diff --git a/mcp/mcp_settings_deep_agent.json b/mcp/mcp_settings_deep_agent.json index ddf9962..3aa61ae 100644 --- a/mcp/mcp_settings_deep_agent.json +++ b/mcp/mcp_settings_deep_agent.json @@ -1,14 +1,5 @@ [ { - "mcpServers": { - "rag_retrieve": { - "transport": "stdio", - "command": "python", - "args": [ - "./mcp/rag_retrieve_server.py", - "{bot_id}" - ] - } - } + "mcpServers": {} } ] diff --git a/prompt/system_prompt_deep_agent.md b/prompt/system_prompt_deep_agent.md index 3ec5258..dcade57 100644 --- a/prompt/system_prompt_deep_agent.md +++ b/prompt/system_prompt_deep_agent.md @@ -1,5 +1,8 @@ + Working directory: {agent_dir_path} +Current User: {user_identifier} +Current Time: {datetime} ### Current Working Directory @@ -8,6 +11,13 @@ The filesystem backend is currently operating in: `{agent_dir_path}` ### File System and Paths +**CRITICAL - Directory Access Restriction:** +- You are **ONLY** allowed to access files and directories within `{agent_dir_path}` +- **NEVER** attempt to access files outside this directory (e.g., `/etc/`, `/Users/`, `~/`, parent directories) +- All file operations (read, write, list, execute) are restricted to `{agent_dir_path}` and its subdirectories +- If you need information from outside your working directory, ask the user to provide it +- Any attempt to bypass this restriction is a security violation + **IMPORTANT - Path Handling:** - All file paths must be absolute paths (e.g., `{agent_dir_path}/file.txt`) - Use the working directory from to construct absolute paths @@ -55,8 +65,4 @@ When using the write_todos tool: - If they want changes, adjust the plan accordingly 6. Update todo status promptly as you complete each item -The todo list is a planning tool - use it judiciously to avoid overwhelming the user with excessive task tracking. - -## System Information -- **Current User**: {user_identifier} -- **Current Time**: {datetime} +The todo list is a planning tool - use it judiciously to avoid overwhelming the user with excessive task tracking. \ No newline at end of file diff --git a/routes/chat.py b/routes/chat.py index 746f5df..0d67f60 100644 --- a/routes/chat.py +++ b/routes/chat.py @@ -286,10 +286,10 @@ async def chat_completions(request: ChatRequest, authorization: Optional[str] = raise HTTPException(status_code=400, detail="bot_id is required") # 创建项目目录（如果有dataset_ids且不是agent类型） - project_dir = create_project_directory(request.dataset_ids, bot_id, request.robot_type) + project_dir = create_project_directory(request.dataset_ids, bot_id, request.robot_type, request.skills) # 收集额外参数作为 generate_cfg - exclude_fields = {'messages', 'model', 'model_server', 'dataset_ids', 'language', 'tool_response', 'system_prompt', 'mcp_settings' ,'stream', 'robot_type', 'bot_id', 'user_identifier', 'session_id', 'enable_thinking'} + exclude_fields = {'messages', 'model', 'model_server', 'dataset_ids', 'language', 'tool_response', 'system_prompt', 'mcp_settings' ,'stream', 'robot_type', 'bot_id', 'user_identifier', 'session_id', 'enable_thinking', 'skills'} generate_cfg = {k: v for k, v in request.model_dump().items() if k not in exclude_fields} # 处理消息 messages = process_messages(request.messages, request.language) @@ -336,10 +336,10 @@ async def chat_warmup_v1(request: ChatRequest, authorization: Optional[str] = He raise HTTPException(status_code=400, detail="bot_id is required") # 创建项目目录（如果有dataset_ids且不是agent类型） - project_dir = create_project_directory(request.dataset_ids, bot_id, request.robot_type) + project_dir = create_project_directory(request.dataset_ids, bot_id, request.robot_type, request.skills) # 收集额外参数作为 generate_cfg - exclude_fields = {'messages', 'model', 'model_server', 'dataset_ids', 'language', 'tool_response', 'system_prompt', 'mcp_settings' ,'stream', 'robot_type', 'bot_id', 'user_identifier', 'session_id', 'enable_thinking'} + exclude_fields = {'messages', 'model', 'model_server', 'dataset_ids', 'language', 'tool_response', 'system_prompt', 'mcp_settings' ,'stream', 'robot_type', 'bot_id', 'user_identifier', 'session_id', 'enable_thinking', 'skills'} generate_cfg = {k: v for k, v in request.model_dump().items() if k not in exclude_fields} # 创建一个空的消息列表用于预热（实际消息不会在warmup中处理） @@ -431,11 +431,12 @@ async def chat_warmup_v2(request: ChatRequestV2, authorization: Optional[str] = # 从后端API获取机器人配置（使用v2的鉴权方式） bot_config = await fetch_bot_config(bot_id) - # 创建项目目录（从后端配置获取dataset_ids） + # 创建项目目录（从后端配置获取dataset_ids和skills） project_dir = create_project_directory( bot_config.get("dataset_ids", []), bot_id, - bot_config.get("robot_type", "general_agent") + bot_config.get("robot_type", "general_agent"), + bot_config.get("skills") ) # 创建一个空的消息列表用于预热（实际消息不会在warmup中处理） @@ -533,11 +534,12 @@ async def chat_completions_v2(request: ChatRequestV2, authorization: Optional[st # 从后端API获取机器人配置（使用v2的鉴权方式） bot_config = await fetch_bot_config(bot_id) - # 创建项目目录（从后端配置获取dataset_ids） + # 创建项目目录（从后端配置获取dataset_ids和skills） project_dir = create_project_directory( bot_config.get("dataset_ids", []), bot_id, - bot_config.get("robot_type", "general_agent") + bot_config.get("robot_type", "general_agent"), + bot_config.get("skills") ) # 处理消息 messages = process_messages(request.messages, request.language) diff --git a/utils/api_models.py b/utils/api_models.py index f1a8044..1ea2d91 100644 --- a/utils/api_models.py +++ b/utils/api_models.py @@ -54,6 +54,7 @@ class ChatRequest(BaseModel): user_identifier: Optional[str] = "" session_id: Optional[str] = None enable_thinking: Optional[bool] = DEFAULT_THINKING_ENABLE + skills: Optional[List[str]] = None class ChatRequestV2(BaseModel): diff --git a/utils/fastapi_utils.py b/utils/fastapi_utils.py index e11e066..e154657 100644 --- a/utils/fastapi_utils.py +++ b/utils/fastapi_utils.py @@ -364,17 +364,17 @@ def format_messages_to_chat_history(messages: List[Dict[str, str]]) -> str: return "\n".join(recent_chat_history) -def create_project_directory(dataset_ids: Optional[List[str]], bot_id: str, robot_type: str = "general_agent") -> Optional[str]: +def create_project_directory(dataset_ids: Optional[List[str]], bot_id: str, robot_type: str = "general_agent", skills: Optional[List[str]] = None) -> Optional[str]: """创建项目目录的公共逻辑""" # 只有当 robot_type == "catalog_agent" 且 dataset_ids 不为空时才创建目录 - + if robot_type == "general_agent": return None try: from utils.multi_project_manager import create_robot_project from pathlib import Path - return create_robot_project(dataset_ids, bot_id, Path("~", ".deepagents")) + return create_robot_project(dataset_ids, bot_id, project_path=Path("~", ".deepagents"), skills=skills) except Exception as e: logger.error(f"Error creating project directory: {e}") return None diff --git a/utils/multi_project_manager.py b/utils/multi_project_manager.py index b7b167a..8547542 100644 --- a/utils/multi_project_manager.py +++ b/utils/multi_project_manager.py @@ -15,6 +15,7 @@ from datetime import datetime logger = logging.getLogger('app') from utils.file_utils import get_document_preview +from utils import settings def generate_robot_directory_tree(robot_dir: str, robot_id: str, max_depth: int = 3) -> str: @@ -301,22 +302,42 @@ def generate_robot_readme(robot_id: str, dataset_ids: List[str], copy_results: L return str(readme_path) +def _get_robot_dir(project_path: Path, bot_id: str) -> Path: + """ + 获取 robot 目录路径，处理软链接情况 + + Args: + project_path: 项目路径 + bot_id: 机器人ID + + Returns: + Path: robot 目录路径（已展开） + """ + resolved_path = project_path.expanduser().resolve() + if resolved_path.name == "robot": + # project_path 已经指向 robot 目录（如 ~/.deepagents -> projects/robot） + return (project_path / bot_id).expanduser() + else: + # project_path 指向 projects 目录 + return (project_path / "robot" / bot_id).expanduser() + + def should_rebuild_robot_project(dataset_ids: List[str], bot_id: str, project_path: Path) -> bool: """ 检查是否需要重建机器人项目 1. 检查机器人项目是否存在 2. 检查是否有新增的dataset_id 3. 检查processing_log.json文件是否更新 - + Args: dataset_ids: 源项目ID列表 bot_id: 机器人ID project_path: 项目路径 - + Returns: bool: 是否需要重建 """ - robot_dir = project_path / "robot" / bot_id + robot_dir = _get_robot_dir(project_path, bot_id) # 如果机器人项目不存在，需要创建 if not robot_dir.exists(): @@ -375,35 +396,39 @@ def should_rebuild_robot_project(dataset_ids: List[str], bot_id: str, project_pa return False -def create_robot_project(dataset_ids: List[str], bot_id: str, force_rebuild: bool = False, project_path: Path = Path("projects")) -> str: +def create_robot_project(dataset_ids: List[str], bot_id: str, force_rebuild: bool = False, project_path: Path = Path("projects"), skills: Optional[List[str]] = None) -> str: """ 创建机器人项目，合并多个源项目的dataset文件夹 - + Args: dataset_ids: 源项目ID列表 bot_id: 机器人ID force_rebuild: 是否强制重建 - + skills: 技能文件名列表（如 ["rag-retrieve", "device_controller.zip"]） + Returns: str: 机器人项目目录路径 """ - logger.info(f"Creating robot project: {bot_id} from sources: {dataset_ids}") + logger.info(f"Creating robot project: {bot_id} from sources: {dataset_ids}, skills: {skills}") # 检查是否需要重建 if not force_rebuild and not should_rebuild_robot_project(dataset_ids, bot_id, project_path): robot_dir = project_path / "robot" / bot_id logger.info(f"Using existing robot project: {robot_dir}") + # 即使使用现有项目，也要处理 skills（如果提供了） + if skills: + _extract_skills_to_robot(robot_dir, skills, project_path) return str(robot_dir) - + # 创建机器人目录结构 - robot_dir = project_path / "robot" / bot_id + robot_dir = _get_robot_dir(project_path, bot_id) dataset_dir = robot_dir / "dataset" - + # 清理已存在的目录（如果需要） if robot_dir.exists(): logger.info(f"Robot directory already exists, cleaning up: {robot_dir}") shutil.rmtree(robot_dir) - + robot_dir.mkdir(parents=True, exist_ok=True) dataset_dir.mkdir(parents=True, exist_ok=True) @@ -435,6 +460,11 @@ def create_robot_project(dataset_ids: List[str], bot_id: str, force_rebuild: boo config_file = robot_dir / "robot_config.json" config_data = { "dataset_ids": dataset_ids, + "bot_id": bot_id, + "env": { + "backend_host": settings.BACKEND_HOST, + "masterkey": settings.MASTERKEY + }, "created_at": datetime.now().isoformat(), "total_folders": len(copy_results), "successful_copies": sum(1 for r in copy_results if r["success"]) @@ -454,7 +484,11 @@ def create_robot_project(dataset_ids: List[str], bot_id: str, force_rebuild: boo logger.info(f" Successful copies: {successful_copies}") logger.info(f" Config saved: {config_file}") logger.info(f" README generated: {readme_path}") - + + # 处理 skills 解压 + if skills: + _extract_skills_to_robot(robot_dir, skills, project_path) + return str(robot_dir) @@ -462,7 +496,58 @@ if __name__ == "__main__": # 测试代码 test_dataset_ids = ["test-project-1", "test-project-2"] test_bot_id = "test-robot-001" - + robot_dir = create_robot_project(test_dataset_ids, test_bot_id) logger.info(f"Created robot project at: {robot_dir}") - + + +def _extract_skills_to_robot(robot_dir: Path, skills: List[str], project_path: Path) -> None: + """ + 解压 skills 到 robot 项目的 skills 文件夹 + + Args: + robot_dir: 机器人项目目录 + skills: 技能文件名列表（如 ["rag-retrieve", "device_controller.zip"]） + project_path: 项目路径 + """ + import zipfile + + # skills 源目录在 projects/skills，需要通过解析软链接获取正确路径 + # project_path 可能是 ~/.deepagents (软链接 -> projects/robot) + # 所以 skills 源目录是 project_path.resolve().parent / "skills" + resolved_path = project_path.expanduser().resolve() + skills_source_dir = resolved_path.parent / "skills" + skills_target_dir = robot_dir / "skills" + + # 先清空 skills_target_dir，然后重新解压 + if skills_target_dir.exists(): + logger.info(f" Removing existing skills directory: {skills_target_dir}") + shutil.rmtree(skills_target_dir) + + skills_target_dir.mkdir(parents=True, exist_ok=True) + logger.info(f"Extracting skills to {skills_target_dir}") + + for skill in skills: + # 规范化文件名（确保有 .zip 后缀） + if not skill.endswith(".zip"): + skill_file = skill + ".zip" + else: + skill_file = skill + + skill_source_path = skills_source_dir / skill_file + + if not skill_source_path.exists(): + logger.warning(f" Skill file not found: {skill_source_path}") + continue + + # 获取解压后的文件夹名称（去掉 .zip 后缀） + folder_name = skill_file.replace(".zip", "") + extract_target = skills_target_dir / folder_name + + # 解压文件 + try: + with zipfile.ZipFile(skill_source_path, 'r') as zip_ref: + zip_ref.extractall(extract_target) + logger.info(f" Extracted: {skill_file} -> {extract_target}") + except Exception as e: + logger.error(f" Failed to extract {skill_file}: {e}") diff --git a/utils/symlink_utils.py b/utils/symlink_utils.py index bd7a3b0..f9710de 100644 --- a/utils/symlink_utils.py +++ b/utils/symlink_utils.py @@ -24,47 +24,12 @@ def setup_deepagents_symlink(): # Create robot directory if it doesn't exist robot_dir.mkdir(parents=True, exist_ok=True) - # If ~/.deepagents already exists and is not a symlink, backup and remove it - if deepagents_dir.exists() and not deepagents_dir.is_symlink(): - backup_dir = deepagents_dir.parent / f"{deepagents_dir.name}.backup" - logger.warning(f"~/.deepagents directory exists but is not a symlink.") - logger.warning(f"Creating backup at {backup_dir}") - - try: - # Create backup - import shutil - if backup_dir.exists(): - shutil.rmtree(backup_dir) - shutil.move(str(deepagents_dir), str(backup_dir)) - logger.info(f"Successfully backed up existing directory to {backup_dir}") - except Exception as backup_error: - logger.error(f"Failed to backup existing directory: {backup_error}") - logger.error("Please manually remove or backup ~/.deepagents to proceed") - return False - - # If ~/.deepagents is already a symlink pointing to the right place, do nothing - if deepagents_dir.is_symlink(): - target = deepagents_dir.resolve() - if target == robot_dir.resolve(): - logger.info(f"~/.deepagents already points to {robot_dir}") - return True - else: - # Remove existing symlink pointing elsewhere - deepagents_dir.unlink() - logger.info(f"Removed existing symlink pointing to {target}") + # If ~/.deepagents already exists, do nothing + if deepagents_dir.exists(): + logger.info(f"~/.deepagents already exists at {deepagents_dir}, skipping symlink creation") + return True # Create the symbolic link - # Check again before creating to handle race conditions - if deepagents_dir.is_symlink() or deepagents_dir.exists(): - logger.warning(f"Path {deepagents_dir} exists, attempting to remove before symlink") - if deepagents_dir.is_symlink(): - deepagents_dir.unlink() - elif deepagents_dir.is_dir(): - import shutil - shutil.rmtree(str(deepagents_dir)) - else: - deepagents_dir.unlink() - os.symlink(robot_dir, deepagents_dir, target_is_directory=True) logger.info(f"Created symbolic link: {deepagents_dir} -> {robot_dir}") return True