from __future__ import annotations import abc from typing import Any from pydantic import BaseModel, Field from app.core.config import settings class SandboxLimits(BaseModel): timeout: int = Field(default_factory=lambda: settings.sandbox_timeout) memory_mb: int = Field(default_factory=lambda: settings.sandbox_memory_mb) cpus: float = Field(default_factory=lambda: settings.sandbox_cpus) network: bool = Field(default_factory=lambda: settings.sandbox_network) max_output: int = Field(default_factory=lambda: settings.sandbox_max_output) class ExecResult(BaseModel): ok: bool = True exit_code: int = 0 stdout: str = "" stderr: str = "" duration_ms: int = 0 timed_out: bool = False truncated: bool = False backend: str = "" files_written: list[str] = Field(default_factory=list) error: str | None = None def summary(self) -> str: if self.timed_out: return f"执行超时(>{'?'}s)" if self.error: return self.error return self.stdout or self.stderr or "(无输出)" class Sandbox(abc.ABC): """代码执行后端的统一接口。""" name: str = "base" @abc.abstractmethod async def run( self, code: str, *, language: str = "python", limits: SandboxLimits | None = None, session_id: str | None = None, env: dict[str, str] | None = None, files: dict[str, str] | None = None, ) -> ExecResult: """执行一段代码。session_id 相同的调用共享同一个文件系统, 这样 agent 可以先写文件、再在下一步读回来。""" async def health(self) -> dict[str, Any]: return {"backend": self.name, "available": True} async def cleanup(self, session_id: str) -> None: return None async def close(self) -> None: return None def truncate(text: str, limit: int) -> tuple[str, bool]: if len(text) <= limit: return text, False half = limit // 2 return ( text[:half] + f"\n\n…(省略 {len(text) - limit} 字符)…\n\n" + text[-half:], True, )