Merge branch 'prod'

agent/agent_config.py

@@ -31,6 +31,7 @@ class AgentConfig:
     user_identifier: Optional[str] = None
     session_id: Optional[str] = None
     dataset_ids: Optional[List[str]] = field(default_factory=list)
+    trace_id: Optional[str] = None  # 请求追踪ID，从 X-Request-ID header 获取
 
     # 响应控制参数
     stream: bool = False
@@ -72,6 +73,7 @@ class AgentConfig:
             'messages': self.messages,
             'enable_memori': self.enable_memori,
             'memori_semantic_search_top_k': self.memori_semantic_search_top_k,
+            'trace_id': self.trace_id,
         }
         
     def safe_print(self):
@@ -93,10 +95,18 @@ class AgentConfig:
         )
         from .checkpoint_utils import prepare_checkpoint_message
         from .checkpoint_manager import get_checkpointer_manager
+        from utils.log_util.context import g
 
         if messages is None:
             messages = []
 
+        # 从全局上下文获取 trace_id
+        trace_id = None
+        try:
+            trace_id = getattr(g, 'trace_id', None)
+        except LookupError:
+            pass
+
         robot_type = request.robot_type
         if robot_type == "catalog_agent":
             robot_type = "deep_agent"
@@ -130,6 +140,7 @@ class AgentConfig:
             dataset_ids=request.dataset_ids,
             enable_memori=enable_memori,
             memori_semantic_search_top_k=getattr(request, 'memori_semantic_search_top_k', None) or MEM0_SEMANTIC_SEARCH_TOP_K,
+            trace_id=trace_id,
         )
 
         # 在创建 config 时尽早准备 checkpoint 消息
@@ -158,9 +169,17 @@ class AgentConfig:
         )
         from .checkpoint_utils import prepare_checkpoint_message
         from .checkpoint_manager import get_checkpointer_manager
+        from utils.log_util.context import g
 
         if messages is None:
             messages = []
+
+        # 从全局上下文获取 trace_id
+        trace_id = None
+        try:
+            trace_id = getattr(g, 'trace_id', None)
+        except LookupError:
+            pass
         language = request.language or bot_config.get("language", "zh")
         preamble_text, system_prompt = get_preamble_text(language, bot_config.get("system_prompt"))
         robot_type = bot_config.get("robot_type", "general_agent")
@@ -194,6 +213,7 @@ class AgentConfig:
             dataset_ids=bot_config.get("dataset_ids", []),  # 从后端配置获取dataset_ids
             enable_memori=enable_memori,
             memori_semantic_search_top_k=bot_config.get("memori_semantic_search_top_k", MEM0_SEMANTIC_SEARCH_TOP_K),
+            trace_id=trace_id,
         )
 
         # 在创建 config 时尽早准备 checkpoint 消息

agent/deep_assistant.py

@@ -137,7 +137,7 @@ async def init_agent(config: AgentConfig):
 
     # 加载配置
     final_system_prompt = await load_system_prompt_async(
-        config.project_dir, config.language, config.system_prompt, config.robot_type, config.bot_id, config.user_identifier
+        config.project_dir, config.language, config.system_prompt, config.robot_type, config.bot_id, config.user_identifier, config.trace_id or ""
     )
     final_mcp_settings = await load_mcp_settings_async(
         config.project_dir, config.mcp_settings, config.bot_id, config.robot_type
@@ -240,7 +240,11 @@ async def init_agent(config: AgentConfig):
             enable_memory=False,
             workspace_root=workspace_root,
             middleware=middleware,
-            checkpointer=checkpointer
+            checkpointer=checkpointer,
+            shell_env={
+              "ASSISTANT_ID": config.bot_id,
+              "TRACE_ID": config.trace_id
+            }
         )
     else:
         # 只有在 enable_thinking 为 True 时才添加 GuidelineMiddleware
@@ -369,6 +373,7 @@ def create_custom_cli_agent(
     workspace_root: str | None = None,
     checkpointer: Checkpointer | None = None,
     store: BaseStore | None = None,
+    shell_env: dict[str, str] | None = None,
 ) -> tuple[Pregel, CompositeBackend]:
     """Create a CLI-configured agent with custom workspace_root for shell commands.
 
@@ -393,6 +398,8 @@ def create_custom_cli_agent(
         workspace_root: Working directory for shell commands. If None, uses Path.cwd().
         checkpointer: Optional checkpointer for persisting conversation state
         store: Optional BaseStore for persisting user preferences and agent memory
+        shell_env: Optional custom environment variables to pass to ShellMiddleware.
+                   These will be merged with os.environ. Custom vars take precedence.
 
