Merge pull request #9635 from zexi/automated-cherry-pick-of-#9634-upstream-master

Automated cherry pick of #9634: feat(ai): LLM 套餐本地路径导入与部署重启/操作增强

Author: zexi

containers/Ai/locales/en.json

@@ -60,6 +60,10 @@
   "aice.instant_app": "Inference Model Library",
   "aice.instant_app.install": "Install Model Data",
   "aice.llm_sku": "Inference Template",
+  "aice.llm_sku.source.huggingface": "Hugging Face",
+  "aice.llm_sku.source.model_scope": "ModelScope",
+  "aice.llm_sku.source.local_path": "Local Path",
+  "aice.llm_sku.source.ollama": "Ollama Registry",
   "aice.app_llm_sku": "Agent Template",
   "aice.desktop_llm": "Desktop Instances",
   "aice.desktop_llm_create": "Create Desktop Instance",
@@ -323,6 +327,8 @@
   "aice.add_port_mapping": "Add Mapping",
   "aice.preferred_model": "Preferred model",
   "aice.customized_args": "Custom arguments",
+  "aice.backend_parameters": "Custom arguments",
+  "aice.backend_parameters.help": "Parameter names become vLLM/SGLang CLI flags, e.g. max-model-len. With a value: --max-model-len=8192; without a value: --enable-auto-tool-choice",
   "aice.add_customized_arg": "Add argument",
   "aice.customized_arg_key": "Key",
   "aice.customized_arg_value": "Value",
@@ -491,6 +497,12 @@
   "aice.llm_deployment.aiproxy_sync_in_progress": "Gateway sync is in progress. Try again later.",
   "aice.llm_deployment.aiproxy_not_registered": "Not registered with AI Gateway",
   "aice.llm_deployment.scale": "Scale",
+  "aice.llm_deployment.restart": "Restart",
+  "aice.llm_deployment.restart_confirm": "This will restart all replicas under this deployment. Service may be briefly unavailable. Continue?",
+  "aice.llm_deployment.restart_disabled_invalid_status": "Restart is not allowed in current status ({0})",
+  "aice.llm_deployment.syncstatus_disabled_invalid_status": "Sync status is not allowed in current status ({0})",
+  "aice.llm_deployment.scale_disabled_invalid_status": "Scale is not allowed in current status ({0})",
+  "aice.llm_deployment.restart_disabled_no_ready_replicas": "No ready replicas; restart is unavailable",
   "aice.llm_deployment.scale.help": "Ready {0} / Desired {1}",
   "aice.llm_deployment.auto_start": "Auto Start",
   "aice.llm_deployment.create.mode": "Create Mode",
@@ -520,6 +532,17 @@
   "aice.llm_deployment.deploy.from_huggingface": "From Hugging Face",
   "aice.llm_deployment.deploy.from_modelscope": "From ModelScope",
   "aice.llm_deployment.deploy.from_local": "From Local Path",
+  "aice.local_path_import.host_model_path": "Host model path",
+  "aice.local_path_import.prefer_hosts": "Hosts",
+  "aice.local_path_import.prefer_hosts_extra": "Select hosts where the model directory exists (multiple allowed). Inherited when creating a deployment; replicas are round-robin scheduled across selected hosts",
+  "aice.local_path_import.prefer_hosts_required": "Select at least one host",
+  "aice.local_path_import.prefer_hosts_deploy_hint": "The model must exist at the same path on selected hosts",
+  "aice.local_path_import.host_model_path_extra": "Absolute path to an existing model directory on the host; mounted into the container via host_paths",
+  "aice.local_path_import.host_model_path_placeholder": "e.g. /data/models/Qwen3-8B",
+  "aice.local_path_import.path_required": "Host model path is required",
+  "aice.local_path_import.path_absolute": "Path must be an absolute path starting with /",
+  "aice.local_path_import.mount_preview": "Mount: {0} → in container {1}",
+  "aice.local_path_import.advanced_host_paths": "Advanced: edit host_paths manually",
   "aice.llm_deployment.deploy.coming_soon": "Coming soon",
   "aice.llm_deployment.deploy.selected_repo": "Selected: {0}",
   "aice.llm_deployment.deploy.change_repo": "Change",

containers/Ai/locales/ja-JP.json

@@ -59,6 +59,10 @@
   "aice.instant_app": "推論モデルライブラリ",
   "aice.instant_app.install": "モデルデータをインストール",
   "aice.llm_sku": "推論テンプレート",
+  "aice.llm_sku.source.huggingface": "Hugging Face",
+  "aice.llm_sku.source.model_scope": "ModelScope",
+  "aice.llm_sku.source.local_path": "ローカルパス",
+  "aice.llm_sku.source.ollama": "Ollama 公式リポジトリ",
   "aice.app_llm_sku": "Agentテンプレート",
   "aice.desktop_llm": "デスクトップインスタンス",
   "aice.desktop_llm_create": "デスクトップインスタンス作成",
@@ -322,6 +326,8 @@
   "aice.add_port_mapping": "マッピングを追加",
   "aice.preferred_model": "優先モデル",
   "aice.customized_args": "カスタム引数",
+  "aice.backend_parameters": "カスタム引数",
+  "aice.backend_parameters.help": "パラメータ名は vLLM/SGLang の CLI フラグになります（例: max-model-len）。値あり: --max-model-len=8192、値なし: --enable-auto-tool-choice",
   "aice.add_customized_arg": "引数を追加",
   "aice.customized_arg_key": "キー",
   "aice.customized_arg_value": "値",
@@ -490,6 +496,12 @@
   "aice.llm_deployment.aiproxy_sync_in_progress": "ゲートウェイ同期中です。しばらくしてから再試行してください。",
   "aice.llm_deployment.aiproxy_not_registered": "AI ゲートウェイ未登録",
   "aice.llm_deployment.scale": "スケール",
+  "aice.llm_deployment.restart": "再起動",
+  "aice.llm_deployment.restart_confirm": "このデプロイの全レプリカを再起動します。サービスが一時的に利用できなくなる場合があります。続行しますか？",
+  "aice.llm_deployment.restart_disabled_invalid_status": "現在の状態（{0}）では再起動できません",
+  "aice.llm_deployment.syncstatus_disabled_invalid_status": "現在の状態（{0}）では状態同期できません",
+  "aice.llm_deployment.scale_disabled_invalid_status": "現在の状態（{0}）ではレプリカ数を変更できません",
+  "aice.llm_deployment.restart_disabled_no_ready_replicas": "準備完了のレプリカがないため、再起動できません",
   "aice.llm_deployment.scale.help": "準備完了 {0} / 期待 {1}",
   "aice.llm_deployment.auto_start": "自動起動",
   "aice.llm_deployment.create.mode": "作成モード",
@@ -519,6 +531,17 @@
   "aice.llm_deployment.deploy.from_huggingface": "Hugging Face から",
   "aice.llm_deployment.deploy.from_modelscope": "ModelScope から",
   "aice.llm_deployment.deploy.from_local": "ローカルパスから",
+  "aice.local_path_import.host_model_path": "ホストモデルパス",
+  "aice.local_path_import.prefer_hosts": "ホスト",
+  "aice.local_path_import.prefer_hosts_extra": "モデルディレクトリが存在するホストを選択（複数可）。デプロイ作成時にデフォルトで継承され、レプリカは選択ホストにラウンドロビンで割り当てられます",
+  "aice.local_path_import.prefer_hosts_required": "少なくとも1台のホストを選択してください",
+  "aice.local_path_import.prefer_hosts_deploy_hint": "モデルは選択したホスト上の同じパスに存在する必要があります",
+  "aice.local_path_import.host_model_path_extra": "ホスト上の既存モデルディレクトリの絶対パス。host_paths でコンテナにマウントします",
+  "aice.local_path_import.host_model_path_placeholder": "例: /data/models/Qwen3-8B",
+  "aice.local_path_import.path_required": "ホストモデルパスを入力してください",
+  "aice.local_path_import.path_absolute": "パスは / で始まる絶対パスである必要があります",
+  "aice.local_path_import.mount_preview": "マウント: {0} → コンテナ内 {1}",
+  "aice.local_path_import.advanced_host_paths": "詳細: host_paths を手動編集",
   "aice.llm_deployment.deploy.coming_soon": "近日対応予定",
   "aice.llm_deployment.deploy.selected_repo": "選択中：{0}",
   "aice.llm_deployment.deploy.change_repo": "変更",

containers/Ai/locales/zh-CN.json

@@ -73,6 +73,10 @@
   "aice.instant_app": "推理模型库",
   "aice.instant_app.install": "安装模型数据",
   "aice.llm_sku": "推理模板",
+  "aice.llm_sku.source.huggingface": "Hugging Face",
+  "aice.llm_sku.source.model_scope": "ModelScope",
+  "aice.llm_sku.source.local_path": "本地路径",
+  "aice.llm_sku.source.ollama": "Ollama 官方仓库",
   "aice.app_llm_sku": "Agent模板",
   "aice.desktop_llm": "桌面实例",
   "aice.desktop_llm_create": "创建桌面实例",
@@ -336,6 +340,8 @@
   "aice.add_port_mapping": "新增映射",
   "aice.preferred_model": "首选模型标识",
   "aice.customized_args": "自定义参数",
+  "aice.backend_parameters": "自定义参数",
+  "aice.backend_parameters.help": "参数名将作为 vLLM/SGLang CLI 标志，例如 max-model-len；有值时提交为 --max-model-len=8192，无值时提交为 --enable-auto-tool-choice",
   "aice.add_customized_arg": "新增参数",
   "aice.customized_arg_key": "参数名",
   "aice.customized_arg_value": "参数值",
@@ -494,6 +500,12 @@
   "aice.llm_deployment.aiproxy_sync_in_progress": "网关同步进行中，请稍后再试",
   "aice.llm_deployment.aiproxy_not_registered": "尚未注册 AI 网关",
   "aice.llm_deployment.scale": "调整副本数",
+  "aice.llm_deployment.restart": "重启",
+  "aice.llm_deployment.restart_confirm": "将重启该部署下的全部副本，期间服务可能短暂不可用，是否继续？",
+  "aice.llm_deployment.restart_disabled_invalid_status": "当前状态（{0}）不允许重启",
+  "aice.llm_deployment.syncstatus_disabled_invalid_status": "当前状态（{0}）不允许同步状态",
+  "aice.llm_deployment.scale_disabled_invalid_status": "当前状态（{0}）不允许调整副本数",
+  "aice.llm_deployment.restart_disabled_no_ready_replicas": "暂无就绪副本，无法重启",
   "aice.llm_deployment.scale.help": "当前就绪 {0} / 期望 {1}",
   "aice.llm_deployment.auto_start": "自动启动",
   "aice.llm_deployment.create.mode": "创建模式",
@@ -523,6 +535,17 @@
   "aice.llm_deployment.deploy.from_huggingface": "从 Hugging Face",
   "aice.llm_deployment.deploy.from_modelscope": "从 ModelScope",
   "aice.llm_deployment.deploy.from_local": "从本地路径",
+  "aice.local_path_import.host_model_path": "宿主机模型路径",
+  "aice.local_path_import.prefer_hosts": "宿主机",
+  "aice.local_path_import.prefer_hosts_extra": "选择模型目录所在的宿主机，可多选；创建部署时默认继承，多副本将轮询分配到所选宿主机",
+  "aice.local_path_import.prefer_hosts_required": "请至少选择一个宿主机",
+  "aice.local_path_import.prefer_hosts_deploy_hint": "模型须存在于所选宿主机上的相同路径",
+  "aice.local_path_import.host_model_path_extra": "填写宿主机上已存在的模型目录绝对路径，将通过 host_paths 挂载到推理容器",
+  "aice.local_path_import.host_model_path_placeholder": "例如 /data/models/Qwen3-8B",
+  "aice.local_path_import.path_required": "请输入宿主机模型路径",
+  "aice.local_path_import.path_absolute": "路径须以 / 开头的绝对路径",
+  "aice.local_path_import.mount_preview": "将挂载：{0} → 容器内 {1}",
+  "aice.local_path_import.advanced_host_paths": "高级：手动编辑 host_paths",
   "aice.llm_deployment.deploy.coming_soon": "敬请期待",
   "aice.llm_deployment.deploy.selected_repo": "已选择：{0}",
   "aice.llm_deployment.deploy.change_repo": "重新选择",

containers/Ai/router/index.js

@@ -7,6 +7,7 @@ const LlmSku = () => import(/* webpackChunkName: "k8s" */ /* webpackPrefetch: tr
 const LlmSkuCreate = () => import(/* webpackChunkName: "k8s" */ /* webpackPrefetch: true */ '@Ai/views/llm-sku/create/index')
 const LlmSkuImportFromModelSets = () => import(/* webpackChunkName: "k8s" */ /* webpackPrefetch: true */ '@Ai/views/llm-sku/import-from-model-sets')
 const LlmSkuImportFromHuggingFace = () => import(/* webpackChunkName: "k8s" */ /* webpackPrefetch: true */ '@Ai/views/llm-sku/import-from-huggingface')
+const LlmSkuImportFromLocalPath = () => import(/* webpackChunkName: "k8s" */ /* webpackPrefetch: true */ '@Ai/views/llm-sku/import-from-local-path')
 const LlmSkuImportFromCommunity = () => import(/* webpackChunkName: "k8s" */ /* webpackPrefetch: true */ '@Ai/views/llm-sku/import-from-community')
 const LlmInstantmodel = () => import(/* webpackChunkName: "k8s" */ /* webpackPrefetch: true */ '@Ai/views/llm-instantmodel')
 const LlmImage = () => import(/* webpackChunkName: "k8s" */ /* webpackPrefetch: true */ '@Ai/views/llm-image')
@@ -206,6 +207,11 @@ export default {
               path: 'import-from-huggingface',
               component: LlmSkuImportFromHuggingFace,
             },
+            {
+              name: 'LlmSkuImportFromLocalPath',
+              path: 'import-from-local-path',
+              component: LlmSkuImportFromLocalPath,
+            },
             {
               name: 'LlmSkuCreate',
               path: 'create',

containers/Ai/sections/LlmSkuSelect.vue

@@ -35,6 +35,7 @@
 <script>
 import { sizestr } from '@/utils/utils'
 import { getSkuModelDisplayText } from '@Ai/views/llm-sku/utils/modelDisplay'
+import { getLlmSkuSourceLabel } from '@Ai/views/llm-sku/utils/skuSourceDisplay'
 
 export default {
   name: 'LlmSkuSelect',
@@ -115,6 +116,9 @@ export default {
       if (item.llm_type) {
         parts.push(`${this.$t('aice.llm_type')}：${item.llm_type}`)
       }
+      if (item.source) {
+        parts.push(`${this.$t('aice.llm_deployment.source')}：${getLlmSkuSourceLabel(item.source)}`)
+      }
       if (item.app_name) {
         parts.push(`${this.$t('aice.llm_image.app_name')}：${item.app_name}`)
       }

containers/Ai/utils/aiproxyDeploymentActions.js

@@ -1,14 +1,40 @@
-export function formatDeploymentAiproxyStatus (status, vm) {
+export function formatDeploymentStatus (status, vm) {
   if (!status) return '-'
   const key = `status.llmDeployment.${status}`
   const text = vm.$t(key)
   return text === key ? status : text
 }
 
+export function formatAiproxySyncStatus (status, vm) {
+  if (!status) return '-'
+  const key = `status.llmDeploymentAiproxy.${status}`
+  const text = vm.$t(key)
+  return text === key ? status : text
+}
+
+// Prefer aiproxy_sync_status; fall back to legacy status values during upgrade.
+export function resolveAiproxySyncStatus (row) {
+  const syncStatus = row?.aiproxy_sync_status
+  if (syncStatus) return syncStatus
+  const legacy = row?.status
+  switch (legacy) {
+    case 'aiproxy_pending':
+      return 'pending'
+    case 'aiproxy_syncing':
+      return 'syncing'
+    case 'aiproxy_partial':
+      return 'partial'
+    case 'aiproxy_sync_failed':
+      return 'failed'
+    default:
+      return row?.auto_register_aiproxy === false ? 'disabled' : ''
+  }
+}
+
 export function isAiproxySyncInProgress (row) {
-  const status = row?.status
-  if (status === 'aiproxy_syncing') return true
-  if (status === 'aiproxy_pending' && row?.auto_register_aiproxy !== false) return true
+  const status = resolveAiproxySyncStatus(row)
+  if (status === 'syncing') return true
+  if (status === 'pending' && row?.auto_register_aiproxy !== false) return true
   return false
 }
 
@@ -21,7 +47,44 @@ export function isAiproxyRegistered (row) {
   return true
 }
 
+const CREATING_STATUSES = [
+  'creating',
+  'importing_model',
+  'creating_sku',
+  'deleting',
+  'import_model_failed',
+  'create_sku_failed',
+  'create_fail',
+  'delete_fail',
+]
+
+export function getScaleMeta (row, vm) {
+  if (CREATING_STATUSES.includes(row?.status)) {
+    return {
+      validate: false,
+      tooltip: vm.$t('aice.llm_deployment.scale_disabled_invalid_status', [
+        formatDeploymentStatus(row?.status, vm),
+      ]),
+    }
+  }
+  if (isAiproxySyncInProgress(row)) {
+    return {
+      validate: false,
+      tooltip: vm.$t('aice.llm_deployment.aiproxy_sync_in_progress'),
+    }
+  }
+  return { validate: true }
+}
+
 export function getRegisterAiproxyMeta (row, vm) {
+  if (CREATING_STATUSES.includes(row?.status)) {
+    return {
+      validate: false,
+      tooltip: vm.$t('aice.llm_deployment.scale_disabled_invalid_status', [
+        formatDeploymentStatus(row?.status, vm),
+      ]),
+    }
+  }
   if (isAiproxySyncInProgress(row)) {
     return {
       validate: false,
@@ -47,6 +110,57 @@ export function getUnregisterAiproxyMeta (row, vm) {
   return { validate: true }
 }
 
+const RESTARTABLE_STATUSES = ['ready', 'partial', 'running', 'start_fail']
+
+export function getRestartMeta (row, vm) {
+  if (CREATING_STATUSES.includes(row?.status)) {
+    return {
+      validate: false,
+      tooltip: vm.$t('aice.llm_deployment.restart_disabled_invalid_status', [
+        formatDeploymentStatus(row?.status, vm),
+      ]),
+    }
+  }
+  if (isAiproxySyncInProgress(row)) {
+    return {
+      validate: false,
+      tooltip: vm.$t('aice.llm_deployment.aiproxy_sync_in_progress'),
+    }
+  }
+  if (!RESTARTABLE_STATUSES.includes(row?.status)) {
+    const statusLabel = formatDeploymentStatus(row?.status, vm)
+    return {
+      validate: false,
+      tooltip: vm.$t('aice.llm_deployment.restart_disabled_invalid_status', [statusLabel]),
+    }
+  }
+  return { validate: true }
+}
+
+const SYNC_FORBIDDEN_STATUSES = [
+  'creating',
+  'importing_model',
+  'creating_sku',
+  'deleting',
+  'start_delete',
+  'deleted',
+  'delete_fail',
+  'delete_failed',
+  'syncing',
+]
+
+export function getSyncStatusMeta (row, vm) {
+  if (SYNC_FORBIDDEN_STATUSES.includes(row?.status)) {
+    return {
+      validate: false,
+      tooltip: vm.$t('aice.llm_deployment.syncstatus_disabled_invalid_status', [
+        formatDeploymentStatus(row?.status, vm),
+      ]),
+    }
+  }
+  return { validate: true }
+}
+
 const POST_ACTION_POLL_MS = 2000
 const POST_ACTION_MAX_POLLS = 15