来改进 +- +-- [ ] 支持更多编程语言 (Node.js, Ruby等) +-- [ ] Python内核持久化，避免重复启动 +-- [ ] 代码执行超时设置 +-- [ ] 更细粒度的权限控制 +-- [ ] 代码执行历史记录 +-- [ ] 性能监控和分析 +Index: core/llm/code/__init__.py +=================================================================== +diff --git a/core/llm/code/__init__.py b/core/llm/code/__init__.py +deleted file mode 100644 +--- a/core/llm/code/__init__.py (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,3 +0,0 @@ +-from .agent import CodeAgent +- +-__all__ = ["CodeAgent"] +\ No newline at end of file +Index: core/computer/keyboard/keyboard.py +=================================================================== +diff --git a/core/computer/keyboard/keyboard.py b/core/computer/keyboard/keyboard.py +deleted file mode 100644 +--- a/core/computer/keyboard/keyboard.py (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,296 +0,0 @@ +-import pyautogui +-import time +-from typing import Union, List, Optional +- +- +-class Keyboard: +- """键盘操作类，提供各种键盘控制功能""" +- +- def __init__(self): +- """初始化键盘模块""" +- pyautogui.PAUSE = 0.05 # 每次操作后暂停0.05秒 +- +- def type_text(self, text: str, interval: float = 0.0) -> dict: +- """ +- 输入文本 +- +- Args: +- text: 要输入的文本 +- interval: 每个字符之间的间隔时间（秒） +- +- Returns: +- 包含操作结果的字典 +- +- Examples: +- >>> keyboard.type_text("Hello World") +- >>> keyboard.type_text("慢速输入", interval=0.1) +- +- Note: +- 对于中文输入，需要系统已经切换到中文输入法 +- """ +- try: +- pyautogui.write(text, interval=interval) +- return { +- 'success': True, +- 'action': 'type_text', +- 'text': text, +- 'length': len(text), +- 'interval': interval +- } +- except Exception as e: +- return { +- 'success': False, +- 'action': 'type_text', +- 'error': str(e) +- } +- +- def press(self, key: str, presses: int = 1, interval: float = 0.0) -> dict: +- """ +- 按下指定键 +- +- Args: +- key: 键名，如 'enter', 'space', 'a', 'ctrl' 等 +- presses: 按键次数 +- interval: 多次按键之间的间隔（秒） +- +- Returns: +- 包含操作结果的字典 +- +- Examples: +- >>> keyboard.press('enter') # 按下回车键 +- >>> keyboard.press('tab', presses=3) # 按3次Tab键 +- >>> keyboard.press('backspace') # 按退格键 +- +- 常用键名: +- - 字母: 'a', 'b', 'c' ... +- - 数字: '0', '1', '2' ... +- - 功能键: 'f1', 'f2' ... 'f12' +- - 方向键: 'up', 'down', 'left', 'right' +- - 特殊键: 'enter', 'space', 'tab', 'backspace', 'delete', 'esc' +- - 修饰键: 'ctrl', 'shift', 'alt', 'win' (或 'command' 在Mac上) +- """ +- try: +- pyautogui.press(key, presses=presses, interval=interval) +- return { +- 'success': True, +- 'action': 'press', +- 'key': key, +- 'presses': presses, +- 'interval': interval +- } +- except Exception as e: +- return { +- 'success': False, +- 'action': 'press', +- 'error': str(e) +- } +- +- def hotkey(self, *keys: str) -> dict: +- """ +- 按下组合键（快捷键） +- +- Args: +- *keys: 要同时按下的键，按顺序传入 +- +- Returns: +- 包含操作结果的字典 +- +- Examples: +- >>> keyboard.hotkey('ctrl', 'c') # Ctrl+C 复制 +- >>> keyboard.hotkey('ctrl', 'v') # Ctrl+V 粘贴 +- >>> keyboard.hotkey('ctrl', 'shift', 's') # Ctrl+Shift+S +- >>> keyboard.hotkey('alt', 'tab') # Alt+Tab 切换窗口 +- >>> keyboard.hotkey('ctrl', 'a') # Ctrl+A 全选 +- """ +- try: +- pyautogui.hotkey(*keys) +- return { +- 'success': True, +- 'action': 'hotkey', +- 'keys': keys, +- 'combination': '+'.join(keys) +- } +- except Exception as e: +- return { +- 'success': False, +- 'action': 'hotkey', +- 'error': str(e) +- } +- +- def key_down(self, key: str) -> dict: +- """ +- 按下键（不释放） +- +- Args: +- key: 键名 +- +- Returns: +- 包含操作结果的字典 +- +- Examples: +- >>> keyboard.key_down('shift') # 按下Shift键 +- """ +- try: +- pyautogui.keyDown(key) +- return { +- 'success': True, +- 'action': 'key_down', +- 'key': key +- } +- except Exception as e: +- return { +- 'success': False, +- 'action': 'key_down', +- 'error': str(e) +- } +- +- def key_up(self, key: str) -> dict: +- """ +- 释放键 +- +- Args: +- key: 键名 +- +- Returns: +- 包含操作结果的字典 +- +- Examples: +- >>> keyboard.key_up('shift') # 释放Shift键 +- """ +- try: +- pyautogui.keyUp(key) +- return { +- 'success': True, +- 'action': 'key_up', +- 'key': key +- } +- except Exception as e: +- return { +- 'success': False, +- 'action': 'key_up', +- 'error': str(e) +- } +- +- def hold_key(self, key: str, duration: float = 1.0) -> dict: +- """ +- 按住一个键一段时间 +- +- Args: +- key: 键名 +- duration: 按住的时间（秒） +- +- Returns: +- 包含操作结果的字典 +- +- Examples: +- >>> keyboard.hold_key('space', duration=2.0) # 按住空格键2秒 +- """ +- try: +- pyautogui.keyDown(key) +- time.sleep(duration) +- pyautogui.keyUp(key) +- return { +- 'success': True, +- 'action': 'hold_key', +- 'key': key, +- 'duration': duration +- } +- except Exception as e: +- return { +- 'success': False, +- 'action': 'hold_key', +- 'error': str(e) +- } +- +- # 常用快捷键的便捷方法 +- +- def copy(self) -> dict: +- """复制 (Ctrl+C)""" +- return self.hotkey('ctrl', 'c') +- +- def paste(self) -> dict: +- """粘贴 (Ctrl+V)""" +- return self.hotkey('ctrl', 'v') +- +- def cut(self) -> dict: +- """剪切 (Ctrl+X)""" +- return self.hotkey('ctrl', 'x') +- +- def select_all(self) -> dict: +- """全选 (Ctrl+A)""" +- return self.hotkey('ctrl', 'a') +- +- def undo(self) -> dict: +- """撤销 (Ctrl+Z)""" +- return self.hotkey('ctrl', 'z') +- +- def redo(self) -> dict: +- """重做 (Ctrl+Y)""" +- return self.hotkey('ctrl', 'y') +- +- def save(self) -> dict: +- """保存 (Ctrl+S)""" +- return self.hotkey('ctrl', 's') +- +- def find(self) -> dict: +- """查找 (Ctrl+F)""" +- return self.hotkey('ctrl', 'f') +- +- def new_tab(self) -> dict: +- """新建标签页 (Ctrl+T)""" +- return self.hotkey('ctrl', 't') +- +- def close_tab(self) -> dict: +- """关闭标签页 (Ctrl+W)""" +- return self.hotkey('ctrl', 'w') +- +- def refresh(self) -> dict: +- """刷新 (F5)""" +- return self.press('f5') +- +- def switch_window(self) -> dict: +- """切换窗口 (Alt+Tab)""" +- return self.hotkey('alt', 'tab') +- +- +-# 创建全局实例，支持直接调用 +-_keyboard_instance = Keyboard() +- +-# 导出便捷函数 +-type_text = _keyboard_instance.type_text +-press = _keyboard_instance.press +-hotkey = _keyboard_instance.hotkey +-key_down = _keyboard_instance.key_down +-key_up = _keyboard_instance.key_up +-hold_key = _keyboard_instance.hold_key +- +-# 导出快捷键函数 +-copy = _keyboard_instance.copy +-paste = _keyboard_instance.paste +-cut = _keyboard_instance.cut +-select_all = _keyboard_instance.select_all +-undo = _keyboard_instance.undo +-redo = _keyboard_instance.redo +-save = _keyboard_instance.save +-find = _keyboard_instance.find +-new_tab = _keyboard_instance.new_tab +-close_tab = _keyboard_instance.close_tab +-refresh = _keyboard_instance.refresh +-switch_window = _keyboard_instance.switch_window +- +- +-if __name__ == "__main__": +- # 简单测试 +- print("测试键盘模块...") +- +- # 测试按键 +- print("按下Esc键...") +- result = press('esc') +- print(f"结果: {result}") +- +- # 测试快捷键 +- print("测试Ctrl+C快捷键...") +- result = copy() +- print(f"结果: {result}") +- +- print("键盘模块测试完成！") +Index: core/computer/code/languages/bash.py +=================================================================== +diff --git a/core/computer/code/languages/bash.py b/core/computer/code/languages/bash.py +deleted file mode 100644 +--- a/core/computer/code/languages/bash.py (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,63 +0,0 @@ +-import queue +-import shutil +-import subprocess +-import threading +-import time +- +-from ..base_language import BaseLanguage +- +- +-class BashLanguage(BaseLanguage): +- def __init__(self): +- super().