@icyfenix-dmla/cli 2026.5.3-1019 → 2026.5.3-1248

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@icyfenix-dmla/cli",
-  "version": "2026.5.3-1019",
+  "version": "2026.5.3-1248",
   "description": "DMLA 沙箱服务命令行工具",
   "type": "module",
   "main": "src/index.js",

package/src/commands/server.js

@@ -326,6 +326,25 @@ function findServerPath() {
   return null
 }
 
+/**
+ * 查找 kernel_runner.py 路径
+ * --dev 模式下需要挂载此文件
+ */
+function findKernelRunnerPath() {
+  // 开发环境路径：packages/cli/src/commands -> ../../../local-server/src/kernel_runner.py
+  const devPath = path.resolve(__dirname, '../../../local-server/src/kernel_runner.py')
+  // npm 包路径：packages/cli/src/commands -> ../server/kernel_runner.py（构建后）
+  const npmPath = path.resolve(__dirname, '../server/kernel_runner.py')
+
+  if (fs.existsSync(devPath)) {
+    return devPath
+  }
+  if (fs.existsSync(npmPath)) {
+    return npmPath
+  }
+  return null
+}
+
 /**
  * 查找共享模块目录
  * --dev 模式下需要挂载此目录
@@ -427,10 +446,17 @@ export async function startServerSync(port, useGpu = false, dev = false) {
 
   // 查找共享模块路径（--dev 模式需要）
   const sharedModulesPath = dev ? findSharedModulesPath() : null
+  // 查找 kernel_runner.py 路径（--dev 模式需要）
+  const kernelRunnerPath = dev ? findKernelRunnerPath() : null
+
   if (dev && !sharedModulesPath) {
     console.log(chalk.yellow('⚠️ --dev 模式需要共享模块目录'))
     console.log(chalk.gray('   未找到 shared_modules，将仅使用镜像内置模块'))
   }
+  if (dev && !kernelRunnerPath) {
+    console.log(chalk.yellow('⚠️ --dev 模式需要 kernel_runner.py'))
+    console.log(chalk.gray('   未找到 kernel_runner.py，将仅使用镜像内置版本'))
+  }
 
   console.log(chalk.gray(`   镜像类型: ${imageResolution.message}`))
   console.log(chalk.gray('   同步模式启动...'))
@@ -438,6 +464,9 @@ export async function startServerSync(port, useGpu = false, dev = false) {
   if (dev && sharedModulesPath) {
     console.log(chalk.gray(`   共享模块: ${sharedModulesPath}`))
   }
+  if (dev && kernelRunnerPath) {
+    console.log(chalk.gray(`   执行器: ${kernelRunnerPath}`))
+  }
   console.log()
 
   // 设置环境变量
@@ -452,6 +481,9 @@ export async function startServerSync(port, useGpu = false, dev = false) {
     if (sharedModulesPath) {
       process.env.SHARED_MODULES_PATH = sharedModulesPath
     }
+    if (kernelRunnerPath) {
+      process.env.KERNEL_RUNNER_PATH = kernelRunnerPath
+    }
   }
 
   // 动态 import 服务器模块并直接运行
@@ -542,9 +574,15 @@ export async function startServer(port, useGpu = false, dev = false) {
 
     // 查找共享模块路径（--dev 模式需要）
     const sharedModulesPath = dev ? findSharedModulesPath() : null
+    // 查找 kernel_runner.py 路径（--dev 模式需要）
+    const kernelRunnerPath = dev ? findKernelRunnerPath() : null
+
     if (dev && sharedModulesPath) {
       console.log(chalk.gray(`   共享模块: ${sharedModulesPath}`))
     }
+    if (dev && kernelRunnerPath) {
+      console.log(chalk.gray(`   执行器: ${kernelRunnerPath}`))
+    }
 
     // 日志文件路径
     const logDir = path.resolve(__dirname, '../../logs')
@@ -574,6 +612,9 @@ export async function startServer(port, useGpu = false, dev = false) {
       if (sharedModulesPath) {
         env.SHARED_MODULES_PATH = sharedModulesPath
       }
+      if (kernelRunnerPath) {
+        env.KERNEL_RUNNER_PATH = kernelRunnerPath
+      }
     }
 
     // 写入启动日志

package/src/server/index.js

@@ -7,6 +7,7 @@ import cors from 'cors'
 import { fileURLToPath } from 'url'
 import { resolve } from 'path'
 import sandboxRouter from './routes/sandbox.js'
+import { cleanupAllContainers } from './sandbox.js'
 
 export const app = express()
 const PORT = process.env.PORT || 3001
@@ -76,14 +77,26 @@ process.on('unhandledRejection', (reason, promise) => {
   log(`UNHANDLED REJECTION: ${reason}`)
 })
 
-// 捕获进程信号
-process.on('SIGTERM', () => {
-  log('Received SIGTERM')
+// 捕获进程信号 - 优雅退出，清理所有容器
+process.on('SIGTERM', async () => {
+  log('Received SIGTERM, cleaning up containers...')
+  try {
+    await cleanupAllContainers()
+    log('All containers cleaned up')
+  } catch (e) {
+    log(`Cleanup error: ${e.message}`)
+  }
   process.exit(0)
 })
 
-process.on('SIGINT', () => {
-  log('Received SIGINT')
+process.on('SIGINT', async () => {
+  log('Received SIGINT (Ctrl+C), cleaning up containers...')
+  try {
+    await cleanupAllContainers()
+    log('All containers cleaned up')
+  } catch (e) {
+    log(`Cleanup error: ${e.message}`)
+  }
   process.exit(0)
 })
 

package/src/server/routes/sandbox.js

@@ -3,7 +3,7 @@
  */
 import { Router } from 'express'
 import Docker from 'dockerode'
-import sandbox, { runPythonCode, checkImageExists, checkGPUAvailable, checkCUDACompatibility } from '../sandbox.js'
+import sandbox, { runPythonCode, checkImageExists, checkGPUAvailable, checkCUDACompatibility, abortExecution } from '../sandbox.js'
 
 const { SANDBOX_CONFIG } = sandbox
 const docker = new Docker()
@@ -232,4 +232,31 @@ router.get('/cuda-compat', async (req, res) => {
   }
 })
 
+/**
+ * 中止执行
+ * POST /api/sandbox/abort
+ * Body: { executionId?: string }  // 可选，不传则中止所有
+ */
+router.post('/abort', async (req, res) => {
+  const { executionId = null } = req.body
+
+  try {
+    const result = await abortExecution(executionId)
+
+    res.json({
+      success: result.success,
+      stopped: result.stopped,
+      message: result.stopped > 0
+        ? `已中止 ${result.stopped} 个执行任务`
+        : '没有正在运行的任务'
+    })
+  } catch (error) {
+    console.error('[Sandbox] Abort error:', error)
+    res.status(500).json({
+      success: false,
+      error: error.message || '中止失败'
+    })
+  }
+})
+
 export default router

package/src/server/sandbox.js

@@ -20,6 +20,91 @@ function log(message) {
 // 启动时记录
 log('Sandbox module initialized')
 
+// ==================== 全局容器追踪 ====================
+// 用于追踪所有活跃的容器，支持中止操作和优雅退出清理
+const activeContainers = new Map()
+
+/**
+ * 生成唯一的执行 ID
+ * 格式: exec-{时间戳}-{随机后缀}
+ */
+function generateExecutionId() {
+  return `exec-${Date.now()}-${Math.random().toString(36).slice(2, 7)}`
+}
+
+/**
+ * 注册容器到活跃列表
+ * @param {string} executionId - 执行 ID
+ * @param {Docker.Container} container - Docker 容器实例
+ */
+function registerContainer(executionId, container) {
+  activeContainers.set(executionId, container)
+  log(`Container registered: ${executionId} -> ${container.id}`)
+}
+
+/**
+ * 从活跃列表移除容器
+ * @param {string} executionId - 执行 ID
+ */
+function unregisterContainer(executionId) {
+  activeContainers.delete(executionId)
+  log(`Container unregistered: ${executionId}`)
+}
+
+/**
+ * 清理所有活跃容器
+ * 用于中止操作和进程退出时清理
+ * @returns {Promise<number>} 清理的容器数量
+ */
+export async function cleanupAllContainers() {
+  const count = activeContainers.size
+  log(`Cleaning up ${count} active containers...`)
+
+  for (const [executionId, container] of activeContainers) {
+    try {
+      await container.stop({ t: 5 })  // 5秒优雅停止
+      await container.remove({ force: true })
+      log(`Container stopped and removed: ${executionId}`)
+    } catch (e) {
+      log(`Failed to cleanup container ${executionId}: ${e.message}`)
+    }
+  }
+
+  activeContainers.clear()
+  log(`All containers cleaned up: ${count}`)
+  return count
+}
+
+/**
+ * 中止指定执行或所有执行
+ * @param {string|null} executionId - 执行 ID，null 表示中止所有
+ * @returns {Promise<{success: boolean, stopped: number}>}
+ */
+export async function abortExecution(executionId = null) {
+  const containersToStop = executionId
+    ? [activeContainers.get(executionId)].filter(c => c)
+    : Array.from(activeContainers.values())
+
+  log(`Aborting ${containersToStop.length} containers (executionId: ${executionId || 'all'})`)
+
+  let stopped = 0
+  for (const [id, container] of containersToStop.map(c => [activeContainers.entries().find(([k, v]) => v === c)?.[0], c])) {
+    if (container) {
+      try {
+        await container.stop({ t: 5 })
+        await container.remove({ force: true })
+        activeContainers.delete(id)
+        stopped++
+        log(`Container aborted: ${id}`)
+      } catch (e) {
+        log(`Failed to abort container ${id}: ${e.message}`)
+      }
+    }
+  }
+
+  return { success: true, stopped }
+}
+
 // 检测运行模式并计算正确的路径
 // 开发模式: 从 local-server/src 运行，项目根目录在上两级
 // 独立模式: 从 packages/cli/src/server 运行，无 shared_modules 目录
@@ -123,6 +208,11 @@ function getProgressReporterPath() {
   if (process.env.PROGRESS_REPORTER_PATH) {
     return process.env.PROGRESS_REPORTER_PATH
   }
+  // --dev 模式下，尝试从 KERNEL_RUNNER_PATH 推断（同一目录）
+  if (process.env.KERNEL_RUNNER_PATH) {
+    const kernelDir = path.dirname(process.env.KERNEL_RUNNER_PATH)
+    return path.join(kernelDir, 'dmla_progress.py')
+  }
   return DEFAULT_PROGRESS_REPORTER_PATH
 }
 
@@ -140,9 +230,15 @@ function shouldMountSharedModules() {
 
 /**
  * 检查是否挂载本地 kernel_runner.py（开发模式）
+ * 条件1: MOUNT_KERNEL_RUNNER 不是 'false'（默认启用）
+ * 条件2: 有可用的路径（PROJECT_ROOT 或环境变量指定）
  */
 function shouldMountKernelRunner() {
-  return process.env.MOUNT_KERNEL_RUNNER !== 'false' && PROJECT_ROOT !== null
+  if (process.env.MOUNT_KERNEL_RUNNER === 'false') {
+    return false
+  }
+  // 有环境变量指定的路径，或者有 PROJECT_ROOT
+  return process.env.KERNEL_RUNNER_PATH || PROJECT_ROOT !== null
 }
 
 /**
@@ -320,12 +416,15 @@ print(json.dumps(result))
 export async function runPythonCode(code, useGpu = false, imageOverride = null, timeoutOverride = null) {
   const startTime = Date.now()
 
+  // 生成唯一执行 ID
+  const executionId = generateExecutionId()
+
   // 计算实际超时时间
   const actualTimeout = timeoutOverride === null
     ? null  // unlimited
     : (timeoutOverride || Math.floor(SANDBOX_CONFIG.timeout / 1000))
 
-  log(`runPythonCode called, useGpu=${useGpu}, code length=${code.length}, imageOverride=${imageOverride}, timeout=${actualTimeout === null ? 'unlimited' : actualTimeout}`)
+  log(`runPythonCode called, executionId=${executionId}, useGpu=${useGpu}, code length=${code.length}, imageOverride=${imageOverride}, timeout=${actualTimeout === null ? 'unlimited' : actualTimeout}`)
 
   // 选择镜像：优先使用指定的镜像，否则根据 useGpu 选择
   const image = imageOverride || (useGpu ? SANDBOX_CONFIG.imageGpu : SANDBOX_CONFIG.imageCpu)
@@ -372,7 +471,8 @@ export async function runPythonCode(code, useGpu = false, imageOverride = null,
           traceback: [errorMessage]
         }],
         executionTime,
-        gpuUsed: false
+        gpuUsed: false,
+        executionId
       }
     }
     log('CUDA compatibility check passed')
@@ -471,6 +571,9 @@ export async function runPythonCode(code, useGpu = false, imageOverride = null,
     container = await docker.createContainer(containerConfig)
     log(`Container created: ${container.id}`)
 
+    // 注册到活跃容器列表
+    registerContainer(executionId, container)
+
     // 设置超时（使用动态计算的超时时间，unlimited 时为 24 小时）
     const containerTimeoutMs = timeoutSeconds * 1000 + 10000  // 转换为毫秒，额外 10 秒用于清理
     const timeoutPromise = new Promise((_, reject) => {
@@ -528,7 +631,8 @@ export async function runPythonCode(code, useGpu = false, imageOverride = null,
           traceback: [stdout.substring(0, 1000)]
         }],
         executionTime,
-        gpuUsed: useGpu
+        gpuUsed: useGpu,
+        executionId
       }
     }
 
@@ -553,7 +657,8 @@ export async function runPythonCode(code, useGpu = false, imageOverride = null,
           traceback: [rawOutput]
         }],
         executionTime,
-        gpuUsed: useGpu
+        gpuUsed: useGpu,
+        executionId
       }
     }
 
@@ -561,7 +666,8 @@ export async function runPythonCode(code, useGpu = false, imageOverride = null,
       success: parsedResult.success,
       outputs: parsedResult.outputs || [],
       executionTime: parsedResult.executionTime || (Date.now() - startTime) / 1000,
-      gpuUsed: useGpu
+      gpuUsed: useGpu,
+      executionId
     }
 
   } catch (error) {
@@ -581,10 +687,14 @@ export async function runPythonCode(code, useGpu = false, imageOverride = null,
         traceback: [error.message || 'Unknown error']
       }],
       executionTime,
-      gpuUsed: useGpu
+      gpuUsed: useGpu,
+      executionId
     }
 
   } finally {
+    // 从活跃列表移除
+    unregisterContainer(executionId)
+
     // 清理容器
     log('Cleaning up container...')
     if (container) {
@@ -720,5 +830,7 @@ export default {
   checkCUDACompatibility,
   checkImageExists,
   pullImage,
+  cleanupAllContainers,
+  abortExecution,
   SANDBOX_CONFIG
 }

package/version.json

@@ -1,4 +1,4 @@
 {
-  "buildTime": "2026-05-03T02:20:30.472Z",
-  "cliVersion": "2026.5.3-1019"
+  "buildTime": "2026-05-03T04:48:50.733Z",
+  "cliVersion": "2026.5.3-1248"
 }