     Returns:
         2-tuple of (agent_graph, composite_backend)
@@ -440,15 +447,18 @@ def create_custom_cli_agent(
         # Add shell middleware (only in local mode)
         if enable_shell:
             # Create environment for shell commands
-            # Restore user's original LANGSMITH_PROJECT so their code traces separately
+            # Start with a copy of current environment
-            shell_env = os.environ.copy()
+            final_shell_env = os.environ.copy()
+            # Merge custom environment variables if provided (custom vars take precedence)
+            if shell_env:
+                final_shell_env.update(shell_env)
             # Use custom workspace_root if provided, otherwise use current directory
             shell_workspace = workspace_root if workspace_root is not None else str(Path.cwd())
 
             agent_middleware.append(
                 ShellMiddleware(
                     workspace_root=shell_workspace,
-                    env=shell_env,
+                    env=final_shell_env,
                 )
             )
     else:

agent/prompt_loader.py

@@ -69,9 +69,9 @@ def format_datetime_by_language(language: str) -> str:
             return utc_now.strftime("%Y-%m-%d %H:%M:%S") + " UTC"
 
 
-async def load_system_prompt_async(project_dir: str, language: str = None, system_prompt: str=None, robot_type: str = "general_agent", bot_id: str="", user_identifier: str = "") -> str:
+async def load_system_prompt_async(project_dir: str, language: str = None, system_prompt: str=None, robot_type: str = "general_agent", bot_id: str="", user_identifier: str = "", trace_id: str = "") -> str:
     """异步版本的系统prompt加载
-    
+
     Args:
         project_dir: 项目目录路径，可以为None
         language: 语言代码，如 'zh', 'en', 'jp' 等
@@ -79,7 +79,8 @@ async def load_system_prompt_async(project_dir: str, language: str = None, syste
         robot_type: 机器人类型，取值 agent/catalog_agent
         bot_id: 机器人ID
         user_identifier: 用户标识符
-        
+        trace_id: 请求追踪ID，用于日志追踪
+
     Returns:
         str: 加载到的系统提示词内容
     """
@@ -122,12 +123,13 @@ async def load_system_prompt_async(project_dir: str, language: str = None, syste
 
         # agent_dir_path = f"~/.deepagents/{bot_id}" #agent_dir_path 其实映射的就是 project_dir目录，只是给ai看的目录路径
         prompt = system_prompt_default.format(
-            readme=str(readme), 
+            readme=str(readme),
             extra_prompt=system_prompt or "",
-            language=language_display, 
+            language=language_display,
-            user_identifier=user_identifier, 
+            user_identifier=user_identifier,
             datetime=datetime_str,
-            agent_dir_path="."
+            agent_dir_path=".",
+            trace_id=trace_id or ""
         )
     elif system_prompt:
         prompt = system_prompt.format(language=language_display, user_identifier=user_identifier, datetime=datetime_str)

fastapi_app.py

@@ -71,7 +71,7 @@ from utils.log_util.logger import init_with_fastapi
 logger = logging.getLogger('app')
 
 # Import route modules
-from routes import chat, files, projects, system, skill_manager
+from routes import chat, files, projects, system, skill_manager, database
 
 
 @asynccontextmanager
@@ -174,6 +174,7 @@ app.include_router(files.router)
 app.include_router(projects.router)
 app.include_router(system.router)
 app.include_router(skill_manager.router)
+app.include_router(database.router)
 
 # 注册文件管理API路由
 app.include_router(file_manager_router)

prompt/system_prompt_deep_agent.md

@@ -47,7 +47,7 @@ When executing scripts from SKILL.md files, you MUST convert relative paths to a
 
 - **`{agent_dir_path}/skills/`** - Skill packages with embedded scripts
 - **`{agent_dir_path}/dataset/`** - Store file datasets and document data
-- **`{agent_dir_path}/scripts/`** - Place generated executable scripts here (not skill scripts)
+- **`{agent_dir_path}/executable_code/`** - Place generated executable scripts here (not skill scripts)
 - **`{agent_dir_path}/download/`** - Store downloaded files and content
 