__init__() +- self.process = None +- +- def is_available(self): +- return shutil.which("bash") is not None +- +- def run(self, code: str): +- message = queue.Queue() +- execution_thread = threading.Thread(target=self._execute, args=(code, message)) +- execution_thread.daemon = True +- execution_thread.start() +- return message +- +- def _execute(self, code: str, message: queue.Queue): +- self.is_running = True +- self.start_time = time.time() +- self.should_stop = False +- +- try: +- self.process = subprocess.Popen( +- ["bash"], +- stdin=subprocess.PIPE, +- stdout=subprocess.PIPE, +- stderr=subprocess.STDOUT, +- text=True, +- bufsize=1, +- universal_newlines=True +- ) +- +- self.process.stdin.write(code) +- self.process.stdin.close() +- +- for line in self.process.stdout: +- message.put({"type": "text", "content": line}) +- +- return_code = self.process.wait() +- message.put({"type": "text", "content": f"Return code: {return_code}"}) +- +- except Exception as e: +- message.put({"type": "error", "content": f"[BashLanguage]Error: {e}"}) +- finally: +- self.is_running = False +- self.process = None +- +- def interrupt(self): +- self.should_stop = True +- if self.process: +- self.process.terminate() +- try: +- self.process.wait(timeout=2) +- except subprocess.TimeoutExpired: +- self.process.kill() +Index: core/computer/code/base_language.py +=================================================================== +diff --git a/core/computer/code/base_language.py b/core/computer/code/base_language.py +deleted file mode 100644 +--- a/core/computer/code/base_language.py (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,19 +0,0 @@ +-import time +- +-class BaseLanguage: +- def __init__(self): +- self.is_running = False +- self.start_time = None +- self.elapsed_time = 0 +- self.should_stop = False +- +- def run(self, code: str): +- raise NotImplementedError("[BaseLanguage]Subclasses must implement this method") +- +- def get_elapsed_time(self): +- if self.is_running and self.start_time: +- return time.time() - self.start_time +- return self.elapsed_time +- +- def interrupt(self): +- self.should_stop = True +\ No newline at end of file +Index: core/computer/README.md +=================================================================== +diff --git a/core/computer/README.md b/core/computer/README.md +deleted file mode 100644 +--- a/core/computer/README.md (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,3 +0,0 @@ +-### 设计 +- +-所有返回，字典由列表或者queue装起来 +\ No newline at end of file +Index: core/llm/code/default_prompt.py +=================================================================== +diff --git a/core/llm/code/default_prompt.py b/core/llm/code/default_prompt.py +deleted file mode 100644 +--- a/core/llm/code/default_prompt.py (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,33 +0,0 @@ +-import getpass +-import platform +- +- +-default_prompt = f""" +-You are Open Interpreter, a world-class programmer that can complete any goal by executing code. +-For advanced requests, start by writing a plan. +-When you execute code, it will be executed **on the user's machine**. The user has given you **full and complete permission** to execute any code necessary to complete the task. +-You can access the internet. Run **any code** to achieve the goal, and if at first you don't succeed, try again and again. +-You can install new packages. +-When a user refers to a filename, they're likely referring to an existing file in the directory you're currently executing code in. +-Write messages to the user in Markdown. +-In general, try to **make plans** with as few steps as possible. +-You should try something, print information about it, then continue from there in tiny, informed steps. +-You will never get it on the first try, and attempting it in one go will often lead to errors you cant see. +-You are capable of **any** task. +-Print your code directly in the output. Do not print any additional information, including code explanation, running guide, etc. +- +-Use the following format to output your code: +- +-```python +-# Your python code here +-print("Result to inspect") +-``` +- +-User's Name: {getpass.getuser()} +-User's OS: {platform.system()} +-Language: {{language}} +-""" +- +-default_prompt_end = "Are we done? If the task is finished, print \"The task is done.\" in the first line, then summarize the result. If the task is not finished yet, what's next?" +-# As for actually executing code to carry out that plan, for *stateful* languages (like python, javascript, shell, but NOT for html which starts from 0 every time) **it's critical not to try to do everything in one code block.** +-# If the task is impossible, print \"The task is impossible.\" in the first line, then explain why. +\ No newline at end of file +Index: lin.py +=================================================================== +diff --git a/lin.py b/lin.py +deleted file mode 100644 +--- a/lin.py (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,3 +0,0 @@ +-import numpy +- +-ss +\ No newline at end of file +Index: .gitignore +=================================================================== +diff --git a/.gitignore b/.gitignore +deleted file mode 100644 +--- a/.gitignore (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,7 +0,0 @@ +-.idea +-.vscode +-__pycache__ +-test +- +-*.log +-.env* +\ No newline at end of file +Index: requirements.txt +=================================================================== +diff --git a/requirements.txt b/requirements.txt +deleted file mode 100644 +--- a/requirements.txt (revision 4057a203831344e54e570576815336c422e4fe9f) ++++ /dev/null (revision 4057a203831344e54e570576815336c422e4fe9f) +@@ -1,3 +0,0 @@ +-litellm +-jupyter_client +-pyautogui +\ No newline at end of file +diff --git a/core/llm/gui/__init__.py b/core/llm/gui/__init__.py +deleted file mode 100644 +diff --git a/core/__init__.py b/core/__init__.py +deleted file mode 100644 +diff --git a/core/computer/__init__.py b/core/computer/__init__.py +deleted file mode 100644 diff --git a/core/__init__.py b/core/__init__.py deleted file mode 100644 index e69de29..0000000 diff --git a/core/computer/README.md b/core/computer/README.md deleted file mode 100644 index ee0d150..0000000 --- a/core/computer/README.md +++ /dev/null @@ -1,3 +0,0 @@ -### 设计 - -所有返回，字典由列表或者queue装起来 \ No newline at end of file diff --git a/core/computer/__init__.py b/core/computer/__init__.py deleted file mode 100644 index e69de29..0000000 diff --git a/core/computer/code/README.md b/core/computer/code/README.md deleted file mode 100644 index e1a9f23..0000000 --- a/core/computer/code/README.md +++ /dev/null @@ -1,214 +0,0 @@ -# Code执行模块 - -这是一个多语言代码执行模块，支持Python、Bash和PowerShell代码的异步执行。 - -## 功能概述 - -该模块提供了一个统一的接口来执行不同编程语言的代码，并通过队列返回执行结果。支持代码中断、执行时间监控等功能。 - -## 架构设计 - -### 核心组件 - -- **Code类**: 主要的代码执行管理器，负责语言选择和执行调度 -- **BaseLanguage类**: 所有语言实现的抽象基类 -- **PythonLanguage类**: Python代码执行器，基于Jupyter内核 -- **BashLanguage类**: Bash/Shell代码执行器，基于subprocess -- **PowerShellLanguage类**: PowerShell代码执行器，基于subprocess - -### 消息系统 - -所有执行结果通过`queue.Queue`对象返回，消息格式为字典： - -```python -{ - "type": "消息类型", - "content": "消息内容" -} -``` - -## run()返回的消息格式 - -- **`text`**: 纯文本输出，包括标准输出、错误输出、执行结果等 -- **`image/png`**: PNG格式的图片数据（base64编码） -- **`image/jpeg`**: JPEG格式的图片数据（base64编码） -- **`html`**: HTML格式的内容 -- **`text/plain`**: 纯文本内容 -- **`application/javascript`**: JavaScript代码 -- **`error`**: 模块的错误信息，不包含代码运行的错误 - -## 使用方法 - -### 基本用法 - -```python -from core.computer.code import Code - -# 创建代码执行器 -code_executor = Code() - -# 执行Python代码 -result_queue = code_executor.run("python", "print('Hello, World!')") - -# 执行Bash代码 -result_queue = code_executor.run("bash", "ls -la") - -# 执行PowerShell代码 -result_queue = code_executor.run("powershell", "Get-ChildItem") - -# 获取执行结果 -while not result_queue.empty(): - message = result_queue.get() - print(f"{message['type']}: {message['content']}") -``` - -### 支持的语言 - -- Python: `python` -- Bash: `bash`, `sh` -- PowerShell: `powershell`, `pwsh` - -### 中断执行 - -```python -# 中断当前正在执行的代码 -result_queue = code_executor.interrupt() -``` - -### 监控执行状态 - -```python -# 检查是否有代码正在运行 -is_running = code_executor.is_running() - -# 获取已运行时间 -elapsed_time = code_executor.get_elapsed_time() -``` - -## 各语言实现详情 - -### Python实现 (PythonLanguage) - -- **技术**: 基于Jupyter内核和jupyter_client -- **特性**: - - 支持matplotlib内联绘图 - - 支持图片、HTML、JavaScript等多种输出格式 - - 完整的Python语法支持 - - 自动添加`%matplotlib inline`以支持可视化 -- **依赖**: jupyter_client, ipython kernel - -### Bash实现 (BashLanguage) - -- **技术**: 基于subprocess.Popen -- **特性**: - - 检测bash是否可用 - - 实时输出流 - - 支持信号中断 -- **可用性检查**: `shutil.which("bash")` - -### PowerShell实现 (PowerShellLanguage) - -- **技术**: 基于subprocess.Popen -- **特性**: - - 优先使用pwsh (PowerShell Core)，回退到powershell - - 跨平台支持 - - 无配置文件执行 (`-NoProfile`) -- **可用性检查**: `shutil.which("powershell") or shutil.which("pwsh")` - -## 错误处理 - -### 模块级错误 - -- 不支持的语言 -- 系统环境问题 -- 内核启动失败 - -### 代码执行错误 - -- 语法错误 -- 运行时错误 -- 依赖缺失 - -## 线程安全 - -- 每次代码执行都在独立的线程中运行 -- 使用队列进行线程间通信 -- 支持优雅的中断机制 - -## 性能考虑 - -- Python内核会重复启动和停止，适合短时间执行 -- Bash和PowerShell为每次执行创建新进程 -- 所有语言都支持实时输出，不会阻塞主线程 - -## 依赖要求 - -### Python环境 -- Python 3.7+ -- jupyter_client -- IPython kernel - -### 系统环境 -- Bash: Git Bash, WSL, 或Linux/macOS环境 -- PowerShell: Windows PowerShell 5.1+ 或 PowerShell Core 6.0+ - -## 示例代码 - -```python -# 完整示例 -from core.computer.code import Code -import time - -def execute_code_example(): - executor = Code() - - # Python可视化示例 - print("执行Python代码...") - result = executor.run("python", """ -import matplotlib.pyplot as plt -import numpy as np - -x = np.linspace(0, 10, 100) -y = np.sin(x) - -plt.figure(figsize=(8, 4)) -plt.plot(x, y) -plt.title('正弦波形') -plt.xlabel('x') -plt.ylabel('sin(x)') -plt.grid(True) -plt.show() -""") - - # 读取结果 - while not result.empty() or executor.is_running(): - if not result.empty(): - msg = result.get() - print(f"收到消息: {msg['type']}") - if msg['type'].startswith('image/'): - print(f"图片数据: {len(msg['content'])} 字节") - else: - print(f"内容: {msg['content'][:100]}...") - time.sleep(0.1) - - print("Python执行完成") - -if __name__ == "__main__": - execute_code_example() -``` - -## 注意事项 - -1. **Python内核**: 每次执行都会重新启动内核，状态不会保留 -2. **安全性**: 该模块设计用于受控环境，避免执行不受信任的代码 -3. **资源管理**: 长时间运行可能产生大量进程，建议监控资源使用 -4. **平台兼容性**: 某些功能在不同操作系统上表现可能不同 - -## 未来改进 - -- [ ] 支持更多编程语言 (Node.js, Ruby等) -- [ ] Python内核持久化，避免重复启动 -- [ ] 代码执行超时设置 -- [ ] 更细粒度的权限控制 -- [ ] 代码执行历史记录 -- [ ] 性能监控和分析 diff --git a/core/computer/code/__init__.py b/core/computer/code/__init__.py deleted file mode 100644 index 49f78e3..0000000 --- a/core/computer/code/__init__.py +++ /dev/null @@ -1,11 +0,0 @@ -""" -代码运行包 -""" - - -from .base_language import BaseLanguage -from .code import Code - -__all__ = ['Code', 'BaseLanguage'] - -__version__ = '1.0.0' \ No newline at end of file diff --git a/core/computer/code/base_language.py b/core/computer/code/base_language.py deleted file mode 100644 index 26ffb40..0000000 --- a/core/computer/code/base_language.py +++ /dev/null @@ -1,19 +0,0 @@ -import time - -class BaseLanguage: - def __init__(self): - self.is_running = False - self.start_time = None - self.elapsed_time = 0 - self.should_stop = False - - def run(self, code: str): - raise NotImplementedError("[BaseLanguage]Subclasses must implement this method") - - def get_elapsed_time(self): - if self.is_running and self.start_time: - return time.time() - self.start_time - return self.elapsed_time - - def interrupt(self): - self.should_stop = True \ No newline at end of file diff --git a/core/computer/code/code.py b/core/computer/code/code.py deleted file mode 100644 index eab47c8..0000000 --- a/core/computer/code/code.py +++ /dev/null @@ -1,74 +0,0 @@ -import queue - -from .languages import PythonLanguage, BashLanguage, PowerShellLanguage - - -class Code: - def __init__(self): - self.python = PythonLanguage() - # self.cmd = CMDLanguage() - self.bash = BashLanguage() - self.powershell = PowerShellLanguage() - self.current_language = None - self.language_map = { - "python": self.python, - # "cmd": self.cmd, - "powershell": self.powershell, - "pwsh": self.powershell, # PowerShell Core别名 - "bash": self.bash, - "sh": self.bash, # Shell别名 - } - self.language_list = [] - for lang, lang_obj in self.language_map.items(): - if hasattr(lang_obj, 'is_available'): - if lang_obj.is_available(): - self.language_list.append(lang) - else: - # Python总是可用的 - self.language_list.append(lang) - - # def language_list(self): - # available = [] - - def run(self, lang: str, code: str): - """ - 运行代码 - - :param lang: 要运行的代码语言 - :type lang: str - :param code: 要运行的代码 - :type code: str - """ - lang = lang.lower() - if lang in self.language_list: - self.current_language = self.language_map[lang] - return self.current_language.run(code) - else: - message = queue.Queue() - message.put({"type": "error", "content": f"[Code]Unsupported language:{lang}"}) - return message - - def interrupt(self): - """ - 中断执行中的代码 - """ - message = queue.Queue() - if self.current_language and self.current_language.is_running: - self.current_language.interrupt() - message.put({"type": "text", "content": "Code execution interrupted"}) - else: - message.put({"type": "text", "content": "No code is running"}) - return message - - def get_elapsed_time(self): - """ - 获取代码已运行时间 - """ - if self.current_language: - return self.current_language.get_elapsed_time() - - def is_running(self): - """ - 代码是否正在运行 - """ - return self.current_language and self.current_language.is_running \ No newline at end of file diff --git a/core/computer/code/languages/__init__.py b/core/computer/code/languages/__init__.py deleted file mode 100644 index 2f0da07..0000000 --- a/core/computer/code/languages/__init__.py +++ /dev/null @@ -1,12 +0,0 @@ -""" - -""" - - -from .bash import BashLanguage -from .powershell import PowerShellLanguage -from .python import PythonLanguage - -__all__ = ['PythonLanguage', 'BashLanguage', 'PowerShellLanguage'] - -__version__ = '1.0.0' \ No newline at end of file diff --git a/core/computer/code/languages/bash.py b/core/computer/code/languages/bash.py deleted file mode 100644 index fde830e..0000000 --- a/core/computer/code/languages/bash.py +++ /dev/null @@ -1,63 +0,0 @@ -import queue -import shutil -import subprocess -import threading -import time - -from ..base_language import BaseLanguage - - -class BashLanguage(BaseLanguage): - def __init__(self): - super().__init__() - self.process = None - - def is_available(self): - return shutil.which("bash") is not None - - def run(self, code: str): - message = queue.Queue() - execution_thread = threading.Thread(target=self._execute, args=(code, message)) - execution_thread.daemon = True - execution_thread.start() - return message - - def _execute(self, code: str, message: queue.Queue): - self.is_running = True - self.start_time = time.time() - self.should_stop = False - - try: - self.process = subprocess.Popen( - ["bash"], - stdin=subprocess.PIPE, - stdout=subprocess.PIPE, - stderr=subprocess.STDOUT, - text=True, - bufsize=1, - universal_newlines=True - ) - - self.process.stdin.write(code) - self.process.stdin.close() - - for line in self.process.stdout: - message.put({"type": "text", "content": line}) - - return_code = self.process.wait() - message.put({"type": "text", "content": f"Return code: {return_code}"}) - - except Exception as e: - message.put({"type": "error", "content": f"[BashLanguage]Error: {e}"}) - finally: - self.is_running = False - self.process = None - - def interrupt(self): - self.should_stop = True - if self.process: - self.process.terminate() - try: - self.process.wait(timeout=2) - except subprocess.TimeoutExpired: - self.process.kill() diff --git a/core/computer/code/languages/powershell.py b/core/computer/code/languages/powershell.py deleted file mode 100644 index 20427f4..0000000 --- a/core/computer/code/languages/powershell.py +++ /dev/null @@ -1,64 +0,0 @@ -import queue -import shutil -import subprocess -import threading -import time - -from ..base_language import BaseLanguage - - -class PowerShellLanguage(BaseLanguage): - def __init__(self): - super().__init__() - self.process = None - - def is_available(self): - return shutil.which("powershell") is not None or shutil.which("pwsh") is not None - - def run(self, code: str): - message = queue.Queue() - execution_thread = threading.Thread(target=self._execute, args=(code, message)) - execution_thread.daemon = True - execution_thread.start() - return message - - def _execute(self, code: str, message: queue.Queue): - self.is_running = True - self.start_time = time.time() - self.should_stop = False - - executable = "pwsh" if shutil.which("pwsh") else "powershell" - - try: - # -Command - tells powershell to read command from stdin - self.process = subprocess.Popen( - [executable, "-NoProfile", "-Command", code], - stdout=subprocess.PIPE, - stderr=subprocess.STDOUT, - text=True, - bufsize=1, - universal_newlines=True - ) - - for line in self.process.stdout: - if self.should_stop: - break - message.put({"type": "text", "content": line}) - - return_code = self.process.wait() - message.put({"type": "text", "content": f"Return code: {return_code}"}) - - except Exception as e: - message.put({"type": "error", "content": f"[PowerShellLanguage]Error: {e}"}) - finally: - self.is_running = False - self.process = None - - def interrupt(self): - super().interrupt() - if self.process: - self.process.terminate() - try: - self.process.wait(timeout=2) - except subprocess.TimeoutExpired: - self.process.kill() diff --git a/core/computer/code/languages/python.py b/core/computer/code/languages/python.py deleted file mode 100644 index 53745c0..0000000 --- a/core/computer/code/languages/python.py +++ /dev/null @@ -1,131 +0,0 @@ -import logging -import queue -import re -import threading -import time - -from jupyter_client.manager import KernelManager - -from ..base_language import BaseLanguage - - -class PythonLanguage(BaseLanguage): - def __init__(self): - super().__init__() - self.km = None - self.kc = None - self.current_msg_id = None - - def start(self): - self.should_stop = False - self.start_time = time.time() - self.elapsed_time = 0 - - if self.km: - self.stop() - try: - self.km = KernelManager(kernel_name='python3') - self.is_running = True - self.km.start_kernel() - self.kc = self.km.client() - self.kc.start_channels() - self.kc.wait_for_ready() - logging.info("[PythonLanguage]Started kernel client&manager") - except Exception as e: - self.stop() - logging.error("[PythonLanguage]Error starting kernel: %s", e) - - def stop(self): - self.is_running = False - if self.start_time: - self.elapsed_time = time.time() - self.start_time - self.start_time = None - - try: - if self.km: - self.km.shutdown_kernel() - self.is_running = False - self.km = None - if self.kc: - self.kc.stop_channels() - self.current_msg_id = None - self.kc = None - logging.info("[PythonLanguage]Stopped kernel client&manager") - except Exception as e: - logging.error("[PythonLanguage]Error during cleanup kernel: %s", e) - - def wait_for_shutdown(self): - while self.is_running or self.kc: - time.sleep(0.1) - - def run(self, code: str): - message = queue.Queue() - - excecution_thread = threading.Thread(target=self._execute_jupyter, args=(code, message)) - excecution_thread.daemon = True - excecution_thread.start() - - return message - - def _execute_jupyter(self, code: str, message: queue.Queue): - self.start() - - if self.km.is_alive() is False: - message.put({"type": "error", "content": "[PythonLanguage]Faild to start Jupyter kernel"}) - return - - # 只在需要时添加matplotlib魔法命令 - if "matplotlib" in code and "%matplotlib" not in code: - code = "%matplotlib inline\n" + code - - try: - self.current_msg_id=self.kc.execute(code) - except Exception as e: - message.put({"type": "error", "content": f"[PythonLanguage]Error while executing code: {e}"}) - logging.error("[PythonLanguage]Error while executing code: %s", e) - print("[PythonLanguage]start runing...") - while True: - if self.should_stop: - self.km.interrupt_kernel() - - try: - msg = self.kc.get_iopub_msg(timeout=1) - except queue.Empty: - continue - - if msg['parent_header'].get('msg_id') != self.current_msg_id: - continue - - msg_type = msg['msg_type'] - content = msg['content'] - - print("get_one_msg") - - if msg_type == "stream": - message.put({"type": "text", "content": content['text']}) - - elif msg_type == "error": - content = "\n".join(content["traceback"]) - # 移除颜色标识 - ansi_escape = re.compile(r"\x1B\[[0-?]*[ -/]*[@-~]") - content = ansi_escape.sub("", content) - message.put({"type": "text", "content": content}) - - elif msg_type in ["display_data", "execute_result"]: - data = content["data"] - if "image/png" in data: - message.put({"type": "image/png", "content": data["image/png"]}) - elif "image/jpeg" in data: - message.put({"type": "image/jpeg", "content": data["image/jpeg"]}) - elif "text/html" in data: - message.put({"type": "html", "content": data["text/html"]}) - elif "text/plain" in data: - message.put({"type": "text", "content": data["text/plain"]}) - elif "application/javascript" in data: - message.put({"type": "javascript", "content": data["application/javascript"]}) - - elif msg_type == 'status': - if content['execution_state'] == 'idle': - break - print("[PythonLanguage]finished.") - self.stop() diff --git a/core/computer/keyboard/__init__.py b/core/computer/keyboard/__init__.py deleted file mode 100644 index 2c19937..0000000 --- a/core/computer/keyboard/__init__.py +++ /dev/null @@ -1,47 +0,0 @@ -""" -键盘操作包 -""" -from .keyboard import ( - Keyboard, - type_text, - press, - hotkey, - key_down, - key_up, - hold_key, - copy, - paste, - cut, - select_all, - undo, - redo, - save, - find, - new_tab, - close_tab, - refresh, - switch_window -) - -__all__ = [ - # Keyboard类和函数 - 'Keyboard', - 'type_text', - 'press', - 'hotkey', - 'key_down', - 'key_up', - 'hold_key', - 'copy', - 'paste', - 'cut', - 'select_all', - 'undo', - 'redo', - 'save', - 'find', - 'new_tab', - 'close_tab', - 'refresh', - 'switch_window', -] diff --git a/core/computer/keyboard/keyboard.py b/core/computer/keyboard/keyboard.py deleted file mode 100644 index 0c71680..0000000 --- a/core/computer/keyboard/keyboard.py +++ /dev/null @@ -1,296 +0,0 @@ -import pyautogui -import time -from typing import Union, List, Optional - - -class Keyboard: - """键盘操作类，提供各种键盘控制功能""" - - def __init__(self): - """初始化键盘模块""" - pyautogui.PAUSE = 0.05 # 每次操作后暂停0.05秒 - - def type_text(self, text: str, interval: float = 0.0) -> dict: - """ - 输入文本 - - Args: - text: 要输入的文本 - interval: 每个字符之间的间隔时间（秒） - - Returns: - 包含操作结果的字典 - - Examples: - >>> keyboard.type_text("Hello World") - >>> keyboard.type_text("慢速输入", interval=0.1) - - Note: - 对于中文输入，需要系统已经切换到中文输入法 - """ - try: - pyautogui.write(text, interval=interval) - return { - 'success': True, - 'action': 'type_text', - 'text': text, - 'length': len(text), - 'interval': interval - } - except Exception as e: - return { - 'success': False, - 'action': 'type_text', - 'error': str(e) - } - - def press(self, key: str, presses: int = 1, interval: float = 0.0) -> dict: - """ - 按下指定键 - - Args: - key: 键名，如 'enter', 'space', 'a', 'ctrl' 等 - presses: 按键次数 - interval: 多次按键之间的间隔（秒） - - Returns: - 包含操作结果的字典 - - Examples: - >>> keyboard.press('enter') # 按下回车键 - >>> keyboard.press('tab', presses=3) # 按3次Tab键 - >>> keyboard.press('backspace') # 按退格键 - - 常用键名: - - 字母: 'a', 'b', 'c' ... - - 数字: '0', '1', '2' ... - - 功能键: 'f1', 'f2' ... 'f12' - - 方向键: 'up', 'down', 'left', 'right' - - 特殊键: 'enter', 'space', 'tab', 'backspace', 'delete', 'esc' - - 修饰键: 'ctrl', 'shift', 'alt', 'win' (或 'command' 在Mac上) - """ - try: - pyautogui.press(key, presses=presses, interval=interval) - return { - 'success': True, - 'action': 'press', - 'key': key, - 'presses': presses, - 'interval': interval - } - except Exception as e: - return { - 'success': False, - 'action': 'press', - 'error': str(e) - } - - def hotkey(self, *keys: str) -> dict: - """ - 按下组合键（快捷键） - - Args: - *keys: 要同时按下的键，按顺序传入 - - Returns: - 包含操作结果的字典 - - Examples: - >>> keyboard.hotkey('ctrl', 'c') # Ctrl+C 复制 - >>> keyboard.hotkey('ctrl', 'v') # Ctrl+V 粘贴 - >>> keyboard.hotkey('ctrl', 'shift', 's') # Ctrl+Shift+S - >>> keyboard.hotkey('alt', 'tab') # Alt+Tab 切换窗口 - >>> keyboard.hotkey('ctrl', 'a') # Ctrl+A 全选 - """ - try: - pyautogui.hotkey(*keys) - return { - 'success': True, - 'action': 'hotkey', - 'keys': keys, - 'combination': '+'.join(keys) - } - except Exception as e: - return { - 'success': False, - 'action': 'hotkey', - 'error': str(e) - } - - def key_down(self, key: str) -> dict: - """ - 按下键（不释放） - - Args: - key: 键名 - - Returns: - 包含操作结果的字典 - - Examples: - >>> keyboard.key_down('shift') # 按下Shift键 - """ - try: - pyautogui.keyDown(key) - return { - 'success': True, - 'action': 'key_down', - 'key': key - } - except Exception as e: - return { - 'success': False, - 'action': 'key_down', - 'error': str(e) - } - - def key_up(self, key: str) -> dict: - """ - 释放键 - - Args: - key: 键名 - - Returns: - 包含操作结果的字典 - - Examples: - >>> keyboard.key_up('shift') # 释放Shift键 - """ - try: - pyautogui.keyUp(key) - return { - 'success': True, - 'action': 'key_up', - 'key': key - } - except Exception as e: - return { - 'success': False, - 'action': 'key_up', - 'error': str(e) - } - - def hold_key(self, key: str, duration: float = 1.0) -> dict: - """ - 按住一个键一段时间 - - Args: - key: 键名 - duration: 按住的时间（秒） - - Returns: - 包含操作结果的字典 - - Examples: - >>> keyboard.hold_key('space', duration=2.0) # 按住空格键2秒 - """ - try: - pyautogui.keyDown(key) - time.sleep(duration) - pyautogui.keyUp(key) - return { - 'success': True, - 'action': 'hold_key', - 'key': key, - 'duration': duration - } - except Exception as e: - return { - 'success': False, - 'action': 'hold_key', - 'error': str(e) - } - - # 常用快捷键的便捷方法 - - def copy(self) -> dict: - """复制 (Ctrl+C)""" - return self.hotkey('ctrl', 'c') - - def paste(self) -> dict: - """粘贴 (Ctrl+V)""" - return self.hotkey('ctrl', 'v') - - def cut(self) -> dict: - """剪切 (Ctrl+X)""" - return self.hotkey('ctrl', 'x') - - def select_all(self) -> dict: - """全选 (Ctrl+A)""" - return self.hotkey('ctrl', 'a') - - def undo(self) -> dict: - """撤销 (Ctrl+Z)""" - return self.hotkey('ctrl', 'z') - - def redo(self) -> dict: - """重做 (Ctrl+Y)""" - return self.hotkey('ctrl', 'y') - - def save(self) -> dict: - """保存 (Ctrl+S)""" - return self.hotkey('ctrl', 's') - - def find(self) -> dict: - """查找 (Ctrl+F)""" - return self.hotkey('ctrl', 'f') - - def new_tab(self) -> dict: - """新建标签页 (Ctrl+T)""" - return self.hotkey('ctrl', 't') - - def close_tab(self) -> dict: - """关闭标签页 (Ctrl+W)""" - return self.hotkey('ctrl', 'w') - - def refresh(self) -> dict: - """刷新 (F5)""" - return self.press('f5') - - def switch_window(self) -> dict: - """切换窗口 (Alt+Tab)""" - return self.hotkey('alt', 'tab') - - -# 创建全局实例，支持直接调用 -_keyboard_instance = Keyboard() - -# 导出便捷函数 -type_text = _keyboard_instance.type_text -press = _keyboard_instance.press -hotkey = _keyboard_instance.hotkey -key_down = _keyboard_instance.key_down -key_up = _keyboard_instance.key_up -hold_key = _keyboard_instance.hold_key - -# 导出快捷键函数 -copy = _keyboard_instance.copy -paste = _keyboard_instance.paste -cut = _keyboard_instance.cut -select_all = _keyboard_instance.select_all -undo = _keyboard_instance.undo -redo = _keyboard_instance.redo -save = _keyboard_instance.save -find = _keyboard_instance.find -new_tab = _keyboard_instance.new_tab -close_tab = _keyboard_instance.close_tab -refresh = _keyboard_instance.refresh -switch_window = _keyboard_instance.switch_window - - -if __name__ == "__main__": - # 简单测试 - print("测试键盘模块...") - - # 测试按键 - print("按下Esc键...") - result = press('esc') - print(f"结果: {result}") - - # 测试快捷键 - print("测试Ctrl+C快捷键...") - result = copy() - print(f"结果: {result}") - - print("键盘模块测试完成！") diff --git a/core/computer/mouse/__init__.py b/core/computer/mouse/__init__.py deleted file mode 100644 index 4abfea7..0000000 --- a/core/computer/mouse/__init__.py +++ /dev/null @@ -1,33 +0,0 @@ -""" -鼠标操作包 -""" -from .mouse import ( - Mouse, - click, - double_click, - right_click, - move, - move_rel, - drag, - drag_rel, - scroll, - get_position, - mouse_down, - mouse_up -) - -__all__ = [ - # Mouse类和函数 - 'Mouse', - 'click', - 'double_click', - 'right_click', - 'move', - 'move_rel', - 'drag', - 'drag_rel', - 'scroll', - 'get_position', - 'mouse_down', - 'mouse_up', -] \ No newline at end of file diff --git a/core/computer/mouse/mouse.py b/core/computer/mouse/mouse.py deleted file mode 100644 index eb1ee10..0000000 --- a/core/computer/mouse/mouse.py +++ /dev/null @@ -1,368 +0,0 @@ -import pyautogui -import time -from typing import Tuple, Optional, Union - - -class Mouse: - """鼠标操作类，提供各种鼠标控制功能""" - - def __init__(self): - """初始化鼠标模块""" - # 设置PyAutoGUI的安全特性 - pyautogui.FAILSAFE = True # 移动鼠标到屏幕左上角会触发异常 - pyautogui.PAUSE = 0.1 # 每次操作后暂停0.1秒 - - def click(self, x: Optional[int] = None, y: Optional[int] = None, - clicks: int = 1, interval: float = 0.0, - button: str = 'left', duration: float = 0.0) -> dict: - """ - 鼠标点击操作 - - Args: - x: X坐标，None表示当前位置 - y: Y坐标，None表示当前位置 - clicks: 点击次数，默认1次 - interval: 多次点击之间的间隔（秒） - button: 鼠标按键 'left'/'right'/'middle' - duration: 移动到目标位置的持续时间（秒） - - Returns: - 包含操作结果的字典 - - Examples: - >>> mouse.click(100, 200) # 点击坐标(100, 200) - >>> mouse.click() # 点击当前位置 - >>> mouse.click(button='right') # 右键点击当前位置 - """ - try: - if x is not None and y is not None: - pyautogui.click(x, y, clicks=clicks, interval=interval, - button=button, duration=duration) - else: - pyautogui.click(clicks=clicks, interval=interval, button=button) - - current_pos = pyautogui.position() - return { - 'success': True, - 'action': 'click', - 'position': (current_pos.x, current_pos.y), - 'button': button, - 'clicks': clicks - } - except Exception as e: - return { - 'success': False, - 'action': 'click', - 'error': str(e) - } - - def double_click(self, x: Optional[int] = None, y: Optional[int] = None, - button: str = 'left', duration: float = 0.0) -> dict: - """ - 鼠标双击操作 - - Args: - x: X坐标，None表示当前位置 - y: Y坐标，None表示当前位置 - button: 鼠标按键 'left'/'right'/'middle' - duration: 移动到目标位置的持续时间（秒） - - Returns: - 包含操作结果的字典 - - Examples: - >>> mouse.double_click(300, 400) # 双击坐标(300, 400) - """ - return self.click(x, y, clicks=2, button=button, duration=duration) - - def right_click(self, x: Optional[int] = None, y: Optional[int] = None, - duration: float = 0.0) -> dict: - """ - 鼠标右键点击 - - Args: - x: X坐标，None表示当前位置 - y: Y坐标，None表示当前位置 - duration: 移动到目标位置的持续时间（秒） - - Returns: - 包含操作结果的字典 - - Examples: - >>> mouse.right_click(500, 300) # 右键点击坐标(500, 300) - """ - return self.click(x, y, button='right', duration=duration) - - def move(self, x: int, y: int, duration: float = 0.0) -> dict: - """ - 移动鼠标到指定位置 - - Args: - x: 目标X坐标 - y: 目标Y坐标 - duration: 移动持续时间（秒），0表示立即移动 - - Returns: - 包含操作结果的字典 - - Examples: - >>> mouse.move(800, 600) # 立即移动到(800, 600) - >>> mouse.move(800, 600, duration=1.0) # 用1秒时间移动到(800, 600) - """ - try: - pyautogui.moveTo(x, y, duration=duration) - current_pos = pyautogui.position() - return { - 'success': True, - 'action': 'move', - 'position': (current_pos.x, current_pos.y), - 'duration': duration - } - except Exception as e: - return { - 'success': False, - 'action': 'move', - 'error': str(e) - } - - def move_rel(self, x_offset: int, y_offset: int, duration: float = 0.0) -> dict: - """ - 相对移动鼠标（从当前位置偏移） - - Args: - x_offset: X轴偏移量 - y_offset: Y轴偏移量 - duration: 移动持续时间（秒） - - Returns: - 包含操作结果的字典 - - Examples: - >>> mouse.move_rel(100, 0) # 向右移动100像素 - """ - try: - pyautogui.moveRel(x_offset, y_offset, duration=duration) - current_pos = pyautogui.position() - return { - 'success': True, - 'action': 'move_rel', - 'position': (current_pos.x, current_pos.y), - 'offset': (x_offset, y_offset), - 'duration': duration - } - except Exception as e: - return { - 'success': False, - 'action': 'move_rel', - 'error': str(e) - } - - def drag(self, x: int, y: int, duration: float = 0.5, - button: str = 'left') -> dict: - """ - 拖拽鼠标到指定位置 - - Args: - x: 目标X坐标 - y: 目标Y坐标 - duration: 拖拽持续时间（秒） - button: 使用的鼠标按键 - - Returns: - 包含操作结果的字典 - - Examples: - >>> mouse.drag(500, 500, duration=1.0) # 拖拽到(500, 500) - """ - try: - pyautogui.drag(x, y, duration=duration, button=button) - current_pos = pyautogui.position() - return { - 'success': True, - 'action': 'drag', - 'position': (current_pos.x, current_pos.y), - 'button': button, - 'duration': duration - } - except Exception as e: - return { - 'success': False, - 'action': 'drag', - 'error': str(e) - } - - def drag_rel(self, x_offset: int, y_offset: int, duration: float = 0.5, - button: str = 'left') -> dict: - """ - 相对拖拽（从当前位置偏移） - - Args: - x_offset: X轴偏移量 - y_offset: Y轴偏移量 - duration: 拖拽持续时间（秒） - button: 使用的鼠标按键 - - Returns: - 包含操作结果的字典 - - Examples: - >>> mouse.drag_rel(100, 50) # 向右下拖拽 - """ - try: - pyautogui.dragRel(x_offset, y_offset, duration=duration, button=button) - current_pos = pyautogui.position() - return { - 'success': True, - 'action': 'drag_rel', - 'position': (current_pos.x, current_pos.y), - 'offset': (x_offset, y_offset), - 'button': button, - 'duration': duration - } - except Exception as e: - return { - 'success': False, - 'action': 'drag_rel', - 'error': str(e) - } - - def scroll(self, clicks: int, x: Optional[int] = None, - y: Optional[int] = None) -> dict: - """ - 滚动鼠标滚轮 - - Args: - clicks: 滚动单位数，正数向上滚动，负数向下滚动 - x: 在指定位置滚动的X坐标（可选） - y: 在指定位置滚动的Y坐标（可选） - - Returns: - 包含操作结果的字典 - - Examples: - >>> mouse.scroll(5) # 向上滚动5个单位 - >>> mouse.scroll(-3) # 向下滚动3个单位 - >>> mouse.scroll(5, 500, 500) # 在(500, 500)位置向上滚动 - """ - try: - if x is not None and y is not None: - pyautogui.scroll(clicks, x, y) - else: - pyautogui.scroll(clicks) - - return { - 'success': True, - 'action': 'scroll', - 'clicks': clicks, - 'position': (x, y) if x and y else 'current' - } - except Exception as e: - return { - 'success': False, - 'action': 'scroll', - 'error': str(e) - } - - def get_position(self) -> dict: - """ - 获取当前鼠标位置 - - Returns: - 包含鼠标位置的字典 - - Examples: - >>> pos = mouse.get_position() - >>> print(f"鼠标位置: ({pos['x']}, {pos['y']})") - """ - try: - pos = pyautogui.position() - return { - 'success': True, - 'x': pos.x, - 'y': pos.y, - 'position': (pos.x, pos.y) - } - except Exception as e: - return { - 'success': False, - 'error': str(e) - } - - def mouse_down(self, x: Optional[int] = None, y: Optional[int] = None, - button: str = 'left') -> dict: - """ - 按下鼠标按键（不释放） - - Args: - x: X坐标（可选） - y: Y坐标（可选） - button: 鼠标按键 - - Returns: - 包含操作结果的字典 - """ - try: - if x is not None and y is not None: - pyautogui.mouseDown(x, y, button=button) - else: - pyautogui.mouseDown(button=button) - - return { - 'success': True, - 'action': 'mouse_down', - 'button': button - } - except Exception as e: - return { - 'success': False, - 'action': 'mouse_down', - 'error': str(e) - } - - def mouse_up(self, x: Optional[int] = None, y: Optional[int] = None, - button: str = 'left') -> dict: - """ - 释放鼠标按键 - - Args: - x: X坐标（可选） - y: Y坐标（可选） - button: 鼠标按键 - - Returns: - 包含操作结果的字典 - """ - try: - if x is not None and y is not None: - pyautogui.mouseUp(x, y, button=button) - else: - pyautogui.mouseUp(button=button) - - return { - 'success': True, - 'action': 'mouse_up', - 'button': button - } - except Exception as e: - return { - 'success': False, - 'action': 'mouse_up', - 'error': str(e) - } - - -# 创建全局实例，支持直接调用 -_mouse_instance = Mouse() - -# 导出便捷函数 -click = _mouse_instance.click -double_click = _mouse_instance.double_click -right_click = _mouse_instance.right_click -move = _mouse_instance.move -move_rel = _mouse_instance.move_rel -drag = _mouse_instance.drag -drag_rel = _mouse_instance.drag_rel -scroll = _mouse_instance.scroll -get_position = _mouse_instance.get_position -mouse_down = _mouse_instance.mouse_down -mouse_up = _mouse_instance.mouse_up \ No newline at end of file diff --git a/core/computer/screen/__init__.py b/core/computer/screen/__init__.py deleted file mode 100644 index 4e5eb6a..0000000 --- a/core/computer/screen/__init__.py +++ /dev/null @@ -1,6 +0,0 @@ -from .screen import Screen - - -__all__ = [Screen] - -__version__ = '1.0.0' \ No newline at end of file diff --git a/core/computer/screen/screen.py b/core/computer/screen/screen.py deleted file mode 100644 index eb5521d..0000000 --- a/core/computer/screen/screen.py +++ /dev/null @@ -1,35 +0,0 @@ -import base64 -import io - -from PIL import Image, ImageGrab - - -class Screen: - def __init__(self): - pass - def screenshot(self, resize_factor: float = 0.5, format: str= "png", quality: int = 100): - """ - 获取截屏 - :param resize_factor: 缩放因子 - :param format: 格式，可选png/jpeg - :param quality: 图片质量，jpeg有效 - :return: [{"type": f"image/{format}", "content": base64_encoded}] - """ - image = ImageGrab.grab() - new_width = int(image.size[0] * resize_factor) - new_height = int(image.size[1] * resize_factor) - image = image.resize((new_width, new_height), Image.LANCZOS) - - img_byte_arr= io.BytesIO() - format = format.lower() - if format == "png": - image.save(img_byte_arr, format=format) - elif format == "jpeg": - image.save(img_byte_arr, format=format, quality=quality) - else: - raise ValueError("[Screen]Unsupported format: ", format) - - img_bytes = img_byte_arr.getvalue() - base64_encoded = base64.b64encode(img_bytes).decode('utf-8') - - return [{"type": f"image/{format}", "content": base64_encoded}] \ No newline at end of file diff --git a/core/llm/code/__init__.py b/core/llm/code/__init__.py deleted file mode 100644 index 3dc5561..0000000 --- a/core/llm/code/__init__.py +++ /dev/null @@ -1,3 +0,0 @@ -from .agent import CodeAgent - -__all__ = ["CodeAgent"] \ No newline at end of file diff --git a/core/llm/code/agent.py b/core/llm/code/agent.py deleted file mode 100644 index d06157b..0000000 --- a/core/llm/code/agent.py +++ /dev/null @@ -1,202 +0,0 @@ -import logging -import os -import time -import queue -import threading -from queue import Queue -from dotenv import load_dotenv -from litellm import completion -from core.computer.code import Code -from .default_prompt import default_prompt, default_prompt_end -from .code_parser import CodeParser - -dotenv_path = os.path.join(os.path.dirname(__file__), ".env") -load_dotenv(dotenv_path) - -class CodeAgent: - def __init__(self): - self.code_executer = Code() - self.SYSTEM_PROMPT = default_prompt.format(language=str(self.code_executer.language_list)) - logging.info("[CodeAgent]"+self.SYSTEM_PROMPT) - self.SYSTEM_PROMPT_END = default_prompt_end - self.loop_breakers = ["The task is done.", "The task is impossible."] - self.model = os.getenv("CodeAgent_MODEL") - self.api_base = os.getenv("CodeAgent_API_BASE") - self.api_key = os.getenv("CodeAgent_API_KEY") - self.stop_code = False - self.stop_agent = False - self.permission = None - - def _listener(self, message_from_client: Queue): - """ - 监听来自客户端的信息 - :param message_from_client: - :return: - """ - while True: - try: - message = message_from_client.get(timeout=0.5) - except queue.Empty: - continue - print("get client message") - # 不是发给CodeAgent的信息，放回 - if message["name"] != "CodeAgent": - message_from_client.put(message) - continue - # 允许执行代码 - if message["type"] == "request": - if message["content"] == "deny": - self.permission = False - logging.info("[CodeAgent]User denied execution") - elif message["content"] == "approve": - self.permission = True - logging.info("[CodeAgent]User approved execution") - # 停止相关 - elif message["content"] == "stop_agent": - self.stop_agent = True - logging.info("[CodeAgent]User stopped agent") - elif message["content"] == "stop_code": - self.stop_code = True - logging.info("[CodeAgent]User stopped code execution") - - def _should_stop(self, ai_content: str = None): - if ai_content: - for loop_breaker in self.loop_breakers: - if loop_breaker in ai_content: - return True - return False - - def task(self, description: str, message_from_client: Queue, message_to_client: Queue): - self.stop_agent = False - messages = [ - {"role": "system", "content": self.SYSTEM_PROMPT}, - {"role": "user", "content": description}] - listener_thread = threading.Thread(target=self._listener, args=(message_from_client,)) - listener_thread.daemon = True - listener_thread.start() - logging.info("[CodeAgent][START]") - message_to_client.put({"name": "CodeAgent", "type": "status", "content": "[START]"}) - - while not self.stop_agent: - # 请求ai对话 - response = completion( - model=self.model, - api_base=self.api_base, - api_key=self.api_key, - messages=messages, - stream=True, - ) - # 流式返回 - message_to_client.put({"name": "CodeAgent", "type": "ai_content", "content": "[BEGIN]"}) - ai_content = "" - for chunk in response: - # 检查是否需要停止 - if self.stop_agent: - logging.info("[CodeAgent][STOP]: User stop") - message_to_client.put({"name": "CodeAgent", "type": "status", "content": "[STOP]"}) - return - if chunk.choices[0].delta.content: - delta = chunk.choices[0].delta.content - message_to_client.put({"name": "CodeAgent", "type": "ai_content", "content": delta}) - ai_content += delta - message_to_client.put({"name": "CodeAgent", "type": "ai_content", "content": "[END]"}) - messages.append({"role": "assistant", "content": ai_content}) - - # 检查ai是否想要停止 - if self._should_stop(ai_content): - break - - # TODO: If the context is too long, use RAG or other methods to handle it - # if len(messages) > 10: - # messages = [messages[0]] + messages[-5:] - - # 获取模型返回的代码 - codes = CodeParser(ai_content) - if not codes: - messages.append({"role": "user", "content": "No code blocks found. Skip execution. " + self.SYSTEM_PROMPT_END}) - continue - - # 依次执行代码 - logging.info(f"Find {len(codes)} code blocks. Executing...") - for i in range(len(codes)): - - logging.info(f"Executing code block {i}..., code: {codes[i]}") - code = codes[i] - self.permission = None - message_to_client.put({"name": "CodeAgent", "type": "status", "content": f"[BLOCK{i}]"}) - message_to_client.put({"name": "CodeAgent", "type": "request", "content": f"[BLOCK{i}]need_permission"}) - - # 阻塞等待客户端返回允许结果 - print("wtf???") - while self.permission is None: - logging.info("waiting permission") - print("waiting permission") - # 检查是否需要停止 - time.sleep(0.5) - if self.stop_agent: - logging.info("[CodeAgent][STOP]: User stop") - message_to_client.put({"name": "CodeAgent", "type": "status", "content": "[STOP]"}) - return - if not self.permission: - logging.info(f"User rejected to excecute code block{i}") - messages.append({"role": "user", "content": f"User rejected to excecute code block{i}"}) - continue - - logging.info(f"User approved to excecute code block {i}, start executing") - results = self.code_executer.run(lang=code["lang"], code=code["code"]) # 正确解包 - code_output_content = [{"type": "text", "text": f"Code block{i} is executed, and the output is as follows:"}] - - self.stop_code = False - while self.code_executer.is_running() or not results.empty(): - logging.info(f"[CodeAgent]: Code block{i} is running, code_executer:{self.code_executer.is_running()}, results.empty():{results.empty()}") - # 检查是否需要停止 - if self.stop_code or self.stop_agent: - self.code_executer.interrupt() - break - - try: - delta = results.get(timeout=0.2) - except queue.Empty: - continue - - # TODO: 有bug，图片先不返回 - if delta["type"] == "image/png" or delta["type"] == "image/jpeg": - # pass - message_to_client.put({"name": "CodeAgent", **delta}) - code_output_content.append( - { - "type": "image_url", - "image_url": { - "url": f"data:{delta["type"]};base64,{delta["content"]}" - } - } - ) - else: # 暂时把其他所有内容转为text - message_to_client.put({"name": "CodeAgent", "type": "text", "content": delta["content"]}) - code_output_content.append( - { - "type": "text", - "text": delta["content"] - } - ) - - logging.info(f"Finish executing code block {i}") - messages.append({"role": "user", "content": code_output_content}) - - messages.append({"role": "user", "content": "All code blocks are executed. " + self.SYSTEM_PROMPT_END}) - - logging.info("[CodeAgent][STOP]: task finished") - message_to_client.put({"name": "CodeAgent", "type": "status", "content": "[STOP]"}) - return messages[-1]["content"] - - - - - - - - - - - - diff --git a/core/llm/code/code_parser.py b/core/llm/code/code_parser.py deleted file mode 100644 index 83b91a9..0000000 --- a/core/llm/code/code_parser.py +++ /dev/null @@ -1,10 +0,0 @@ -import re - -def CodeParser(text: str): - pattern = r"```([\w\+\-\.]*)\n([\s\S]*?)\n```" - matches = re.findall(pattern, text) - - results = [] - for lang, code in matches: - results.append({"lang": lang.strip() if lang else "text", "code": code}) - return results \ No newline at end of file diff --git a/core/llm/code/default_prompt.py b/core/llm/code/default_prompt.py deleted file mode 100644 index 63a6852..0000000 --- a/core/llm/code/default_prompt.py +++ /dev/null @@ -1,33 +0,0 @@ -import getpass -import platform - - -default_prompt = f""" -You are Open Interpreter, a world-class programmer that can complete any goal by executing code. -For advanced requests, start by writing a plan. -When you execute code, it will be executed **on the user's machine**. The user has given you **full and complete permission** to execute any code necessary to complete the task. -You can access the internet. Run **any code** to achieve the goal, and if at first you don't succeed, try again and again. -You can install new packages. -When a user refers to a filename, they're likely referring to an existing file in the directory you're currently executing code in. -Write messages to the user in Markdown. -In general, try to **make plans** with as few steps as possible. -You should try something, print information about it, then continue from there in tiny, informed steps. -You will never get it on the first try, and attempting it in one go will often lead to errors you cant see. -You are capable of **any** task. -Print your code directly in the output. Do not print any additional information, including code explanation, running guide, etc. - -Use the following format to output your code: - -```python -# Your python code here -print("Result to inspect") -``` - -User's Name: {getpass.getuser()} -User's OS: {platform.system()} -Language: {{language}} -""" - -default_prompt_end = "Are we done? If the task is finished, print \"The task is done.\" in the first line, then summarize the result. If the task is not finished yet, what's next?" -# As for actually executing code to carry out that plan, for *stateful* languages (like python, javascript, shell, but NOT for html which starts from 0 every time) **it's critical not to try to do everything in one code block.** -# If the task is impossible, print \"The task is impossible.\" in the first line, then explain why. \ No newline at end of file diff --git "a/core/llm/code/\346\266\210\346\201\257\350\256\276\350\256\241.md" "b/core/llm/code/\346\266\210\346\201\257\350\256\276\350\256\241.md" deleted file mode 100644 index d780a77..0000000 --- "a/core/llm/code/\346\266\210\346\201\257\350\256\276\350\256\241.md" +++ /dev/null @@ -1,83 +0,0 @@ -# 消息设计 (Message Design) - -Agent 与客户端（UI）之间通过 `Queue` 进行通信，消息格式为 JSON 字典。 - -## 1. 客户端 -> Agent (Client to Agent) -客户端发送的消息主要用于控制 Agent 的行为和响应权限请求。 - -**基本格式:** -```json -{ - "name": "CodeAgent", // 给谁发送就填谁 - "type": "request", - "content": "指令内容" -} -``` - -**指令内容 (content):** -- `approve`: 允许执行当前代码块。 -- `deny`: 拒绝执行当前代码块。 -- `stop_agent`: 停止整个 Agent 任务。 -- `stop_code`: 停止当前正在执行的代码（中断执行）。 - -## 3. Agent -> 客户端 (Agent to Client) -Agent 发送的消息用于更新状态、流式传输 AI 回复、请求权限以及展示代码执行结果。 - -**基本格式:** -```json -{ - "name": "CodeAgent", - "type": "消息类型", - "content": "消息内容" -} -``` - -### 3.1 状态消息 (type: "status") -用于通知 UI 当前 Agent 的工作阶段。 -- `[START]`: 任务开始。 -- `[STOP]`: 任务结束（完成或被用户停止）。 -- `[BLOCK{i}]`: 正在处理第 `i` 个代码块（准备执行）。 - -### 3.2 AI 内容流 (type: "ai_content") -用于流式展示 LLM 的思考和回复过程。 -- `[BEGIN]`: 开始生成内容。 -- `[END]`: 内容生成结束。 -- ``: AI 生成的文本片段（增量更新）。 - -### 3.3 权限请求 (type: "request") -当 Agent 解析出代码块并准备执行时，向 UI 请求权限。 -- `[BLOCK{i}]need_permission`: 请求执行第 `i` 个代码块的权限。 - - UI 收到此消息后应显示“批准/拒绝”按钮。 - -### 3.4 代码执行输出 (type: "text" / "image/png" 等) -展示代码执行的实时输出。 - -- **文本输出 (type: "text"):** - - `content`: 标准输出 (stdout) 或标准错误 (stderr) 的文本内容。 - -- **图片输出 (type: "image/png" / "image/jpeg"):** - - *注意: 目前 `agent.py` 中图片处理逻辑被注释/跳过，需要修复才能正常接收图片。* - - `content`: Base64 编码的图片数据。 - -- **其他类型:** - - `html`: HTML 内容。 - - `javascript`: JS 代码。 - - `error`: 执行错误信息。 - -## 4. 交互流程示例 (Interaction Flow) - -调用`CodeAgent.task("任务描述")`之后： - -1. **任务开始**: Agent 发送 `status: [START]`。 -2. **AI 思考**: Agent 发送 `ai_content: [BEGIN]`, 随后发送多条内容片段, 最后发送 `ai_content: [END]`。 -3. **代码解析**: Agent 发现代码块。(Agent发现的代码块并不会传给客户端，客户端需要自己按顺序切分代码片段) -4. **请求权限**: Agent 发送 `status: [BLOCK0]`, `request: [BLOCK0]need_permission`。 -5. **用户确认**: 客户端发送 `content: approve`。 -6. **代码执行**: Agent 执行代码，持续发送 `type: text` (日志/结果)。 -7. **执行完成**: Agent 继续下一轮或结束。 -8. **任务结束**: Agent 发送 `status: [STOP]`。 - -## 5. TODO (Suggestions) -1. **修复图片传输**: base64太长，之后可以采取压缩方法 -2. **结构化状态**: 目前的 `[START]` 等状态码混在 `content` 字符串中，建议后续优化为独立的枚举字段，方便前端解析。 -3. **上下文未处理** diff --git a/core/llm/gui/__init__.py b/core/llm/gui/__init__.py deleted file mode 100644 index e69de29..0000000 diff --git a/requirements.txt b/requirements.txt deleted file mode 100644 index cac7992..0000000 --- a/requirements.txt +++ /dev/null @@ -1,3 +0,0 @@ -litellm -jupyter_client -pyautogui \ No newline at end of file