 **Path Examples:**

routes/database.py

@@ -0,0 +1,291 @@
+"""
+数据库操作 API 路由
+提供数据库迁移、表结构变更等功能
+"""
+import logging
+from typing import Optional
+from fastapi import APIRouter, HTTPException, Header
+from pydantic import BaseModel
+
+from agent.db_pool_manager import get_db_pool_manager
+from utils.settings import MASTERKEY
+from utils.fastapi_utils import extract_api_key_from_auth
+
+logger = logging.getLogger('app')
+
+router = APIRouter()
+
+
+def verify_database_auth(authorization: Optional[str]) -> None:
+    """
+    验证数据库操作 API 的认证
+
+    直接使用 MASTERKEY 进行验证
+
+    Args:
+        authorization: Authorization header 值
+
+    Raises:
+        HTTPException: 认证失败时抛出 401/403 错误
+    """
+    # 提取提供的 token
+    provided_token = extract_api_key_from_auth(authorization)
+
+    if not provided_token:
+        raise HTTPException(
+            status_code=401,
+            detail="Authorization header is required"
+        )
+
+    if provided_token != MASTERKEY:
+        raise HTTPException(
+            status_code=403,
+            detail="Invalid authorization token"
+        )
+
+
+class DatabaseMigrationResponse(BaseModel):
+    """数据库迁移响应"""
+    success: bool
+    message: str
+    steps_completed: list[str]
+    error: Optional[str] = None
+
+
+class ExecuteSQLRequest(BaseModel):
+    """执行 SQL 请求"""
+    sql: str
+    autocommit: bool = True
+
+
+class ExecuteSQLResponse(BaseModel):
+    """执行 SQL 响应"""
+    success: bool
+    rows_affected: Optional[int] = None
+    message: str
+    columns: Optional[list[str]] = None
+    data: Optional[list[list]] = None
+    error: Optional[str] = None
+
+
+@router.post("/api/v1/database/migrate-pgvector", response_model=DatabaseMigrationResponse)
+async def migrate_pgvector(authorization: Optional[str] = Header(None)):
+    """
+    执行 pgvector 扩展安装迁移
+
+    执行步骤：
+    1. 将 public.vector 表重命名为 public.vector_legacy
+    2. 创建 pgvector 扩展 (CREATE EXTENSION vector)
+
+    注意：此操作会修改数据库结构，请确保在执行前已做好备份。
+
+    Authentication:
+        - Authorization header should contain: Bearer {MASTERKEY}
+
+    Returns:
+        DatabaseMigrationResponse: 迁移结果
+    """
+    # 验证认证
+    verify_database_auth(authorization)
+
+    steps_completed = []
+    pool_manager = get_db_pool_manager()
+
+    try:
+        # 获取异步连接
+        pool = pool_manager.pool
+
+        # 步骤 1: 重命名 vector 表为 vector_legacy
+        async with pool.connection() as conn:
+            async with conn.cursor() as cursor:
+                # 检查 vector 表是否存在
+                await cursor.execute("""
+                    SELECT EXISTS (
+                        SELECT FROM information_schema.tables
+                        WHERE table_schema = 'public' AND table_name = 'vector'
+                    )
+                """)
+                vector_exists = (await cursor.fetchone())[0]
+
+                if vector_exists:
+                    # 检查 vector_legacy 是否已存在
+                    await cursor.execute("""
+                        SELECT EXISTS (
+                            SELECT FROM information_schema.tables
+                            WHERE table_schema = 'public' AND table_name = 'vector_legacy'
+                        )
+                    """)
+                    legacy_exists = (await cursor.fetchone())[0]
+
+                    if legacy_exists:
+                        steps_completed.append("vector_legacy 表已存在，跳过重命名")
+                    else:
+                        # 执行重命名
+                        await cursor.execute("ALTER TABLE public.vector RENAME TO vector_legacy")
+                        steps_completed.append("已将 public.vector 表重命名为 public.vector_legacy")
+                        logger.info("Renamed public.vector to public.vector_legacy")
+                else:
+                    steps_completed.append("public.vector 表不存在，跳过重命名")
+
+                # 提交事务
+                await conn.commit()
+
+        # 步骤 2: 创建 pgvector 扩展
+        async with pool.connection() as conn:
+            async with conn.cursor() as cursor:
+                # 检查 pgvector 扩展是否已安装
+                await cursor.execute("""
+                    SELECT EXISTS (
+                        SELECT 1 FROM pg_extension WHERE extname = 'vector'
+                    )
+                """)
+                extension_exists = (await cursor.fetchone())[0]
+
+                if extension_exists:
+                    steps_completed.append("pgvector 扩展已安装")
+                else:
+                    # 创建 pgvector 扩展
+                    await cursor.execute("CREATE EXTENSION vector")
+                    steps_completed.append("已成功安装 pgvector 扩展")
+                    logger.info("Created pgvector extension")
+
+                # 提交事务
+                await conn.commit()
+
+        return DatabaseMigrationResponse(
+            success=True,
+            message="pgvector 迁移完成",
+            steps_completed=steps_completed
+        )
+
+    except HTTPException:
+        raise
+    except Exception as e:
+        logger.error(f"pgvector 迁移失败: {e}")
+        return DatabaseMigrationResponse(
+            success=False,
+            message="pgvector 迁移失败",
+            steps_completed=steps_completed,
+            error=str(e)
+        )
+
+
+@router.post("/api/v1/database/execute-sql", response_model=ExecuteSQLResponse)
+async def execute_sql(request: ExecuteSQLRequest, authorization: Optional[str] = Header(None)):
+    """
+    执行自定义 SQL 语句
+
+    注意：此接口具有较高权限，请谨慎使用。
+
+    Authentication:
+        - Authorization header should contain: Bearer {MASTERKEY}
+
+    Args:
+        request: 包含 SQL 语句和是否自动提交的请求
+
+    Returns:
+        ExecuteSQLResponse: 执行结果
+    """
+    # 验证认证
+    verify_database_auth(authorization)
+
+    pool_manager = get_db_pool_manager()
+
+    try:
+        pool = pool_manager.pool
+
+        async with pool.connection() as conn:
+            async with conn.cursor() as cursor:
+                await cursor.execute(request.sql)
+                rows_affected = cursor.rowcount
+
+                # 获取列名
+                columns = None
+                data = None
+                if cursor.description:
+                    columns = [desc.name for desc in cursor.description]
+                    # 获取所有行数据
+                    rows = await cursor.fetchall()
+                    data = [list(row) for row in rows]
+
+                if request.autocommit:
+                    await conn.commit()
+
+                return ExecuteSQLResponse(
+                    success=True,
+                    rows_affected=rows_affected,
+                    message=f"SQL 执行成功，影响行数: {rows_affected}, 返回数据: {len(data) if data else 0} 行",
+                    columns=columns,
+                    data=data
+                )
+
+    except HTTPException:
+        raise
+    except Exception as e:
+        logger.error(f"SQL 执行失败: {e}")
+        raise HTTPException(
+            status_code=500,
+            detail=f"SQL 执行失败: {str(e)}"
+        )
+
+
+@router.get("/api/v1/database/check-pgvector")
+async def check_pgvector(authorization: Optional[str] = Header(None)):
+    """
+    检查 pgvector 扩展安装状态
+
+    Authentication:
+        - Authorization header should contain: Bearer {MASTERKEY}
+
+    Returns:
+        pgvector 扩展的状态信息
+    """
+    # 验证认证
+    verify_database_auth(authorization)
+
+    pool_manager = get_db_pool_manager()
+
+    try:
+        pool = pool_manager.pool
+
+        async with pool.connection() as conn:
+            async with conn.cursor() as cursor:
+                # 检查扩展是否存在
+                await cursor.execute("""
+                    SELECT
+                        extname,
+                        extversion
+                    FROM pg_extension
+                    WHERE extname = 'vector'
+                """)
+                extension_result = await cursor.fetchone()
+
+                # 检查 vector 相关表
+                await cursor.execute("""
+                    SELECT
+                        table_name,
+                        table_type
+                    FROM information_schema.tables
+                    WHERE table_schema = 'public'
+                    AND table_name LIKE '%vector%'
+                    ORDER BY table_name
+                """)
+                tables = await cursor.fetchall()
+
+                return {
+                    "extension_installed": extension_result is not None,
+                    "extension_version": extension_result[1] if extension_result else None,
+                    "vector_tables": [
+                        {"name": row[0], "type": row[1]}
+                        for row in tables
+                    ]
+                }
+
+    except HTTPException:
+        raise
+    except Exception as e:
+        logger.error(f"检查 pgvector 状态失败: {e}")
+        raise HTTPException(
+            status_code=500,
+            detail=f"检查失败: {str(e)}"
+        )

routes/skill_manager.py

@@ -4,6 +4,7 @@ import shutil
 import zipfile
 import logging
 import asyncio
+import yaml
 from typing import List, Optional
 from fastapi import APIRouter, HTTPException, Query, UploadFile, File, Form
 from pydantic import BaseModel
@@ -294,14 +295,12 @@ def parse_skill_frontmatter(skill_md_path: str) -> Optional[dict]:
             return None
 
         frontmatter = frontmatter_match.group(1)
-        metadata = {}
 
-        # Parse key: value pairs from frontmatter
+        # Parse YAML using yaml.safe_load
-        for line in frontmatter.split('\n'):
+        metadata = yaml.safe_load(frontmatter)
-            line = line.strip()
+        if not isinstance(metadata, dict):
-            if ':' in line:
+            logger.warning(f"Invalid frontmatter format in {skill_md_path}")
-                key, value = line.split(':', 1)
+            return None
-                metadata[key.strip()] = value.strip()
 
         # Return name and description if both exist
         if 'name' in metadata and 'description' in metadata: