Action and opensandbox deploy (#6572)

* action

* action

* action

* build: integrate OpenSandbox as Agent Execution Sandbox (#6490)

* Update action (#6571)

* action

* action

* action

* action

* action

* build: integrate OpenSandbox as Agent Execution Sandbox

# Conflicts:
#	deploy/args.json
#	deploy/dev/docker-compose.cn.yml
#	deploy/dev/docker-compose.yml
#	deploy/docker/cn/docker-compose.milvus.yml
#	deploy/docker/cn/docker-compose.oceanbase.yml
#	deploy/docker/cn/docker-compose.pg.yml
#	deploy/docker/cn/docker-compose.seekdb.yml
#	deploy/docker/cn/docker-compose.zilliz.yml
#	deploy/docker/global/docker-compose.milvus.yml
#	deploy/docker/global/docker-compose.oceanbase.yml
#	deploy/docker/global/docker-compose.pg.yml
#	deploy/docker/global/docker-compose.seekdb.yml
#	deploy/docker/global/docker-compose.ziliiz.yml
#	deploy/templates/docker-compose.prod.yml
#	document/public/deploy/docker/cn/docker-compose.milvus.yml
#	document/public/deploy/docker/cn/docker-compose.oceanbase.yml
#	document/public/deploy/docker/cn/docker-compose.pg.yml
#	document/public/deploy/docker/cn/docker-compose.seekdb.yml
#	document/public/deploy/docker/cn/docker-compose.zilliz.yml
#	document/public/deploy/docker/global/docker-compose.milvus.yml
#	document/public/deploy/docker/global/docker-compose.oceanbase.yml
#	document/public/deploy/docker/global/docker-compose.pg.yml
#	document/public/deploy/docker/global/docker-compose.seekdb.yml
#	document/public/deploy/docker/global/docker-compose.ziliiz.yml

* remove invalid  action

---------

Co-authored-by: Archer <545436317@qq.com>
Co-authored-by: xqvvu <whoeverimf5@gmail.com>

* action

---------

Co-authored-by: chanzany <chenzhi@sangfor.com.cn>
Co-authored-by: xqvvu <whoeverimf5@gmail.com>

.github/workflows/preview-docs-build.yml

@@ -5,18 +5,9 @@ on:
     paths:
     paths:
       - 'document/**'
       - 'document/**'
     types: [opened, synchronize, reopened]
     types: [opened, synchronize, reopened]
-  pull_request_target:
-    paths:
-      - 'document/**'
-    types: [opened, synchronize, reopened]
 
 
 jobs:
 jobs:
   build-docs-image:
   build-docs-image:
-    # 内部和外部贡献者都可以触发构建
-    if: |
-      (github.event_name == 'pull_request') ||
-      (github.event_name == 'pull_request_target')
-
     permissions:
     permissions:
       contents: read
       contents: read
       pull-requests: write
       pull-requests: write
@@ -26,9 +17,6 @@ jobs:
     steps:
     steps:
       - name: Checkout PR code
       - name: Checkout PR code
         uses: actions/checkout@v4
         uses: actions/checkout@v4
-        with:
-          # 对于 pull_request_target，检出 PR 的代码
-          ref: ${{ github.event_name == 'pull_request_target' && github.event.pull_request.head.sha || github.sha }}
 
 
       - name: Get current datetime
       - name: Get current datetime
         id: datetime
         id: datetime
@@ -71,7 +59,11 @@ jobs:
       pull-requests: write
       pull-requests: write
       issues: write
       issues: write
     uses: ./.github/workflows/preview-docs-push.yml
     uses: ./.github/workflows/preview-docs-push.yml
-    secrets: inherit
+    secrets:
+      ALI_IMAGE_USER: ${{ secrets.FASTGPT_ALI_IMAGE_USER }}
+      ALI_IMAGE_PSW: ${{ secrets.FASTGPT_ALI_IMAGE_PSW }}
+      ALI_IMAGE_PREFIX: ${{ secrets.FASTGPT_ALI_IMAGE_PREFIX }}
+      KUBE_CONFIG_CN: ${{ secrets.KUBE_CONFIG_CN }}
     with:
     with:
       pr_number: ${{ format('{0}', github.event.pull_request.number) }}
       pr_number: ${{ format('{0}', github.event.pull_request.number) }}
       datetime: ${{ needs.build-docs-image.outputs.datetime }}
       datetime: ${{ needs.build-docs-image.outputs.datetime }}

.github/workflows/preview-docs-push.yml

@@ -12,6 +12,15 @@ on:
       run_id:
       run_id:
         required: true
         required: true
         type: string
         type: string
+    secrets:
+      ALI_IMAGE_USER:
+        required: true
+      ALI_IMAGE_PSW:
+        required: true
+      ALI_IMAGE_PREFIX:
+        required: true
+      KUBE_CONFIG_CN:
+        required: true
 
 
 permissions:
 permissions:
   contents: read
   contents: read
@@ -55,14 +64,14 @@ jobs:
         uses: docker/login-action@v3
         uses: docker/login-action@v3
         with:
         with:
           registry: registry.cn-hangzhou.aliyuncs.com
           registry: registry.cn-hangzhou.aliyuncs.com
-          username: ${{ secrets.FASTGPT_ALI_IMAGE_USER }}
-          password: ${{ secrets.FASTGPT_ALI_IMAGE_PSW }}
+          username: ${{ secrets.ALI_IMAGE_USER }}
+          password: ${{ secrets.ALI_IMAGE_PSW }}
 
 
       - name: Tag and push image
       - name: Tag and push image
         run: |
         run: |
           docker tag fastgpt-docs:${{ steps.artifacts.outputs.datetime }} \
           docker tag fastgpt-docs:${{ steps.artifacts.outputs.datetime }} \
-            ${{ secrets.FASTGPT_ALI_IMAGE_PREFIX }}/fastgpt-docs:${{ steps.artifacts.outputs.datetime }}
-          docker push ${{ secrets.FASTGPT_ALI_IMAGE_PREFIX }}/fastgpt-docs:${{ steps.artifacts.outputs.datetime }}
+            ${{ secrets.ALI_IMAGE_PREFIX }}/fastgpt-docs:${{ steps.artifacts.outputs.datetime }}
+          docker push ${{ secrets.ALI_IMAGE_PREFIX }}/fastgpt-docs:${{ steps.artifacts.outputs.datetime }}
 
 
       - name: Setup kubeconfig
       - name: Setup kubeconfig
         run: |
         run: |
@@ -73,12 +82,12 @@ jobs:
       - name: Update deployment image
       - name: Update deployment image
         run: |
         run: |
           kubectl set image deployment/fastgpt-docs-preview \
           kubectl set image deployment/fastgpt-docs-preview \
-            fastgpt-docs-preview=${{ secrets.FASTGPT_ALI_IMAGE_PREFIX }}/fastgpt-docs:${{ steps.artifacts.outputs.datetime }}
+            fastgpt-docs-preview=${{ secrets.ALI_IMAGE_PREFIX }}/fastgpt-docs:${{ steps.artifacts.outputs.datetime }}
 
 
       - name: Annotate deployment
       - name: Annotate deployment
         run: |
         run: |
           kubectl annotate deployment/fastgpt-docs-preview \
           kubectl annotate deployment/fastgpt-docs-preview \
-            originImageName="${{ secrets.FASTGPT_ALI_IMAGE_PREFIX }}/fastgpt-docs:${{ steps.artifacts.outputs.datetime }}" --overwrite
+            originImageName="${{ secrets.ALI_IMAGE_PREFIX }}/fastgpt-docs:${{ steps.artifacts.outputs.datetime }}" --overwrite
 
 
       - name: Comment deployment status
       - name: Comment deployment status
         uses: FinleyGe/github-tools@0.0.1
         uses: FinleyGe/github-tools@0.0.1

.github/workflows/preview-fastgpt-build.yml

@@ -2,19 +2,10 @@ name: Preview fastgpt build
 
 
 on:
 on:
   pull_request:
   pull_request:
-    # 支持所有分支
-    types: [opened, synchronize, reopened]
-  pull_request_target:
-    # 外部贡献者也支持自动构建
     types: [opened, synchronize, reopened]
     types: [opened, synchronize, reopened]
 
 
 jobs:
 jobs:
   build-preview-images:
   build-preview-images:
-    # 内部和外部贡献者都可以触发构建
-    if: |
-      (github.event_name == 'pull_request') ||
-      (github.event_name == 'pull_request_target')
-
     permissions:
     permissions:
       contents: read
       contents: read
       pull-requests: write
       pull-requests: write
@@ -29,8 +20,6 @@ jobs:
       - name: Checkout PR code
       - name: Checkout PR code
         uses: actions/checkout@v4
         uses: actions/checkout@v4
         with:
         with:
-          # 对于 pull_request_target，检出 PR 的代码
-          ref: ${{ github.event_name == 'pull_request_target' && github.event.pull_request.head.sha || github.sha }}
           fetch-depth: 0
           fetch-depth: 0
 
 
       - name: Set up Docker Buildx
       - name: Set up Docker Buildx
@@ -103,7 +92,10 @@ jobs:
       pull-requests: write
       pull-requests: write
       issues: write
       issues: write
     uses: ./.github/workflows/preview-fastgpt-push.yml
     uses: ./.github/workflows/preview-fastgpt-push.yml
-    secrets: inherit
+    secrets:
+      ALI_IMAGE_USER: ${{ secrets.FASTGPT_ALI_IMAGE_USER }}
+      ALI_IMAGE_PSW: ${{ secrets.FASTGPT_ALI_IMAGE_PSW }}
+      ALI_IMAGE_PREFIX: ${{ secrets.FASTGPT_ALI_IMAGE_PREFIX }}
     with:
     with:
       pr_number: ${{ format('{0}', github.event.pull_request.number) }}
       pr_number: ${{ format('{0}', github.event.pull_request.number) }}
       pr_sha: ${{ github.sha }}
       pr_sha: ${{ github.sha }}

.github/workflows/preview-fastgpt-push.yml

@@ -15,6 +15,21 @@ on:
       image:
       image:
         required: true
         required: true
         type: string
         type: string
+    secrets:
+      ALI_IMAGE_USER:
+        required: true
+      ALI_IMAGE_PSW:
+        required: true
+      ALI_IMAGE_PREFIX:
+        required: true
+
+permissions:
+  contents: read
+  packages: write
+  attestations: write
+  id-token: write
+  pull-requests: write
+  issues: write
 
 
 permissions:
 permissions:
   contents: read
   contents: read
@@ -45,15 +60,15 @@ jobs:
           if [[ "${{ inputs.image }}" == "fastgpt" ]]; then
           if [[ "${{ inputs.image }}" == "fastgpt" ]]; then
             echo "IMAGE_NAME=fastgpt" >> $GITHUB_OUTPUT
             echo "IMAGE_NAME=fastgpt" >> $GITHUB_OUTPUT
             echo "DESCRIPTION=fastgpt-pr image" >> $GITHUB_OUTPUT
             echo "DESCRIPTION=fastgpt-pr image" >> $GITHUB_OUTPUT
-            echo "DOCKER_REPO_TAGGED=${{ secrets.FASTGPT_ALI_IMAGE_PREFIX }}/fastgpt-pr:fastgpt_${SHA}" >> $GITHUB_OUTPUT
+            echo "DOCKER_REPO_TAGGED=${{ secrets.ALI_IMAGE_PREFIX }}/fastgpt-pr:fastgpt_${SHA}" >> $GITHUB_OUTPUT
           elif [[ "${{ inputs.image }}" == "sandbox" ]]; then
           elif [[ "${{ inputs.image }}" == "sandbox" ]]; then
             echo "IMAGE_NAME=fastgpt-sandbox" >> $GITHUB_OUTPUT
             echo "IMAGE_NAME=fastgpt-sandbox" >> $GITHUB_OUTPUT
             echo "DESCRIPTION=fastgpt-sandbox-pr image" >> $GITHUB_OUTPUT
             echo "DESCRIPTION=fastgpt-sandbox-pr image" >> $GITHUB_OUTPUT
-            echo "DOCKER_REPO_TAGGED=${{ secrets.FASTGPT_ALI_IMAGE_PREFIX }}/fastgpt-pr:fastgpt_sandbox_${SHA}" >> $GITHUB_OUTPUT
+            echo "DOCKER_REPO_TAGGED=${{ secrets.ALI_IMAGE_PREFIX }}/fastgpt-pr:fastgpt_sandbox_${SHA}" >> $GITHUB_OUTPUT
           elif [[ "${{ inputs.image }}" == "mcp_server" ]]; then
           elif [[ "${{ inputs.image }}" == "mcp_server" ]]; then
             echo "IMAGE_NAME=fastgpt-mcp-server" >> $GITHUB_OUTPUT
             echo "IMAGE_NAME=fastgpt-mcp-server" >> $GITHUB_OUTPUT
             echo "DESCRIPTION=fastgpt-mcp_server-pr image" >> $GITHUB_OUTPUT
             echo "DESCRIPTION=fastgpt-mcp_server-pr image" >> $GITHUB_OUTPUT
-            echo "DOCKER_REPO_TAGGED=${{ secrets.FASTGPT_ALI_IMAGE_PREFIX }}/fastgpt-pr:fastgpt_mcp_server_${SHA}" >> $GITHUB_OUTPUT
+            echo "DOCKER_REPO_TAGGED=${{ secrets.ALI_IMAGE_PREFIX }}/fastgpt-pr:fastgpt_mcp_server_${SHA}" >> $GITHUB_OUTPUT
           fi
           fi
 
 
       - name: Download image artifact
       - name: Download image artifact
@@ -84,8 +99,8 @@ jobs:
         uses: docker/login-action@v3
         uses: docker/login-action@v3
         with:
         with:
           registry: registry.cn-hangzhou.aliyuncs.com
           registry: registry.cn-hangzhou.aliyuncs.com
-          username: ${{ secrets.FASTGPT_ALI_IMAGE_USER }}
-          password: ${{ secrets.FASTGPT_ALI_IMAGE_PSW }}
+          username: ${{ secrets.ALI_IMAGE_USER }}
+          password: ${{ secrets.ALI_IMAGE_PSW }}
 
 
       - name: Tag and push image
       - name: Tag and push image
         run: |
         run: |

deploy/args.json

@@ -14,7 +14,10 @@
     "milvus-etcd": "v3.5.5",
     "milvus-etcd": "v3.5.5",
     "milvus-standalone": "v2.4.3",
     "milvus-standalone": "v2.4.3",
     "oceanbase": "4.3.5-lts",
     "oceanbase": "4.3.5-lts",
-    "seekdb": "1.0.1.0-100000392025122619"
+    "seekdb": "1.0.1.0-100000392025122619",
+    "opensandbox-server": "v0.1.7",
+    "opensandbox-execd": "v1.0.7",
+    "opensandbox-egress": "v1.0.1"
   },
   },
   "images": {
   "images": {
     "cn": {
     "cn": {
@@ -32,7 +35,10 @@
       "milvus-etcd": "quay.io/coreos/etcd",
       "milvus-etcd": "quay.io/coreos/etcd",
       "milvus-standalone": "milvusdb/milvus",
       "milvus-standalone": "milvusdb/milvus",
       "oceanbase": "oceanbase/oceanbase-ce",
       "oceanbase": "oceanbase/oceanbase-ce",
-      "seekdb": "oceanbase/seekdb"
+      "seekdb": "oceanbase/seekdb",
+      "opensandbox-server": "sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/server",
+      "opensandbox-execd": "sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd",
+      "opensandbox-egress": "sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/egress"
     },
     },
     "global": {
     "global": {
       "fastgpt": "ghcr.io/labring/fastgpt",
       "fastgpt": "ghcr.io/labring/fastgpt",
@@ -49,7 +55,10 @@
       "milvus-etcd": "quay.io/coreos/etcd",
       "milvus-etcd": "quay.io/coreos/etcd",
       "milvus-standalone": "milvusdb/milvus",
       "milvus-standalone": "milvusdb/milvus",
       "oceanbase": "oceanbase/oceanbase-ce",
       "oceanbase": "oceanbase/oceanbase-ce",
-      "seekdb": "oceanbase/seekdb"
+      "seekdb": "oceanbase/seekdb",
+      "opensandbox-server": "opensandbox/server",
+      "opensandbox-execd": "opensandbox/execd",
+      "opensandbox-egress": "opensandbox/egress"
     }
     }
   }
   }
 }
 }

deploy/helm/opensandbox/.helmignore

+# Patterns to ignore when building packages.
+# This supports shell glob matching, relative path matching, and
+# negation (prefixed with !). Only one pattern per line.
+.DS_Store
+# Common VCS dirs
+.git/
+.gitignore
+.bzr/
+.bzrignore
+.hg/
+.hgignore
+.svn/
+# Common backup files
+*.swp
+*.bak
+*.tmp
+*.orig
+*~
+# Various IDEs
+.project
+.idea/
+*.tmproj
+.vscode/

deploy/helm/opensandbox/CHANGELOG.md

+# Changelog
+
+All notable changes to the OpenSandbox Helm Chart will be documented in this file.
+
+The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
+and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
+
+## [0.1.0] - Initial Release
+
+### Added
+
+#### Core Features
+- OpenSandbox Kubernetes Controller Helm chart
+- **OpenSandbox Server deployment with FastAPI control plane for SDK integration**
+- Support for deploying controller with configurable replicas and resources
+- BatchSandbox and Pool CRD definitions
+- RBAC resources (ClusterRole, ClusterRoleBinding, ServiceAccount)
+- Leader election configuration for high availability
+
+#### Server Features
+- Server Deployment with configurable replicas and resources
+- Server Service with ClusterIP/NodePort/LoadBalancer support
+- ConfigMap-based configuration management
+- Optional Ingress support for external access
+- Health probes for liveness and readiness checks
+- In-cluster Kubernetes configuration
+- API key authentication support (optional)
+- SDK-compatible REST API on port 8080
+
+#### Pool Management
+- Default agent-pool with execd and task-executor sidecar
+- Pool template support for creating pre-warmed Pod pools
+- Configurable Pool capacity (bufferMin, bufferMax, poolMin, poolMax)
+- SDK-compatible Pool configuration with execd on port 44772
+
+#### Multiple Values Files
+- `values.yaml` - Default configuration with agent-pool enabled
+- `values-e2e.yaml` - End-to-end testing with minimal resources (2-5 pods)
+- Use `--set` or custom values files for production/development overrides
+
+#### Templates
+- `deployment.yaml` - Controller manager deployment
+- **`server-deployment.yaml` - Server deployment**
+- **`server-service.yaml` - Server service**
+- **`server-configmap.yaml` - Server configuration**
+- **`server-ingress.yaml` - Server ingress (optional)**
+- `pools.yaml` - Dynamic Pool resource generation from values
+- `serviceaccount.yaml` - Service account for controller
+- `clusterrole.yaml` - RBAC cluster role
+- `clusterrolebinding.yaml` - RBAC cluster role binding
+- `leader-election-role.yaml` - Leader election RBAC
+- `leader-election-rolebinding.yaml` - Leader election binding
+- `metrics-service.yaml` - Metrics service endpoint
+- `metrics-rbac.yaml` - Metrics RBAC resources
+- `servicemonitor.yaml` - Prometheus ServiceMonitor (optional)
+- `extra-roles.yaml` - User management roles (viewer, editor)
+- `poddisruptionbudget.yaml` - High availability Pod disruption budget
+- `NOTES.txt` - Post-installation guidance
+- `_helpers.tpl` - Template helper functions
+
+#### Scripts
+- `scripts/install.sh` - Interactive installation wizard with environment selection
+- `scripts/uninstall.sh` - Safe uninstallation with resource cleanup
+- `scripts/e2e-test.sh` - End-to-end validation (Install → Server → Pool → SDK → Uninstall)
+- `scripts/README.md` - Comprehensive script documentation and troubleshooting guide
+
+#### Configuration Options
+- `nameOverride` and `fullnameOverride` for custom resource naming
+- **`server.enabled` - Enable/disable server deployment (default: true)**
+- **`server.service.type` - Service type (ClusterIP/NodePort/LoadBalancer)**
+- **`server.service.nodePort` - NodePort value (optional)**
+- **`server.ingress.enabled` - Enable Ingress for external access**
+- **`server.config.server.apiKey` - Optional API key authentication**
+- `healthProbePort` - Configurable health check port (default: 8081)
+- `healthProbes.liveness` - Liveness probe timing configuration
+- `healthProbes.readiness` - Readiness probe timing configuration
+- `podDisruptionBudget.enabled` - Optional PDB for HA deployments
+- `namespaceOverride` - Custom namespace (default: opensandbox)
+
+#### Documentation
+- Comprehensive README.md with installation and configuration guide
+- examples/README.md with usage scenarios and best practices
+- examples/pool-agent-production.yaml with production-ready Pool configuration
+- examples/DIRECTORY_STRUCTURE.md explaining file organization
+- Example YAML files for Pool and BatchSandbox resources
+
+### Configuration Defaults
+- Controller image: `opensandbox/controller:dev`
+- **Server image: `opensandbox/server:v0.1.0`**
+- Task executor image: `opensandbox/task-executor:dev`
+- Image pull policy: `Never` (for local development)
+- Namespace: `opensandbox`
+- Controller replicas: 1 (3 in production values)
+- **Server replicas: 1**
+- **Server enabled: true (required for SDK usage)**
+- Default Pool enabled: `agent-pool` with 2-5 pods (E2E) or 10-100 pods (default)
+
+### Notes
+- This is the initial release of the Helm chart
+- All templates have been tested with `helm lint` and E2E validation
+- Chart supports Kubernetes 1.19+

deploy/helm/opensandbox/Chart.yaml

+apiVersion: v2
+name: opensandbox-controller
+description: A Helm chart for deploying OpenSandbox Kubernetes Controller
+type: application
+version: 0.1.0
+appVersion: "0.0.1"
+keywords:
+  - opensandbox
+  - sandbox
+  - kubernetes
+  - operator
+  - controller
+home: https://github.com/alibaba/OpenSandbox
+maintainers:
+  - name: OpenSandbox Team
+    url: https://github.com/alibaba/OpenSandbox
+icon: https://avatars.githubusercontent.com/u/1961952

deploy/helm/opensandbox/Makefile

+# Helm Chart Makefile for OpenSandbox Controller
+
+CHART_NAME := opensandbox-controller
+RELEASE_NAME := opensandbox-controller
+NAMESPACE := opensandbox
+VALUES_FILE := values.yaml
+
+# Helm commands
+HELM := helm
+KUBECTL := kubectl
+
+.PHONY: help
+help: ## Display this help
+	@awk 'BEGIN {FS = ":.*##"; printf "\nUsage:\n  make \033[36m<target>\033[0m\n\nTargets:\n"} /^[a-zA-Z_-]+:.*?##/ { printf "  \033[36m%-15s\033[0m %s\n", $$1, $$2 }' $(MAKEFILE_LIST)
+
+.PHONY: lint
+lint: ## Lint the Helm chart
+	$(HELM) lint .
+
+.PHONY: template
+template: ## Render chart templates to stdout
+	$(HELM) template $(RELEASE_NAME) . -f $(VALUES_FILE)
+
+.PHONY: template-debug
+template-debug: ## Render chart templates with debug output
+	$(HELM) template $(RELEASE_NAME) . -f $(VALUES_FILE) --debug
+
+.PHONY: dry-run
+dry-run: ## Perform a dry-run installation
+	$(HELM) install $(RELEASE_NAME) . \
+		-f $(VALUES_FILE) \
+		--namespace $(NAMESPACE) \
+		--create-namespace \
+		--dry-run --debug
+
+.PHONY: install
+install: ## Install the chart
+	$(HELM) install $(RELEASE_NAME) . \
+		-f $(VALUES_FILE) \
+		--namespace $(NAMESPACE) \
+		--create-namespace
+
+.PHONY: install-e2e
+install-e2e: ## Install the chart with e2e test values
+	$(HELM) install $(RELEASE_NAME) . \
+		-f values-e2e.yaml \
+		--namespace $(NAMESPACE) \
+		--create-namespace
+
+.PHONY: upgrade
+upgrade: ## Upgrade the chart
+	$(HELM) upgrade $(RELEASE_NAME) . \
+		-f $(VALUES_FILE) \
+		--namespace $(NAMESPACE)
+
+.PHONY: upgrade-e2e
+upgrade-e2e: ## Upgrade with e2e test values
+	$(HELM) upgrade $(RELEASE_NAME) . \
+		-f values-e2e.yaml \
+		--namespace $(NAMESPACE)
+
+.PHONY: uninstall
+uninstall: ## Uninstall the chart
+	$(HELM) uninstall $(RELEASE_NAME) --namespace $(NAMESPACE)
+
+.PHONY: status
+status: ## Show release status
+	$(HELM) status $(RELEASE_NAME) --namespace $(NAMESPACE)
+
+.PHONY: list
+list: ## List all releases
+	$(HELM) list --namespace $(NAMESPACE)
+
+.PHONY: get-values
+get-values: ## Get values for the release
+	$(HELM) get values $(RELEASE_NAME) --namespace $(NAMESPACE)
+
+.PHONY: get-all
+get-all: ## Get all information about the release
+	$(HELM) get all $(RELEASE_NAME) --namespace $(NAMESPACE)
+
+.PHONY: package
+package: ## Package the chart into an archive
+	$(HELM) package .
+
+.PHONY: verify-install
+verify-install: ## Verify the installation
+	@echo "Checking deployment..."
+	$(KUBECTL) get deployment -n $(NAMESPACE)
+	@echo "\nChecking pods..."
+	$(KUBECTL) get pods -n $(NAMESPACE)
+	@echo "\nChecking CRDs..."
+	$(KUBECTL) get crds | grep sandbox.opensandbox.io
+
+.PHONY: logs
+logs: ## Show controller logs
+	$(KUBECTL) logs -n $(NAMESPACE) -l control-plane=controller-manager -f
+
+.PHONY: clean-crds
+clean-crds: ## Delete CRDs (use with caution!)
+	$(KUBECTL) delete crd batchsandboxes.sandbox.opensandbox.io
+	$(KUBECTL) delete crd pools.sandbox.opensandbox.io
+
+.PHONY: test-connection
+test-connection: ## Test if kubectl can connect to the cluster
+	$(KUBECTL) cluster-info
+	$(KUBECTL) get nodes
+
+.PHONY: create-namespace
+create-namespace: ## Create the namespace
+	$(KUBECTL) create namespace $(NAMESPACE) --dry-run=client -o yaml | $(KUBECTL) apply -f -
+
+.PHONY: delete-namespace
+delete-namespace: ## Delete the namespace
+	$(KUBECTL) delete namespace $(NAMESPACE)
+
+# Advanced targets
+.PHONY: diff
+diff: ## Show diff between current release and chart
+	@command -v helm-diff >/dev/null 2>&1 || { echo "helm-diff plugin required. Install: helm plugin install https://github.com/databus23/helm-diff"; exit 1; }
+	$(HELM) diff upgrade $(RELEASE_NAME) . -f $(VALUES_FILE) --namespace $(NAMESPACE)
+
+.PHONY: history
+history: ## Show release history
+	$(HELM) history $(RELEASE_NAME) --namespace $(NAMESPACE)
+
+.PHONY: rollback
+rollback: ## Rollback to previous release
+	$(HELM) rollback $(RELEASE_NAME) --namespace $(NAMESPACE)
+
+.PHONY: rollback-to
+rollback-to: ## Rollback to specific revision (usage: make rollback-to REVISION=2)
+	@if [ -z "$(REVISION)" ]; then echo "Please specify REVISION=<number>"; exit 1; fi
+	$(HELM) rollback $(RELEASE_NAME) $(REVISION) --namespace $(NAMESPACE)

deploy/helm/opensandbox/README.md

+# OpenSandbox Controller Helm Chart
+
+This Helm chart deploys the OpenSandbox Kubernetes Controller, which manages sandbox environments through custom resources.
+
+## Prerequisites
+
+- Kubernetes 1.19+
+- Helm 3.0+
+- Container runtime (Docker, containerd, etc.)
+- **Three container images required**:
+  1. **Controller image**: The main controller manager
+  2. **Server image**: FastAPI control plane for SDK usage
+  3. **Task Executor image**: Sidecar container for task execution (optional but required for task features)
+
+## Important: Image Requirements
+
+OpenSandbox requires **three separate images**:
+
+### 1. Controller Image
+The main controller that manages BatchSandbox and Pool resources.
+
+```bash
+# Build controller image
+make docker-build IMG=your-registry/opensandbox-controller:v1.0.0
+docker push your-registry/opensandbox-controller:v1.0.0
+```
+
+### 2. Server Image
+FastAPI control plane that exposes REST API for SDK usage. **This is the entry point for SDK clients**.
+
+```bash
+# Build server image (from server directory)
+cd ../../../server
+TAG=v1.0.0 ./build.sh
+# Or manually:
+docker build -t your-registry/opensandbox-server:v1.0.0 .
+docker push your-registry/opensandbox-server:v1.0.0
+```
+
+**Note**: The server is **required for SDK usage**. If you only use `kubectl` to manage CRDs directly, you can disable it by setting `server.enabled=false`.
+
+### 3. Task Executor Image
+A sidecar container injected into Pool pods for task execution. **This is not deployed as a separate Deployment**, but configured in Pool resources.
+
+```bash
+# Build task-executor image
+make docker-build-task-executor TASK_EXECUTOR_IMG=your-registry/opensandbox-task-executor:v1.0.0
+docker push your-registry/opensandbox-task-executor:v1.0.0
+```
+
+**Note**: The task-executor image is only needed if you want to use task execution features. For basic sandbox management without tasks, only the controller and server images are required.
+
+## Features
+
+- **SDK Control Plane**: FastAPI server for Python SDK integration
+- **Batch Sandbox Management**: Create and manage multiple identical sandbox environments
+- **Resource Pooling**: Maintain pre-warmed resource pools for rapid provisioning
+- **Task Orchestration**: Optional integrated task execution engine
+- **High Availability**: Leader election support for multiple replicas
+- **Metrics & Monitoring**: Prometheus metrics endpoint with optional ServiceMonitor
+- **Flexible Access**: ClusterIP, NodePort, or Ingress support for server access
+
+## Installation
+
+### Quick Start
+
+```bash
+# Add the chart repository (if published)
+helm repo add opensandbox https://charts.opensandbox.io
+helm repo update
+
+# Install the chart with all images
+helm install opensandbox-controller opensandbox/opensandbox-controller \
+  --set controllerManager.image.repository=your-registry/opensandbox-controller \
+  --set controllerManager.image.tag=v1.0.0 \
+  --set server.image.repository=your-registry/opensandbox-server \
+  --set server.image.tag=v1.0.0 \
+  --set taskExecutor.image.repository=your-registry/opensandbox-task-executor \
+  --set taskExecutor.image.tag=v1.0.0
+
+# Or install from local directory
+helm install opensandbox-controller ./opensandbox-controller \
+  --set controllerManager.image.repository=your-registry/opensandbox-controller \
+  --set controllerManager.image.tag=v1.0.0 \
+  --set server.image.repository=your-registry/opensandbox-server \
+  --set server.image.tag=v1.0.0 \
+  --set taskExecutor.image.repository=your-registry/opensandbox-task-executor \
+  --set taskExecutor.image.tag=v1.0.0
+```
+
+### Custom Installation
+
+```bash
+# Install with custom values
+helm install opensandbox-controller ./opensandbox-controller \
+  --set controllerManager.image.repository=your-registry/sandbox-controller \
+  --set controllerManager.image.tag=v1.0.0 \
+  --namespace opensandbox \
+  --create-namespace
+
+# Install with values file
+helm install opensandbox-controller ./opensandbox-controller \
+  -f custom-values.yaml
+```
+
+## Configuration
+
+The following table lists the configurable parameters of the chart and their default values.
+
+### Controller Manager Configuration
+
+| Parameter | Description | Default |
+|-----------|-------------|---------|
+| `controllerManager.image.repository` | Controller image repository | `opensandbox/controller` |
+| `controllerManager.image.tag` | Controller image tag | `dev` |
+| `controllerManager.image.pullPolicy` | Image pull policy | `Never` |
+| `controllerManager.replicas` | Number of controller replicas | `1` |
+| `controllerManager.resources.limits.cpu` | CPU limit | `500m` |
+| `controllerManager.resources.limits.memory` | Memory limit | `128Mi` |
+| `controllerManager.resources.requests.cpu` | CPU request | `10m` |
+| `controllerManager.resources.requests.memory` | Memory request | `64Mi` |
+| `controllerManager.leaderElect` | Enable leader election | `true` |
+| `controllerManager.logLevel` | Log verbosity level | `3` |
+
+### Task Executor Configuration
+
+**Important**: The task-executor is not deployed as a separate service. It is configured as a sidecar container in Pool resources. These settings provide the default image and resource configurations for reference when creating Pools.
+
+| Parameter | Description | Default |
+|-----------|-------------|---------|
+| `taskExecutor.image.repository` | Task Executor image repository | `opensandbox/task-executor` |
+| `taskExecutor.image.tag` | Task Executor image tag | `dev` |
+| `taskExecutor.image.pullPolicy` | Image pull policy | `Never` |
+| `taskExecutor.resources.limits.cpu` | Recommended CPU limit for sidecar | `500m` |
+| `taskExecutor.resources.limits.memory` | Recommended memory limit for sidecar | `256Mi` |
+| `taskExecutor.resources.requests.cpu` | Recommended CPU request for sidecar | `100m` |
+| `taskExecutor.resources.requests.memory` | Recommended memory request for sidecar | `128Mi` |
+
+### Server Configuration
+
+**Important**: The server is a FastAPI control plane that exposes REST API for SDK usage. It is **required for SDK integration** but can be disabled if you only use `kubectl` to manage CRDs.
+
+| Parameter | Description | Default |
+|-----------|-------------|---------|
+| `server.enabled` | Enable server deployment | `true` |
+| `server.image.repository` | Server image repository | `opensandbox/server` |
+| `server.image.tag` | Server image tag | `v0.1.0` |
+| `server.image.pullPolicy` | Image pull policy | `Never` |
+| `server.replicas` | Number of server replicas | `1` |
+| `server.resources.limits.cpu` | CPU limit | `1` |
+| `server.resources.limits.memory` | Memory limit | `512Mi` |
+| `server.resources.requests.cpu` | CPU request | `100m` |
+| `server.resources.requests.memory` | Memory request | `256Mi` |
+| `server.config.server.host` | Server listen host | `0.0.0.0` |
+| `server.config.server.port` | Server listen port | `8080` |
+| `server.config.server.logLevel` | Log level (INFO/DEBUG/WARNING/ERROR) | `INFO` |
+| `server.config.server.apiKey` | Optional API key for authentication | `""` |
+| `server.config.runtime.type` | Runtime type (kubernetes/docker) | `kubernetes` |
+| `server.config.runtime.execdImage` | execd image for non-pool mode | `opensandbox/execd:v1.0.5` |
+| `server.config.kubernetes.workloadProvider` | Workload provider type | `batchsandbox` |
+| `server.service.type` | Service type (ClusterIP/NodePort/LoadBalancer) | `ClusterIP` |
+| `server.service.port` | Service port | `8080` |
+| `server.service.nodePort` | NodePort (when type=NodePort) | `""` |
+| `server.ingress.enabled` | Enable Ingress | `false` |
+| `server.ingress.className` | Ingress class name | `""` |
+| `server.ingress.hosts` | Ingress host configuration | `[]` |
+
+### Namespace Configuration
+
+| Parameter | Description | Default |
+|-----------|-------------|---------|
+| `namespaceOverride` | Override the default namespace name | `"opensandbox"` |
+
+**Note**: Both the controller, server, and user resources (Pool, BatchSandbox) use the same namespace for simplicity.
+
+The server automatically uses in-cluster Kubernetes configuration and reads the namespace from the Helm chart configuration.
+
+### Accessing the Server
+
+#### Option 1: Port Forward (Development)
+
+```bash
+# Forward local port to server
+kubectl port-forward -n opensandbox svc/opensandbox-controller-server 8080:8080
+
+# Test connection
+curl http://localhost:8080/health
+```
+
+#### Option 2: NodePort (Local Development)
+
+```bash
+# Install with NodePort
+helm install opensandbox-controller ./opensandbox-controller \
+  --set server.service.type=NodePort \
+  --set server.service.nodePort=30080
+
+# Access via node IP
+curl http://<node-ip>:30080/health
+```
+
+#### Option 3: Ingress (Production)
+
+```bash
+# Install with Ingress
+helm install opensandbox-controller ./opensandbox-controller \
+  --set server.ingress.enabled=true \
+  --set server.ingress.className=nginx \
+  --set server.ingress.hosts[0].host=opensandbox.example.com \
+  --set server.ingress.hosts[0].paths[0].path=/ \
+  --set server.ingress.hosts[0].paths[0].pathType=Prefix
+
+# Access via domain
+curl https://opensandbox.example.com/health
+```
+
+### RBAC Configuration
+
+| Parameter | Description | Default |
+|-----------|-------------|---------|
+| `rbac.create` | Create RBAC resources | `true` |
+| `rbac.serviceAccount.create` | Create ServiceAccount | `true` |
+| `rbac.serviceAccount.name` | ServiceAccount name (if not created) | `""` |
+
+### Metrics Configuration
+
+| Parameter | Description | Default |
+|-----------|-------------|---------|
+| `metrics.enabled` | Enable metrics service | `true` |
+| `metrics.service.type` | Metrics service type | `ClusterIP` |
+| `metrics.service.port` | Metrics service port | `8443` |
+| `metrics.serviceMonitor.enabled` | Create ServiceMonitor (Prometheus Operator) | `false` |
+| `metrics.serviceMonitor.interval` | Scrape interval | `30s` |
+
+### CRD Configuration
+
+| Parameter | Description | Default |
+|-----------|-------------|---------|
+| `crds.install` | Install CRDs | `true` |
+
+### Extra Roles Configuration
+
+| Parameter | Description | Default |
+|-----------|-------------|---------|
+| `extraRoles.batchsandboxEditor.enabled` | Create BatchSandbox editor role | `true` |
+| `extraRoles.batchsandboxViewer.enabled` | Create BatchSandbox viewer role | `true` |
+| `extraRoles.poolEditor.enabled` | Create Pool editor role | `true` |
+| `extraRoles.poolViewer.enabled` | Create Pool viewer role | `true` |
+
+## Usage Examples
+
+### Example 1: Install with Custom Image
+
+```bash
+helm install opensandbox-controller ./opensandbox-controller \
+  --set controllerManager.image.repository=myregistry.com/sandbox-controller \
+  --set controllerManager.image.tag=latest
+```
+
+### Example 2: Install with High Availability
+
+```bash
+helm install opensandbox-controller ./opensandbox-controller \
+  --set controllerManager.replicas=3 \
+  --set controllerManager.resources.requests.cpu=100m \
+  --set controllerManager.resources.requests.memory=256Mi
+```
+
+### Example 3: Install with Prometheus Monitoring
+
+```bash
+helm install opensandbox-controller ./opensandbox-controller \
+  --set metrics.serviceMonitor.enabled=true
+```
+
+### Example 4: Install without CRDs (for upgrades)
+
+```bash
+helm upgrade opensandbox-controller ./opensandbox-controller \
+  --set crds.install=false
+```
+
+## Creating Resources
+
+After installation, you can create OpenSandbox resources:
+
+### Create a Pool
+
+```yaml
+apiVersion: sandbox.opensandbox.io/v1alpha1
+kind: Pool
+metadata:
+  name: example-pool
+spec:
+  minBufferSize: 2
+  maxBufferSize: 5
+  capacity: 10
+  sandboxTemplate:
+    spec:
+      image: ubuntu:latest
+      command: ["sleep", "infinity"]
+```
+
+### Create a BatchSandbox
+
+```yaml
+apiVersion: sandbox.opensandbox.io/v1alpha1
+kind: BatchSandbox
+metadata:
+  name: example-batchsandbox
+spec:
+  replicas: 3
+  ttlSecondsAfterFinished: 3600
+  sandboxTemplate:
+    spec:
+      image: ubuntu:latest
+      command: ["sleep", "infinity"]
+```
+
+## Using with SDK
+
+The OpenSandbox Python SDK connects to the server to manage sandboxes. The server must be accessible from where you run the SDK.
+
+### Access Methods
+
+#### 1. Port Forward (Recommended for Development)
+
+```bash
+# Forward local port to server
+kubectl port-forward -n opensandbox svc/opensandbox-controller-server 8080:8080
+```
+
+Then use SDK with `localhost:8080`:
+
+```python
+from opensandbox import Sandbox
+from opensandbox.config import ConnectionConfig
+
+sandbox = await Sandbox.create(
+    "ubuntu:latest",
+    entrypoint=["sleep", "infinity"],
+    connection_config=ConnectionConfig(domain="localhost:8080"),
+    extensions={"poolRef": "agent-pool"}
+)
+```
+
+#### 2. In-Cluster Access
+
+If running SDK inside the same Kubernetes cluster:
+
+```python
+sandbox = await Sandbox.create(
+    "ubuntu:latest",
+    entrypoint=["sleep", "infinity"],
+    connection_config=ConnectionConfig(
+        domain="opensandbox-controller-server.opensandbox.svc.cluster.local:8080"
+    ),
+    extensions={"poolRef": "agent-pool"}
+)
+```
+
+#### 3. NodePort / LoadBalancer / Ingress
+
+For external access, configure the service type accordingly and use the appropriate domain.
+
+### SDK Usage Examples
+
+The OpenSandbox Python SDK supports two creation modes:
+
+### Pooled Mode (Recommended)
+
+Fast creation using pre-warmed pools. **Image must match the Pool's configuration**:
+
+```python
+from opensandbox import Sandbox
+from opensandbox.config import ConnectionConfig
+
+sandbox = await Sandbox.create(
+    "ubuntu:latest",  # Must match Pool's image
+    entrypoint=["sleep", "infinity"],
+    connection_config=ConnectionConfig(domain="localhost:8080"),  # Server address
+    extensions={"poolRef": "agent-pool"}  # Reference to Pool name
+)
+```
+
+**Important**: When using `poolRef`, the SDK's `image` parameter will be **ignored** - the Pool's pre-configured image is used instead. Only `entrypoint` and `env` can be customized.
+
+### Non-pooled Mode
+
+Direct creation with custom image and resources:
+
+```python
+sandbox = await Sandbox.create(
+    "python:3.11",  # Any image
+    resource={"cpu": "1", "memory": "500Mi"},
+    connection_config=ConnectionConfig(domain="localhost:8080")
+    # No poolRef specified
+)
+```
+
+### Connect to Existing Sandbox
+
+```python
+# List all sandboxes
+from opensandbox import SandboxManager
+manager = SandboxManager(connection_config=ConnectionConfig(domain="localhost:8080"))
+sandboxes = await manager.list_sandbox_infos(SandboxFilter())
+
+# Connect to existing
+sandbox = await Sandbox.connect(
+    sandbox_id="<sandbox-id>",
+    connection_config=ConnectionConfig(domain="localhost:8080")
+)
+```
+
+For detailed SDK integration guide including troubleshooting, see [examples/README.md](examples/README.md)
+
+## Upgrading
+
+```bash
+# Upgrade to a new version
+helm upgrade opensandbox-controller ./opensandbox-controller \
+  --set controllerManager.image.tag=v1.1.0
+
+# Upgrade with new values
+helm upgrade opensandbox-controller ./opensandbox-controller \
+  -f new-values.yaml
+```
+
+## Uninstalling
+
+```bash
+# Uninstall the release
+helm uninstall opensandbox-controller
+
+# Note: CRDs are not automatically deleted. To remove them:
+kubectl delete crd batchsandboxes.sandbox.opensandbox.io
+kubectl delete crd pools.sandbox.opensandbox.io
+```
+
+## Troubleshooting
+
+### Check Controller Status
+
+```bash
+# Check deployment
+kubectl get deployment -n opensandbox
+
+# Check pods
+kubectl get pods -n opensandbox
+
+# Check logs
+kubectl logs -n opensandbox -l control-plane=controller-manager
+```
+
+### Verify CRDs
+
+```bash
+# List CRDs
+kubectl get crds | grep sandbox.opensandbox.io
+
+# Describe CRD
+kubectl describe crd batchsandboxes.sandbox.opensandbox.io
+```
+
+### Check RBAC
+
+```bash
+# Check ServiceAccount
+kubectl get sa -n opensandbox
+
+# Check ClusterRoles
+kubectl get clusterrole | grep sandbox-k8s
+
+# Check ClusterRoleBindings
+kubectl get clusterrolebinding | grep sandbox-k8s
+```
+
+## Development
+
+### Quick Start Scripts
+
+The chart includes utility scripts in the `scripts/` directory:
+
+- **`scripts/install.sh`** - Interactive installation wizard
+- **`scripts/uninstall.sh`** - Safe uninstallation with cleanup
+- **`scripts/e2e-test.sh`** - End-to-end validation
+
+See [scripts/README.md](scripts/README.md) for detailed documentation.
+
+### Linting the Chart
+
+```bash
+helm lint ./opensandbox-controller
+```
+
+### Testing the Chart
+
+```bash
+# Dry run
+helm install opensandbox-controller ./opensandbox-controller --dry-run --debug
+
+# Template rendering
+helm template opensandbox-controller ./opensandbox-controller
+```
+
+### Package the Chart
+
+```bash
+helm package ./opensandbox-controller
+```
+
+## Contributing
+
+Please refer to the main [OpenSandbox repository](https://github.com/alibaba/OpenSandbox) for contribution guidelines.
+
+## License
+
+Apache License 2.0
+
+## Support
+
+- Documentation: https://github.com/alibaba/OpenSandbox
+- Issues: https://github.com/alibaba/OpenSandbox/issues

deploy/helm/opensandbox/crds/sandbox.opensandbox.io_batchsandboxes.yaml

+---
+apiVersion: apiextensions.k8s.io/v1
+kind: CustomResourceDefinition
+metadata:
+  annotations:
+    controller-gen.kubebuilder.io/version: v0.18.0
+  name: batchsandboxes.sandbox.opensandbox.io
+spec:
+  group: sandbox.opensandbox.io
+  names:
+    kind: BatchSandbox
+    listKind: BatchSandboxList
+    plural: batchsandboxes
+    shortNames:
+    - bsbx
+    singular: batchsandbox
+  scope: Namespaced
+  versions:
+  - additionalPrinterColumns:
+    - description: The desired number of pods.
+      jsonPath: .spec.replicas
+      name: DESIRED
+      type: integer
+    - description: The number of currently all pods.
+      jsonPath: .status.replicas
+      name: TOTAL
+      type: integer
+    - description: The number of currently all allocated pods.
+      jsonPath: .status.allocated
+      name: ALLOCATED
+      type: integer
+    - description: The number of currently all ready pods.
+      jsonPath: .status.ready
+      name: Ready
+      type: integer
+    - description: The number of currently all running tasks.
+      jsonPath: .status.taskRunning
+      name: TASK_RUNNING
+      priority: 1
+      type: integer
+    - description: The number of currently all succeed tasks.
+      jsonPath: .status.taskSucceed
+      name: TASK_SUCCEED
+      priority: 1
+      type: integer
+    - description: The number of currently all failed tasks.
+      jsonPath: .status.taskFailed
+      name: TASK_FAILED
+      priority: 1
+      type: integer
+    - description: The number of currently all unknown tasks.
+      jsonPath: .status.taskUnknown
+      name: TASK_UNKNOWN
+      priority: 1
+      type: integer
+    - description: sandbox expire time
+      jsonPath: .spec.expireTime
+      name: EXPIRE
+      type: string
+    - description: CreationTimestamp is a timestamp representing the server time when
+        this object was created. It is not guaranteed to be set in happens-before
+        order across separate operations. Clients may not set this value. It is represented
+        in RFC3339 form and is in UTC.
+      jsonPath: .metadata.creationTimestamp
+      name: AGE
+      type: date
+    name: v1alpha1
+    schema:
+      openAPIV3Schema:
+        description: BatchSandbox is the Schema for the batchsandboxes API.
+        properties:
+          apiVersion:
+            description: |-
+              APIVersion defines the versioned schema of this representation of an object.
+              Servers should convert recognized schemas to the latest internal value, and
+              may reject unrecognized values.
+              More info: https://git.k8s.io/community/contributors/devel/sig-architecture/api-conventions.md#resources
+            type: string
+          kind:
+            description: |-
+              Kind is a string value representing the REST resource this object represents.
+              Servers may infer this from the endpoint the client submits requests to.
+              Cannot be updated.
+              In CamelCase.
+              More info: https://git.k8s.io/community/contributors/devel/sig-architecture/api-conventions.md#types-kinds
+            type: string
+          metadata:
+            type: object
+          spec:
+            description: BatchSandboxSpec defines the desired state of BatchSandbox.
+            properties:
+              expireTime:
+                description: |-
+                  ExpireTime - Absolute time when the batch-sandbox is deleted.
+                  If a time in the past is provided, the batch-sandbox will be deleted immediately.
+                format: date-time
+                type: string
+              poolRef:
+                description: |-
+                  PoolRef references the Pool resource name for pooled sandbox creation.
+                  Mutually exclusive with Template - use PoolRef for pool-based allocation or Template for direct sandbox creation.
+                type: string
+              replicas:
+                default: 1
+                description: Replicas is the number of desired replicas.
+                format: int32
+                minimum: 0
+                type: integer
+              shardPatches:
+                description: ShardPatches indicates patching to the Template for BatchSandbox.
+                x-kubernetes-preserve-unknown-fields: true
+              shardTaskPatches:
+                description: ShardTaskPatches indicates patching to the TaskTemplate
+                  for individual Task.
+                x-kubernetes-preserve-unknown-fields: true
+              taskResourcePolicyWhenCompleted:
+                default: Retain
+                description: |-
+                  TaskResourcePolicyWhenCompleted specifies how resources should be handled once a task reaches a completed state (SUCCEEDED or FAILED).
+                  - Retain: Keep the resources until the BatchSandbox is deleted.
+                  - Release: Free the resources immediately when the task completes.
+                type: string
+              taskTemplate:
+                description: |-
+                  Task is a custom task spec that is automatically dispatched after the sandbox is successfully created.
+                  The Sandbox is responsible for managing the lifecycle of the task.
+                x-kubernetes-preserve-unknown-fields: true
+              template:
+                description: Template describes the pods that will be created.
+                x-kubernetes-preserve-unknown-fields: true
+            required:
+            - replicas
+            type: object
+          status:
+            description: BatchSandboxStatus defines the observed state of BatchSandbox.
+            properties:
+              allocated:
+                description: "\tAllocated is the number of actual scheduled Pod"
+                format: int32
+                type: integer
+              observedGeneration:
+                description: |-
+                  ObservedGeneration is the most recent generation observed for this BatchSandbox. It corresponds to the
+                  BatchSandbox's generation, which is updated on mutation by the API Server.
+                format: int64
+                type: integer
+              ready:
+                description: "\tReady is the number of actual Ready Pod"
+                format: int32
+                type: integer
+              replicas:
+                description: Replicas is the number of actual Pods
+                format: int32
+                type: integer
+              taskFailed:
+                description: TaskFailed is the number of Failed task
+                format: int32
+                type: integer
+              taskPending:
+                description: TaskPending is the number of Pending task which is unassigned
+                format: int32
+                type: integer
+              taskRunning:
+                description: TaskRunning is the number of Running task
+                format: int32
+                type: integer
+              taskSucceed:
+                description: TaskSucceed is the number of Succeed task
+                format: int32
+                type: integer
+              taskUnknown:
+                description: TaskUnknown is the number of Unknown task
+                format: int32
+                type: integer
+            required:
+            - allocated
+            - ready
+            - replicas
+            - taskFailed
+            - taskPending
+            - taskRunning
+            - taskSucceed
+            - taskUnknown
+            type: object
+        type: object
+    served: true
+    storage: true
+    subresources:
+      status: {}

deploy/helm/opensandbox/crds/sandbox.opensandbox.io_pools.yaml

+---
+apiVersion: apiextensions.k8s.io/v1
+kind: CustomResourceDefinition
+metadata:
+  annotations:
+    controller-gen.kubebuilder.io/version: v0.18.0
+  name: pools.sandbox.opensandbox.io
+spec:
+  group: sandbox.opensandbox.io
+  names:
+    kind: Pool
+    listKind: PoolList
+    plural: pools
+    singular: pool
+  scope: Namespaced
+  versions:
+  - additionalPrinterColumns:
+    - description: The number of all nodes in pool.
+      jsonPath: .status.total
+      name: TOTAL
+      type: integer
+    - description: The number of allocated nodes in pool.
+      jsonPath: .status.allocated
+      name: ALLOCATED
+      type: integer
+    - description: The number of available nodes in pool.
+      jsonPath: .status.available
+      name: AVAILABLE
+      type: integer
+    name: v1alpha1
+    schema:
+      openAPIV3Schema:
+        description: Pool is the Schema for the pools API.
+        properties:
+          apiVersion:
+            description: |-
+              APIVersion defines the versioned schema of this representation of an object.
+              Servers should convert recognized schemas to the latest internal value, and
+              may reject unrecognized values.
+              More info: https://git.k8s.io/community/contributors/devel/sig-architecture/api-conventions.md#resources
+            type: string
+          kind:
+            description: |-
+              Kind is a string value representing the REST resource this object represents.
+              Servers may infer this from the endpoint the client submits requests to.
+              Cannot be updated.
+              In CamelCase.
+              More info: https://git.k8s.io/community/contributors/devel/sig-architecture/api-conventions.md#types-kinds
+            type: string
+          metadata:
+            type: object
+          spec:
+            description: PoolSpec defines the desired state of Pool.
+            properties:
+              capacitySpec:
+                description: CapacitySpec controls the size of the resource pool.
+                properties:
+                  bufferMax:
+                    description: BufferMax is the maximum number of nodes kept in
+                      the warm buffer.
+                    format: int32
+                    minimum: 0
+                    type: integer
+                  bufferMin:
+                    description: BufferMin is the minimum number of nodes that must
+                      remain in the buffer.
+                    format: int32
+                    minimum: 0
+                    type: integer
+                  poolMax:
+                    description: PoolMax is the maximum total number of nodes allowed
+                      in the entire pool.
+                    format: int32
+                    minimum: 0
+                    type: integer
+                  poolMin:
+                    description: PoolMin is the minimum total size of the pool.
+                    format: int32
+                    minimum: 0
+                    type: integer
+                required:
+                - bufferMax
+                - bufferMin
+                - poolMax
+                - poolMin
+                type: object
+              template:
+                description: Pod Template used to create pre-warmed nodes in the pool.
+                x-kubernetes-preserve-unknown-fields: true
+            required:
+            - capacitySpec
+            type: object
+          status:
+            description: PoolStatus defines the observed state of Pool.
+            properties:
+              allocated:
+                description: Allocated is the number of nodes currently allocated
+                  to sandboxes.
+                format: int32
+                type: integer
+              available:
+                description: Available is the number of nodes currently available
+                  in the pool.
+                format: int32
+                type: integer
+              observedGeneration:
+                description: |-
+                  ObservedGeneration is the most recent generation observed for this BatchSandbox. It corresponds to the
+                  BatchSandbox's generation, which is updated on mutation by the API Server.
+                format: int64
+                type: integer
+              revision:
+                description: Revision is the latest version of pool
+                type: string
+              total:
+                description: Total is the total number of nodes in the pool.
+                format: int32
+                type: integer
+            required:
+            - allocated
+            - available
+            - revision
+            - total
+            type: object
+        type: object
+    served: true
+    storage: true
+    subresources:
+      status: {}

deploy/helm/opensandbox/examples/DIRECTORY_STRUCTURE.md

+# Examples 目录结构
+
+本文档说明 `examples/` 目录的组织结构和设计原则。
+
+## 📂 目录结构
+
+```
+examples/
+├── README.md                          # 主文档：快速开始、最佳实践
+├── DIRECTORY_STRUCTURE.md             # 本文档：目录结构说明
+├── pool-examples.md                   # 详细的 Pool 配置说明
+│
+├── pool-agent-production.yaml         # 🌟 生产级 Agent Pool（推荐）
+├── pool-sdk-compatible.yaml           # SDK 基础 Pool（execd only）
+├── pool-sdk-with-tasks.yaml           # SDK 完整 Pool（execd + task-executor）
+│
+├── batchsandbox-basic.yaml            # Non-pooled 模式示例
+└── batchsandbox-with-tasks.yaml       # Pooled 批量任务示例
+```
+
+## 📝 文件分类
+
+### Pool 配置文件（3 个）
+
+| 文件 | 类型 | SDK 支持 | 自定义 entrypoint | 推荐度 |
+|------|------|----------|-----------------|--------|
+| `pool-agent-production.yaml` | 生产级 | ✅ | ✅ | ⭐⭐⭐⭐⭐ |
+| `pool-sdk-with-tasks.yaml` | 完整功能 | ✅ | ✅ | ⭐⭐⭐⭐ |
+| `pool-sdk-compatible.yaml` | 基础功能 | ✅ | ❌ | ⭐⭐⭐ |
+
+**选择建议**：
+- **生产环境**：使用 `pool-agent-production.yaml`（包含详细注释和最佳实践）
+- **快速测试**：使用 `pool-sdk-with-tasks.yaml`（更简洁）
+- **特殊需求**：如果不需要自定义 entrypoint，使用 `pool-sdk-compatible.yaml`
+
+### BatchSandbox 配置文件（2 个）
+
+| 文件 | 模式 | 依赖 | 用途 |
+|------|------|------|------|
+| `batchsandbox-basic.yaml` | Non-pooled | 无 | 演示直接创建 Pod |
+| `batchsandbox-with-tasks.yaml` | Pooled | Pool | 演示批量异构任务 |
+
+**使用说明**：
+- **SDK 场景**：通常不需要手动创建 BatchSandbox（SDK 自动创建）
+- **kubectl 场景**：用于批量任务执行（RL 训练、压力测试等）
+
+### 文档文件（3 个）
+
+| 文件 | 内容 | 面向用户 |
+|------|------|---------|
+| `README.md` | 快速开始、最佳实践、故障排查 | 所有用户 |
+| `pool-examples.md` | Pool 详细配置说明 | 高级用户 |
+| `DIRECTORY_STRUCTURE.md` | 目录结构说明 | 开发者 |
+
+## 🎯 设计原则
+
+### 1. 简化选择
+
+**问题**：之前有太多相似的示例（pool-basic, pool-with-execd, pool-with-task-executor...），用户不知道选哪个。
+
+**解决**：
+- 保留 3 个 Pool 示例，明确分类和推荐度
+- 删除误导性示例（pool-basic 无 execd，pool-with-task-executor 无 execd）
+- 突出推荐 `pool-agent-production.yaml`
+
+### 2. 面向实际场景
+
+**问题**：示例文件缺乏实际使用场景说明。
+
+**解决**：
+- 明确标注适用场景（Agent 服务、RL 训练、压力测试等）
+- 提供完整的 SDK 使用代码示例
+- 包含部署、验证、监控的完整流程
+
+### 3. 最佳实践优先
+
+**问题**：缺少生产级配置参考。
+
+**解决**：
+- 创建 `pool-agent-production.yaml` 包含：
+  - 详细的配置注释
+  - 容量规划建议
+  - 安全最佳实践
+  - 监控和调试指南
+
+### 4. 纠正常见误区
+
+**问题**：用户容易误解 Pool 的使用方式。
+
+**解决**：
+- 在 README.md 中突出"常见误区"章节
+- 明确说明：
+  - Pool 是 Pod 池，不是 Sandbox 池
+  - 不需要预创建 BatchSandbox
+  - SDK 每次 create() 创建新 BatchSandbox
+
+## 🔄 文件变更历史
+
+### 删除的文件（4 个）
+
+| 文件 | 删除原因 |
+|------|---------|
+| `pool-basic.yaml` | 无 execd，SDK 无法使用，误导性 |
+| `pool-with-task-executor.yaml` | 只有 task-executor 无 execd，不完整 |
+| `pool-with-execd.yaml` | execd 启动方式不标准，已被 pool-sdk-compatible.yaml 替代 |
+| `batchsandbox-pooled.yaml` | 依赖不存在的 basic-pool，无效 |
+
+### 新增的文件（1 个）
+
+| 文件 | 内容 |
+|------|------|
+| `pool-agent-production.yaml` | 生产级 Agent Pool 配置，包含详细注释和最佳实践 |
+
+### 修改的文件（3 个）
+
+| 文件 | 修改内容 |
+|------|---------|
+| `README.md` | 重写，添加场景分类、常见误区、容量规划等 |
+| `batchsandbox-with-tasks.yaml` | 修改 poolRef 为 agent-pool，添加注释 |
+| `batchsandbox-basic.yaml` | 添加详细注释说明 |
+
+## 📊 使用场景映射
+
+### 场景 A：多 Agent 并发使用
+
+```
+用户需求: Agent 服务、Code Interpreter、动态工作流
+      ↓
+推荐配置: pool-agent-production.yaml
+      ↓
+  使用方式: SDK 动态创建 sandbox
+      ↓
+    流程: Helm 部署 Pool → SDK.create() → SDK.kill()
+```
+
+### 场景 B：批量任务执行
+
+```
+用户需求: RL 训练、压力测试、批量数据处理
+      ↓
+推荐配置: pool-agent-production.yaml + batchsandbox-with-tasks.yaml
+      ↓
+  使用方式: kubectl 创建 BatchSandbox
+      ↓
+    流程: kubectl apply pool → kubectl apply batchsandbox → 自动清理
+```
+
+### 场景 C：测试和开发
+
+```
+用户需求: 测试特定镜像、验证功能
+      ↓
+推荐配置: batchsandbox-basic.yaml (non-pooled)
+      ↓
+  使用方式: kubectl 直接创建
+      ↓
+    流程: kubectl apply → kubectl delete
+```
+
+## 🔗 相关资源
+
+- **深度分析**：`/data/home/cz/sandbox-test/pool-analysis/`
+  - Pool 使用指南
+  - 架构流程图
+  - 验证测试脚本
+
+- **Helm 配置**：`/data/home/cz/OpenSandbox/kubernetes/helm-chart/`
+  - `values.yaml` - 默认配置
+  - `values-e2e.yaml` - E2E 测试配置
+  - 生产/开发环境配置：使用 `--set` 或自定义 values 文件
+
+- **主文档**：`/data/home/cz/OpenSandbox/kubernetes/README.md`
+  - Kubernetes 部署完整指南
+
+## 💡 维护建议
+
+### 添加新示例时
+
+1. **明确场景**：每个示例应对应明确的使用场景
+2. **完整注释**：包含配置说明、使用方式、注意事项
+3. **验证测试**：确保示例可以正常运行
+4. **更新文档**：同步更新 README.md 和本文档
+
+### 修改现有示例时
+
+1. **保持兼容**：避免破坏性变更
+2. **版本说明**：在注释中说明版本要求
+3. **测试验证**：修改后进行完整测试
+4. **文档同步**：更新相关文档
+
+### 删除示例时
+
+1. **评估影响**：确认没有外部依赖
+2. **提供替代**：在文档中说明替代方案
+3. **迁移指南**：如果有用户使用，提供迁移步骤

deploy/helm/opensandbox/examples/README.md

+# OpenSandbox Examples
+
+This directory contains various usage examples and best practices for the OpenSandbox Kubernetes Controller.
+
+## 🎉 Important Update
+
+**Starting from version v0.2.0, the Helm chart deploys an agent-pool by default** without manual creation!
+
+```bash
+# Default installation automatically creates agent-pool
+helm install opensandbox opensandbox-controller
+
+# View the automatically created Pool
+kubectl get pools -n opensandbox
+```
+
+If you don't need the Pool, you can disable it:
+```bash
+helm install opensandbox opensandbox-controller --set pools[0].enabled=false
+```
+
+## 📁 File List
+
+### Pool Examples
+
+| File | SDK Compatible | Custom Entrypoint | Purpose |
+|------|----------------|-------------------|---------|
+| `pool-sdk-compatible.yaml` | ✅ Supported | ❌ Not Supported | SDK Basic Mode (execd only)|
+| `pool-sdk-with-tasks.yaml` | ✅ Supported | ✅ Supported | SDK Complete Mode (execd + task-executor)|
+| **`pool-agent-production.yaml`** | ✅ Supported | ✅ Supported | **🌟 Production-Grade Agent Pool (Recommended)** |
+
+### BatchSandbox Examples
+
+| File | Mode | Purpose |
+|------|------|---------|
+| `batchsandbox-basic.yaml` | Non-pooled | Direct Pod creation without using Pool |
+| `batchsandbox-with-tasks.yaml` | Pooled | Batch heterogeneous task example |
+
+### Documentation
+
+| File | Content |
+|------|---------|
+| `README.md` | This document |
+| `pool-examples.md` | Detailed Pool configuration guide |
+
+## 🎯 Core Concepts
+
+### ❌ Common Misconceptions
+
+> **Misconception 1**: Pool is a pre-created pool of Sandboxes that Agents can reuse
+> **Correct**: Pool is a **Pod pool**, not a Sandbox pool
+
+> **Misconception 2**: Need to pre-create BatchSandbox during Helm deployment for Agent use
+> **Correct**: SDK creates a **new BatchSandbox** with each create() call, no pre-creation needed
+
+> **Misconception 3**: Pool without execd can work with SDK
+> **Correct**: SDK **requires** Pool to contain execd (port 44772)
+
+### ✅ Correct Understanding
+
+```
+During Helm deployment:
+└─> Only create Pool (long-running, maintains pre-warmed Pods)
+
+During SDK runtime:
+├─> Agent-1: SDK.create() → Creates BatchSandbox-1 (allocates Pod-1)
+├─> Agent-2: SDK.create() → Creates BatchSandbox-2 (allocates Pod-2)
+└─> Agent-1: SDK.kill()  → Deletes BatchSandbox-1 (Pod-1 returns to Pool)
+
+Next request:
+└─> Agent-3: SDK.create() → Creates BatchSandbox-3 (reuses Pod-1) ← Fast!
+```
+
+**Key Points**:
+- ✅ Pool maintains **Pods** (pre-warmed containers)
+- ✅ Each SDK.create() creates a **new BatchSandbox**
+- ✅ Pods are reused, BatchSandboxes are not
+- ❌ Don't pre-create BatchSandboxes
+
+## 🚀 Quick Start
+
+### Scenario A: Multi-Agent Concurrent Usage (Recommended)
+
+**Use Cases**: Agent services, Code Interpreter, dynamic workflows
+
+```bash
+# 1. Install Helm chart (automatically creates agent-pool)
+helm install opensandbox opensandbox-controller
+
+# 2. Verify deployment
+kubectl get deployment -n opensandbox
+kubectl get pool -n opensandbox
+
+# 3. Check Pool status
+kubectl get pool agent-pool -n opensandbox -o jsonpath='{.status}' | jq
+# Example output:
+# {
+#   "total": 10,      # Total Pods
+#   "allocated": 0,   # Allocated
+#   "available": 10   # Available
+# }
+
+# 4. View Pool Pods
+kubectl get pods -l pool=agent-pool -n opensandbox
+```
+
+**SDK Usage**:
+
+```python
+from opensandbox import Sandbox
+from opensandbox.config import ConnectionConfig
+from datetime import timedelta
+
+async def handle_agent_request(agent_id: str, task: str):
+    """Create a new sandbox for each Agent request"""
+    # SDK.create() will allocate a Pod from agent-pool
+    sandbox = await Sandbox.create(
+        "nginx:latest",  # Will be ignored, uses image from Pool
+        entrypoint=["/bin/sh", "-c", "sleep infinity"],
+        env={"AGENT_ID": agent_id},
+        timeout=timedelta(hours=1),
+        connection_config=ConnectionConfig(domain="<server-ip>:8088"),
+        extensions={"poolRef": "agent-pool"}  # Use default agent-pool
+    )
+
+    try:
+        # Use sandbox
+        result = await sandbox.commands.run(task)
+        return result
+    finally:
+        # Delete BatchSandbox, Pod returns to Pool
+        await sandbox.kill()
+```
+
+### Scenario B: Custom Pool Capacity (High Concurrency Scenarios)
+
+If you need higher concurrency capacity, override default capacity parameters using `--set`:
+
+```bash
+# Use configuration optimized for multiple Agents (bufferMin: 50, poolMax: 300)
+helm install opensandbox opensandbox-controller \
+  --set pools[0].capacitySpec.bufferMin=50 \
+  --set pools[0].capacitySpec.bufferMax=100 \
+  --set pools[0].capacitySpec.poolMin=50 \
+  --set pools[0].capacitySpec.poolMax=300
+```
+
+### Scenario C: Batch Task Execution (kubectl)
+
+**Use Cases**: RL training, stress testing, batch data processing
+
+```bash
+# 1. Ensure Pool is deployed (automatically created by default)
+kubectl get pool agent-pool -n opensandbox
+
+# 2. Create BatchSandbox to execute batch tasks
+kubectl apply -f batchsandbox-with-tasks.yaml
+
+# 3. View task execution status
+kubectl get batchsandbox task-batch-sandbox -n opensandbox -o wide
+
+# 4. View task logs
+POD_NAME=$(kubectl get pods -l batchsandbox=task-batch-sandbox -n opensandbox -o jsonpath='{.items[0].metadata.name}')
+kubectl logs $POD_NAME -c sandbox-container -n opensandbox
+kubectl logs $POD_NAME -c task-executor -n opensandbox
+
+# 5. Automatic cleanup after task completion (ttlSecondsAfterFinished)
+```
+
+### Scenario D: Non-pooled Direct Creation
+
+**Use Cases**: Testing environments, special image requirements
+
+```bash
+# Directly create BatchSandbox (without using Pool)
+kubectl apply -f batchsandbox-basic.yaml
+
+# View created Pods
+kubectl get pods -l batchsandbox=basic-batch-sandbox -n opensandbox
+```
+
+## 📊 Pool Configuration Type Comparison
+
+### Type 1: SDK Pool (Basic) - pool-sdk-compatible.yaml
+
+```yaml
+# ✅ SDK compatible - contains execd only
+# ❌ Does not support custom entrypoint
+initContainers:
+- name: execd-installer
+  image: opensandbox/execd:v1.0.5
+containers:
+- name: sandbox-container
+  command: ["/opt/opensandbox/bin/bootstrap.sh", "nginx", "-g", "daemon off;"]
+  ports:
+  - containerPort: 44772
+    name: execd
+```
+
+**SDK Usage**:
+```python
+sandbox = await Sandbox.create(
+    "nginx:latest",
+    # ❌ Cannot pass entrypoint
+    env={"VAR": "value"},  # ✅ Can pass environment variables
+    extensions={"poolRef": "sdk-pool"}
+)
+```
+
+### Type 2: Task Pool (Complete) - pool-agent-production.yaml (Recommended)
+
+```yaml
+# ✅ SDK compatible - contains execd + task-executor
+# ✅ Supports custom entrypoint
+spec:
+  shareProcessNamespace: true  # Required by task-executor
+  initContainers:
+  - name: execd-installer
+    image: opensandbox/execd:v1.0.5
+  containers:
+  - name: sandbox-container
+    command: ["/opt/opensandbox/bin/bootstrap.sh", "sleep", "infinity"]
+    ports:
+    - containerPort: 44772
+      name: execd
+  - name: task-executor  # Add task-executor sidecar
+    image: opensandbox/task-executor:dev
+    securityContext:
+      capabilities:
+        add: ["SYS_PTRACE"]
+```
+
+**SDK Usage** (with custom entrypoint):
+```python
+sandbox = await Sandbox.create(
+    "nginx:latest",
+    entrypoint=["/bin/sh", "-c", "custom command"],  # ✅ Can customize
+    env={"VAR": "value"},
+    extensions={"poolRef": "agent-pool"}
+)
+```
+
+## 🔍 Monitoring and Debugging
+
+### Monitor Pool Utilization
+
+```bash
+# Real-time monitoring
+watch kubectl get pool agent-pool -o jsonpath='{.status}' | jq
+
+# View detailed information
+kubectl describe pool agent-pool
+
+# View Pool Pod list
+kubectl get pods -l pool=agent-pool -o wide
+```
+
+**Optimization Recommendations**:
+- If `available` is frequently 0 → Increase `bufferMax`
+- If `available` is always close to `total` → Decrease `bufferMin`
+- If `total` frequently reaches `poolMax` → Increase `poolMax` or optimize Agent usage
+
+### Verify Pool Configuration
+
+```bash
+# Check if Pod contains execd
+kubectl exec -it <pod-name> -c sandbox-container -- ps aux | grep execd
+
+# Check execd port
+kubectl exec -it <pod-name> -c sandbox-container -- nc -zv localhost 44772
+
+# Check task-executor (if present)
+kubectl get pods -l pool=agent-pool -o jsonpath='{.items[0].spec.containers[*].name}'
+# Output should include: sandbox-container task-executor
+```
+
+### View BatchSandbox Status
+
+```bash
+# List all BatchSandboxes
+kubectl get batchsandboxes
+
+# View detailed status
+kubectl describe batchsandbox <name>
+
+# View task execution statistics
+kubectl get batchsandbox -o custom-columns=\
+NAME:.metadata.name,\
+REPLICAS:.spec.replicas,\
+RUNNING:.status.taskRunning,\
+SUCCEED:.status.taskSucceed,\
+FAILED:.status.taskFailed
+```
+
+## 🛠️ Troubleshooting
+
+### Pool Pod Fails to Start
+
+```bash
+# View Pod events
+kubectl describe pod <pod-name>
+
+# View container logs
+kubectl logs <pod-name> -c sandbox-container
+kubectl logs <pod-name> -c task-executor  # If present
+
+# Check image pull
+kubectl describe pod <pod-name> | grep -A 5 Events
+```
+
+### SDK Sandbox Creation Timeout
+
+**Symptom**: SDK error `Health check timeout`
+
+**Possible Causes**:
+1. Pool doesn't have execd → Use `pool-agent-production.yaml`
+2. execd not started → Check execd process in Pod
+3. Network issues → Check network connectivity between Server and Pod
+
+**Troubleshooting Steps**:
+```bash
+# 1. Confirm Pool contains execd
+kubectl get pool agent-pool -o yaml | grep -A 10 initContainers
+
+# 2. Check execd process
+kubectl exec -it <pod-name> -c sandbox-container -- ps aux | grep execd
+
+# 3. Check execd port
+kubectl exec -it <pod-name> -c sandbox-container -- nc -zv localhost 44772
+
+# 4. View Server logs
+kubectl logs -l app=opensandbox-server -n opensandbox
+```
+
+### task-executor Permission Issues
+
+```bash
+# Check security context
+kubectl get pod <pod-name> -o yaml | grep -A 10 securityContext
+
+# Should contain:
+# capabilities:
+#   add: ["SYS_PTRACE"]
+
+# Check process namespace sharing
+kubectl get pod <pod-name> -o jsonpath='{.spec.shareProcessNamespace}'
+# Should output: true
+```
+
+## 📦 Capacity Planning Recommendations
+
+Plan Pool capacity based on concurrent Agent count:
+
+| Concurrent Agents | bufferMin | bufferMax | poolMin | poolMax | Description |
+|------------------|-----------|-----------|---------|---------|-------------|
+| 1-10             | 2         | 5         | 2       | 20      | Small-scale testing |
+| 10-50            | 10        | 20        | 10      | 100     | Small to medium applications |
+| 50-200           | 50        | 100       | 50      | 300     | Medium to large applications |
+| 200+             | 100       | 200       | 100     | 500     | Large-scale production |
+
+**Parameter Descriptions**:
+- `bufferMin`: Minimum buffer, ensures fast response
+- `bufferMax`: Maximum buffer, controls pre-warming cost
+- `poolMin`: Minimum capacity during low traffic periods
+- `poolMax`: Maximum capacity during peak periods
+
+**Cost Optimization**:
+- Low traffic periods: Pool scales down to `poolMin`, saving resources
+- Peak periods: Pool expands to `poolMax`, ensuring response speed
+- Buffer zone: `bufferMin` ensures fast response, `bufferMax` avoids excessive pre-warming
+
+## 📚 Related Documentation
+
+### In-depth Analysis Documents
+
+- **Pool Usage Guide**: `/data/home/cz/sandbox-test/pool-analysis/opensandbox_pool_usage_guide.md`
+- **Architecture Diagrams**: `/data/home/cz/sandbox-test/pool-analysis/pool_architecture.txt`
+- **Verification Test Scripts**: `/data/home/cz/sandbox-test/pool-analysis/test_pool_behavior.py`
+
+### Helm Deployment Configuration
+
+- **Main values configuration**: `/data/home/cz/OpenSandbox/kubernetes/helm-chart/values.yaml`
+- **E2E test configuration**: `/data/home/cz/OpenSandbox/kubernetes/helm-chart/values-e2e.yaml`
+- **Main README**: `/data/home/cz/OpenSandbox/kubernetes/README.md`
+
+### API Reference
+
+```bash
+# View Pool CRD definition
+kubectl explain pool
+kubectl explain pool.spec
+kubectl explain pool.spec.capacitySpec
+
+# View BatchSandbox CRD definition
+kubectl explain batchsandbox
+kubectl explain batchsandbox.spec
+kubectl explain batchsandbox.spec.taskTemplate
+```
+
+## 🧹 Resource Cleanup
+
+```bash
+# Delete BatchSandbox
+kubectl delete batchsandbox --all -n default
+
+# Delete Pool (automatically cleans up related Pods)
+kubectl delete pool --all -n default
+
+# Delete all resources in namespace
+kubectl delete all --all -n default
+```
+
+## 💡 Best Practices Summary
+
+1. **Pool is a Pod pool, not a Sandbox pool**
+2. **SDK creates a new BatchSandbox with each create() call**
+3. **No need to pre-create BatchSandboxes for reuse**
+4. **Only create Pool during Helm deployment**
+5. **Dynamically create/delete BatchSandboxes at runtime**
+6. **Configure Pool capacity parameters appropriately to optimize cost and performance**
+7. **Use `pool-agent-production.yaml` as production environment template**
+
+## 🔗 More Examples
+
+- **Detailed Pool configuration guide**: See `pool-examples.md`
+- **SDK integration examples**: See "Quick Start" section in this document
+- **Helm Chart configuration**: See `values.yaml` and `values-e2e.yaml`

deploy/helm/opensandbox/examples/batchsandbox-basic.yaml

+# ==============================================================================
+# BatchSandbox - Non-pooled 模式（直接创建 Pod）
+# ==============================================================================
+#
+# 用途：演示不使用 Pool 直接创建 BatchSandbox
+#
+# 使用场景：
+#   - 需要使用特定镜像或资源配置（Pool 无法满足）
+#   - 一次性任务，不需要预热优化
+#   - 测试和开发环境
+#
+# 特点：
+#   - 可自由指定镜像、资源、配置
+#   - 创建速度较慢（需要拉取镜像、启动 Pod）
+#   - 不复用预热资源
+#
+# 注意：
+#   - SDK 使用场景通常不需要预创建 BatchSandbox（动态创建即可）
+#   - 此示例主要用于演示 kubectl 直接创建
+#
+# ==============================================================================
+
+apiVersion: sandbox.opensandbox.io/v1alpha1
+kind: BatchSandbox
+metadata:
+  name: basic-batch-sandbox
+  namespace: default
+  labels:
+    mode: non-pooled
+  annotations:
+    description: "Non-pooled 模式示例"
+spec:
+  # 创建3个沙箱副本
+  replicas: 3
+
+  # 不使用资源池，直接创建Pod
+  # poolRef: ""  # 留空表示 non-pooled 模式
+
+  # TTL：3600秒（1小时）后自动清理
+  ttlSecondsAfterFinished: 3600
+
+  # 沙箱模板
+  template:
+    spec:
+      containers:
+      - name: sandbox-container
+        image: ubuntu:22.04
+        command: ["sleep", "infinity"]
+        resources:
+          requests:
+            cpu: "100m"
+            memory: "128Mi"
+          limits:
+            cpu: "500m"
+            memory: "256Mi"

deploy/helm/opensandbox/examples/batchsandbox-with-tasks.yaml

+# ==============================================================================
+# BatchSandbox with Heterogeneous Tasks（带异构任务的批量沙箱）
+# ==============================================================================
+#
+# 用途：演示如何使用 Pool 批量执行异构任务（每个 sandbox 执行不同任务）
+#
+# 使用场景：
+#   - RL 训练：批量创建训练环境，每个环境执行不同策略
+#   - 压力测试：批量执行不同的测试用例
+#   - 数据处理：并行处理多个数据分片
+#
+# 注意：
+#   - 此示例主要用于 kubectl 直接创建的批量任务场景
+#   - SDK 使用场景通常不需要预创建 BatchSandbox（动态创建即可）
+#   - 需要先创建 pool-agent-production.yaml
+#
+# 相关文档：
+#   - /data/home/cz/sandbox-test/pool-analysis/opensandbox_pool_usage_guide.md
+#
+# ==============================================================================
+
+apiVersion: sandbox.opensandbox.io/v1alpha1
+kind: BatchSandbox
+metadata:
+  name: task-batch-sandbox
+  namespace: default
+  labels:
+    use-case: batch-training
+  annotations:
+    description: "批量异构任务示例"
+spec:
+  # 创建3个沙箱副本，每个执行不同的任务
+  replicas: 3
+
+  # 使用包含 execd + task-executor 的 Pool
+  # 注意：需要先创建 pool-agent-production.yaml
+  poolRef: agent-pool
+
+  # TTL：3600秒（1小时）后自动清理
+  ttlSecondsAfterFinished: 3600
+
+  # 默认任务模板（如果shardTaskPatches没有覆盖，则使用此模板）
+  taskTemplate:
+    spec:
+      process:
+        command: ["echo", "Default task message"]
+
+  # 异构任务：为每个沙箱自定义不同的任务
+  shardTaskPatches:
+  # 第1个沙箱的任务
+  - spec:
+      process:
+        command: ["bash", "-c"]
+        args:
+        - |
+          echo "Task 1: Running Python script"
+          python3 -c "
+          import time
+          print('Task 1 started')
+          time.sleep(2)
+          print('Task 1 completed')
+          "
+
+  # 第2个沙箱的任务
+  - spec:
+      process:
+        command: ["bash", "-c"]
+        args:
+        - |
+          echo "Task 2: Running shell commands"
+          date
+          uname -a
+          sleep 1
+          echo "Task 2 completed"
+
+  # 第3个沙箱的任务
+  - spec:
+      process:
+        command: ["bash", "-c"]
+        args:
+        - |
+          echo "Task 3: System info check"
+          cat /etc/os-release
+          df -h
+          free -h
+          echo "Task 3 completed"

deploy/helm/opensandbox/examples/pool-agent-production.yaml

+# ==============================================================================
+# OpenSandbox Agent Pool - 生产级配置
+# ==============================================================================
+#
+# 用途：为多 Agent 并发场景提供预热的沙箱资源池（SDK 模式）
+#
+# 注意：此配置适用于 SDK 场景（需要 execd）
+# 如果只使用 kubectl 管理 BatchSandbox，可以移除 execd 相关配置
+#
+# 架构说明：
+#   Pool 维护预热的 Pod 池（不是 Sandbox 池）
+#   - 每个 Pod 包含 execd（SDK 通信）+ task-executor（任务注入）
+#   - SDK.create() 从 Pool 分配 Pod，创建新的 BatchSandbox
+#   - SDK.kill() 删除 BatchSandbox，Pod 返回 Pool
+#
+# 使用模式：
+#   1. Helm 部署时创建此 Pool（长期运行）
+#   2. Agent 运行时通过 SDK 动态创建/删除 BatchSandbox
+#   3. 不需要预创建 BatchSandbox
+#
+# 容量规划建议：
+#   | 并发 Agent 数 | bufferMin | bufferMax | poolMin | poolMax |
+#   |--------------|-----------|-----------|---------|---------|
+#   | 1-10         | 2         | 5         | 2       | 20      |
+#   | 10-50        | 10        | 20        | 10      | 100     |
+#   | 50-200       | 50        | 100       | 50      | 300     |
+#   | 200+         | 100       | 200       | 100     | 500     |
+#
+# 相关文档：
+#   - /data/home/cz/sandbox-test/pool-analysis/opensandbox_pool_usage_guide.md
+#   - /data/home/cz/OpenSandbox/kubernetes/helm-chart/values.yaml (查看 pools 配置)
+#
+# ==============================================================================
+
+apiVersion: sandbox.opensandbox.io/v1alpha1
+kind: Pool
+metadata:
+  name: agent-pool
+  namespace: default  # 使用与 Helm chart namespaceOverride 一致的命名空间
+  labels:
+    app: opensandbox
+    component: agent-pool
+  annotations:
+    description: "生产级 Agent Pool，支持 SDK 动态创建 sandbox"
+spec:
+  template:
+    metadata:
+      labels:
+        pool: agent-pool
+        sdk-compatible: "true"
+    spec:
+      # ========================================
+      # 必需：共享进程命名空间
+      # ========================================
+      # task-executor 需要访问 sandbox 容器的进程树
+      shareProcessNamespace: true
+
+      # ========================================
+      # Init Container：安装 execd（仅 SDK 场景需要）
+      # ========================================
+      # 注意：如果只使用 kubectl 管理 BatchSandbox（不使用 SDK），
+      # 可以移除此 init container 和相关的 execd 配置。
+      #
+      # execd 的作用：
+      #   - 提供 SDK 与 Pod 的通信接口（44772 端口）
+      #   - 执行 SDK 发送的命令（commands.run(), files.read_file() 等）
+      #
+      # 使用场景：
+      #   ✅ 需要：SDK 动态创建 sandbox（Agent、Code Interpreter 等）
+      #   ❌ 不需要：纯 kubectl 批量任务（RL 训练、压力测试等）
+      initContainers:
+      - name: execd-installer
+        image: opensandbox/execd:v1.0.5
+        imagePullPolicy: IfNotPresent
+        command: ["/bin/sh", "-c"]
+        args:
+          - |
+            # 复制 execd 二进制和启动脚本
+            cp ./execd /opt/opensandbox/bin/execd && \
+            cp ./bootstrap.sh /opt/opensandbox/bin/bootstrap.sh && \
+            chmod +x /opt/opensandbox/bin/execd && \
+            chmod +x /opt/opensandbox/bin/bootstrap.sh && \
+            echo "execd installed successfully"
+        volumeMounts:
+        - name: opensandbox-bin
+          mountPath: /opt/opensandbox/bin
+
+      # ========================================
+      # 主容器：Sandbox 环境
+      # ========================================
+      containers:
+      - name: sandbox-container
+        # 镜像说明：
+        # - Pool 中的镜像由 Pool 预定义，SDK 指定的镜像会被忽略
+        # - 根据 Agent 需求选择合适的基础镜像（nginx, ubuntu, python, etc.）
+        # - 确保镜像包含 /bin/sh 用于执行 bootstrap.sh
+        image: nginx:latest
+        imagePullPolicy: IfNotPresent
+
+        # 启动命令：使用 bootstrap.sh 启动 execd
+        # bootstrap.sh 会：
+        #   1. 后台启动 execd（监听 44772 端口）
+        #   2. 执行用户指定的命令（这里是 sleep infinity）
+        command: ["/opt/opensandbox/bin/bootstrap.sh", "sleep", "infinity"]
+
+        # 环境变量
+        env:
+        - name: EXECD
+          value: /opt/opensandbox/bin/execd
+        # 可添加其他环境变量：
+        # - name: CUSTOM_VAR
+        #   value: "custom-value"
+
+        # 端口配置
+        ports:
+        - containerPort: 44772
+          name: execd
+          protocol: TCP
+        # 可暴露其他端口（如应用端口）：
+        # - containerPort: 8080
+        #   name: app
+        #   protocol: TCP
+
+        # 资源配置
+        # 根据 Agent 任务复杂度调整
+        resources:
+          requests:
+            cpu: "100m"      # 最小 CPU（保证调度）
+            memory: "128Mi"  # 最小内存
+          limits:
+            cpu: "500m"      # 最大 CPU（防止资源抢占）
+            memory: "256Mi"  # 最大内存
+
+        # 卷挂载
+        volumeMounts:
+        - name: opensandbox-bin
+          mountPath: /opt/opensandbox/bin
+          readOnly: true
+
+        # 健康检查（可选）
+        # livenessProbe:
+        #   tcpSocket:
+        #     port: execd
+        #   initialDelaySeconds: 10
+        #   periodSeconds: 30
+        # readinessProbe:
+        #   tcpSocket:
+        #     port: execd
+        #   initialDelaySeconds: 5
+        #   periodSeconds: 10
+
+      # ========================================
+      # Sidecar：Task Executor
+      # ========================================
+      - name: task-executor
+        # 镜像说明：
+        # - 使用 Helm chart 中配置的 task-executor 镜像
+        # - 版本应与 controller 版本保持一致
+        # - 开发环境可用：opensandbox/task-executor:dev
+        # - 生产环境建议：your-registry/opensandbox-task-executor:v1.0.0
+        image: opensandbox/task-executor:dev
+        imagePullPolicy: Never  # 开发环境使用本地镜像；生产环境改为 IfNotPresent
+
+        # 端口配置
+        ports:
+        - containerPort: 5758
+          name: task-executor
+          protocol: TCP
+
+        # 资源配置
+        # task-executor 需要更多 CPU 用于进程注入
+        resources:
+          requests:
+            cpu: "100m"
+            memory: "128Mi"
+          limits:
+            cpu: "500m"      # 可能需要更多 CPU
+            memory: "256Mi"
+
+        # 安全上下文
+        # 必需：SYS_PTRACE 权限用于注入进程到 sandbox 容器
+        securityContext:
+          capabilities:
+            add: ["SYS_PTRACE"]
+          # 生产环境建议添加其他安全设置：
+          # runAsNonRoot: true
+          # runAsUser: 1000
+          # allowPrivilegeEscalation: false
+
+      # ========================================
+      # 卷配置
+      # ========================================
+      volumes:
+      - name: opensandbox-bin
+        emptyDir: {}
+      # 可添加其他卷（如配置文件、数据持久化）：
+      # - name: config
+      #   configMap:
+      #     name: agent-config
+      # - name: data
+      #   persistentVolumeClaim:
+      #     claimName: agent-data-pvc
+
+      # ========================================
+      # 调度配置（可选）
+      # ========================================
+      # 节点选择器
+      # nodeSelector:
+      #   workload-type: agent
+      #   zone: production
+
+      # 容忍度（允许调度到特定污点的节点）
+      # tolerations:
+      # - key: "workload"
+      #   operator: "Equal"
+      #   value: "agent"
+      #   effect: "NoSchedule"
+
+      # 亲和性（控制 Pod 分布）
+      # affinity:
+      #   podAntiAffinity:
+      #     preferredDuringSchedulingIgnoredDuringExecution:
+      #     - weight: 100
+      #       podAffinityTerm:
+      #         labelSelector:
+      #           matchExpressions:
+      #           - key: pool
+      #             operator: In
+      #             values:
+      #             - agent-pool
+      #         topologyKey: kubernetes.io/hostname
+
+  # ========================================
+  # Pool 容量配置
+  # ========================================
+  capacitySpec:
+    # bufferMin: 最小缓冲 - Pool 保证至少有这么多可用 Pod
+    # - 保证快速响应，避免 Agent 等待
+    # - 根据并发 Agent 峰值设置
+    bufferMin: 10
+
+    # bufferMax: 最大缓冲 - Pool 最多预热这么多 Pod
+    # - 控制预热成本，避免资源浪费
+    # - 通常设为 bufferMin 的 2-5 倍
+    bufferMax: 50
+
+    # poolMin: Pool 最小容量 - 即使没人用，也保持这么多 Pod
+    # - 低峰期保底容量，避免冷启动
+    # - 通常与 bufferMin 相同或稍大
+    poolMin: 10
+
+    # poolMax: Pool 最大容量 - 高峰期最多这么多 Pod
+    # - 限制最大资源使用，避免集群资源耗尽
+    # - 根据集群资源和业务峰值设置
+    poolMax: 200
+
+    # 容量规划示例：
+    # 场景：200 个并发 Agent，平均每个 Agent 会话 5 分钟
+    #   bufferMin: 50  （保证 50 个 Agent 立即可用）
+    #   bufferMax: 100 （预热 100 个，覆盖短期突发）
+    #   poolMin: 50    （低峰期保持 50 个）
+    #   poolMax: 300   （高峰期最多 300 个）
+
+# ==============================================================================
+# SDK 使用示例
+# ==============================================================================
+#
+# Python SDK 使用此 Pool：
+#
+# ```python
+# import asyncio
+# from datetime import timedelta
+# from opensandbox import Sandbox
+# from opensandbox.config import ConnectionConfig
+#
+# async def create_agent_sandbox(agent_id: str):
+#     """为 Agent 创建 sandbox"""
+#     sandbox = await Sandbox.create(
+#         "nginx:latest",  # 镜像会被忽略，使用 Pool 中的镜像
+#         entrypoint=["/bin/sh", "-c", "sleep infinity"],  # 可自定义
+#         env={"AGENT_ID": agent_id},  # 可传递环境变量
+#         timeout=timedelta(hours=1),
+#         connection_config=ConnectionConfig(domain="<server-ip>:8088"),
+#         extensions={"poolRef": "agent-pool"}  # 指定 Pool 名称
+#     )
+#     return sandbox
+#
+# async def handle_agent_request(agent_id: str, task: str):
+#     """处理单个 Agent 请求"""
+#     sandbox = await create_agent_sandbox(agent_id)
+#     try:
+#         result = await sandbox.commands.run(task)
+#         return result
+#     finally:
+#         await sandbox.kill()  # Pod 返回 Pool
+# ```
+#
+# ==============================================================================
+# 部署和监控
+# ==============================================================================
+#
+# 1. 部署 Pool：
+#    kubectl apply -f pool-agent-production.yaml
+#
+# 2. 验证 Pool 状态：
+#    kubectl get pool agent-pool -n default
+#    kubectl get pool agent-pool -n default -o jsonpath='{.status}' | jq
+#
+# 3. 查看 Pool 的 Pod：
+#    kubectl get pods -l pool=agent-pool -n default
+#
+# 4. 监控 Pool 使用率：
+#    watch kubectl get pool agent-pool -n default -o jsonpath='{.status}'
+#    # 输出示例：
+#    # {
+#    #   "total": 50,      # 总 Pod 数
+#    #   "allocated": 30,  # 已分配
+#    #   "available": 20   # 可用
+#    # }
+#
+# 5. 优化建议：
+#    - 如果 available 经常为 0 → 增加 bufferMax
+#    - 如果 available 总是接近 total → 减少 bufferMin
+#    - 如果 total 经常达到 poolMax → 增加 poolMax 或优化 Agent 使用
+#
+# ==============================================================================
+# 故障排查
+# ==============================================================================
+#
+# Pool Pod 无法启动：
+#   kubectl describe pod -l pool=agent-pool -n default
+#   kubectl logs -l pool=agent-pool -n default -c sandbox-container
+#   kubectl logs -l pool=agent-pool -n default -c task-executor
+#
+# execd 连接失败：
+#   kubectl exec -it <pod-name> -n default -c sandbox-container -- ps aux | grep execd
+#   kubectl exec -it <pod-name> -n default -c sandbox-container -- nc -zv localhost 44772
+#
+# task-executor 权限问题：
+#   kubectl get pod <pod-name> -n default -o yaml | grep -A 10 securityContext
+#
+# ==============================================================================

deploy/helm/opensandbox/examples/pool-examples.md

+# Pool示例 - 包含Task Executor Sidecar
+
+## 基本Pool（不包含任务执行）
+
+```yaml
+apiVersion: sandbox.opensandbox.io/v1alpha1
+kind: Pool
+metadata:
+  name: basic-pool
+  namespace: default
+spec:
+  template:
+    spec:
+      containers:
+      - name: sandbox-container
+        image: ubuntu:22.04
+        command: ["sleep", "infinity"]
+        resources:
+          requests:
+            cpu: "100m"
+            memory: "128Mi"
+          limits:
+            cpu: "500m"
+            memory: "256Mi"
+  capacitySpec:
+    bufferMax: 10
+    bufferMin: 2
+    poolMax: 20
+    poolMin: 5
+```
+
+## Pool with Task Executor（支持任务执行）
+
+**重要提示**：
+- Task Executor作为sidecar容器运行在Pool的Pod中
+- 必须启用`shareProcessNamespace: true`以共享进程命名空间
+- Task Executor需要`SYS_PTRACE`权限来注入进程
+
+```yaml
+apiVersion: sandbox.opensandbox.io/v1alpha1
+kind: Pool
+metadata:
+  name: task-enabled-pool
+  namespace: default
+spec:
+  template:
+    spec:
+      # 必需：共享进程命名空间，允许task-executor访问sandbox容器的进程
+      shareProcessNamespace: true
+      containers:
+      # 主容器：沙箱环境
+      - name: sandbox-container
+        image: ubuntu:22.04
+        command: ["sleep", "infinity"]
+        resources:
+          requests:
+            cpu: "100m"
+            memory: "128Mi"
+          limits:
+            cpu: "500m"
+            memory: "256Mi"
+
+      # Sidecar：Task Executor（用于任务注入）
+      - name: task-executor
+        # 使用Helm values中配置的镜像
+        # {{ .Values.taskExecutor.image.repository }}:{{ .Values.taskExecutor.image.tag }}
+        image: opensandbox.io/task-executor:v0.0.1
+        imagePullPolicy: IfNotPresent
+        resources:
+          requests:
+            cpu: "100m"
+            memory: "128Mi"
+          limits:
+            cpu: "500m"
+            memory: "256Mi"
+        securityContext:
+          # 必需：需要ptrace权限来注入进程到sandbox容器
+          capabilities:
+            add: ["SYS_PTRACE"]
+
+  capacitySpec:
+    bufferMax: 10
+    bufferMin: 2
+    poolMax: 20
+    poolMin: 5
+```
+
+## BatchSandbox with Tasks（使用Pool执行任务）
+
+创建使用上述Pool的BatchSandbox，并执行异构任务：
+
+```yaml
+apiVersion: sandbox.opensandbox.io/v1alpha1
+kind: BatchSandbox
+metadata:
+  name: task-batch-sandbox
+  namespace: default
+spec:
+  # 副本数量
+  replicas: 3
+
+  # 引用包含task-executor的Pool
+  poolRef: task-enabled-pool
+
+  # TTL：3600秒后自动清理
+  ttlSecondsAfterFinished: 3600
+
+  # 默认任务模板（所有沙箱共享）
+  taskTemplate:
+    spec:
+      process:
+        command: ["echo", "Default task"]
+
+  # 异构任务：为每个沙箱自定义不同的任务
+  shardTaskPatches:
+  - spec:
+      process:
+        command: ["python3", "-c", "print('Task for sandbox 0')"]
+  - spec:
+      process:
+        command: ["bash", "-c", "echo 'Task for sandbox 1' && sleep 5"]
+  - spec:
+      process:
+        command: ["node", "-e", "console.log('Task for sandbox 2')"]
+```
+
+## 镜像配置说明
+
+### 方式1：使用Helm Values配置
+
+在`values.yaml`中配置task-executor镜像：
+
+```yaml
+taskExecutor:
+  image:
+    repository: your-registry/opensandbox-task-executor
+    tag: "v1.0.0"
+    pullPolicy: IfNotPresent
+```
+
+然后在Pool YAML中引用：
+
+```yaml
+image: your-registry/opensandbox-task-executor:v1.0.0
+```
+
+### 方式2：使用环境变量（ConfigMap）
+
+创建ConfigMap存储镜像信息：
+
+```yaml
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: opensandbox-images
+  namespace: default
+data:
+  taskExecutorImage: "your-registry/opensandbox-task-executor:v1.0.0"
+```
+
+在应用层读取ConfigMap并创建Pool。
+
+### 方式3：使用Kustomize替换
+
+使用Kustomize的镜像替换功能：
+
+```yaml
+# kustomization.yaml
+images:
+- name: opensandbox.io/task-executor
+  newName: your-registry/opensandbox-task-executor
+  newTag: v1.0.0
+```
+
+## 验证Task Executor
+
+创建资源后，验证task-executor是否正常运行：
+
+```bash
+# 查看Pool状态
+kubectl get pools task-enabled-pool
+
+# 查看Pool创建的Pod
+kubectl get pods -l pool=task-enabled-pool
+
+# 检查Pod中是否有task-executor容器
+kubectl get pods -l pool=task-enabled-pool -o jsonpath='{.items[0].spec.containers[*].name}'
+# 输出应包含: sandbox-container task-executor
+
+# 查看task-executor日志
+kubectl logs <pod-name> -c task-executor
+
+# 查看BatchSandbox任务状态
+kubectl get batchsandbox task-batch-sandbox -o wide
+# 应显示: TASK_RUNNING, TASK_SUCCEED, TASK_FAILED 等状态
+```
+
+## 故障排查
+
+### Task Executor无法启动
+
+```bash
+# 检查容器状态
+kubectl describe pod <pod-name>
+
+# 检查权限问题
+kubectl get pod <pod-name> -o jsonpath='{.spec.containers[1].securityContext}'
+# 应显示: {"capabilities":{"add":["SYS_PTRACE"]}}
+
+# 检查进程命名空间共享
+kubectl get pod <pod-name> -o jsonpath='{.spec.shareProcessNamespace}'
+# 应显示: true
+```
+
+### 任务执行失败
+
+```bash
+# 查看任务状态
+kubectl describe batchsandbox task-batch-sandbox
+
+# 查看task-executor日志
+kubectl logs <pod-name> -c task-executor -f
+
+# 查看sandbox容器日志
+kubectl logs <pod-name> -c sandbox-container
+```
+
+## 性能考虑
+
+- **资源配置**：根据任务复杂度调整task-executor的资源限制
+- **并发控制**：Pool的`bufferMax`和`poolMax`控制并发沙箱数量
+- **任务超时**：在taskTemplate中配置超时时间防止任务卡死
+- **清理策略**：使用`ttlSecondsAfterFinished`自动清理完成的沙箱
+
+## 最佳实践
+
+1. **镜像版本管理**：controller和task-executor镜像版本保持一致
+2. **资源限制**：task-executor通常需要更多CPU用于进程注入
+3. **安全配置**：只在需要时启用`SYS_PTRACE`权限
+4. **任务设计**：将长时间运行的任务拆分为多个短任务
+5. **监控告警**：监控任务失败率和执行时间

deploy/helm/opensandbox/examples/pool-kubectl-only.yaml

+# ==============================================================================
+# Pool for kubectl-only scenarios（纯 kubectl 场景 Pool）
+# ==============================================================================
+#
+# 用途：纯 kubectl 场景，不使用 SDK
+#
+# 适用场景：
+#   - RL 训练批量任务
+#   - 压力测试
+#   - 批量数据处理
+#   - 不需要 SDK 动态交互的场景
+#
+# 特点：
+#   - 没有 execd（不支持 SDK）
+#   - 包含 task-executor（支持 taskTemplate）
+#   - 配置简洁，资源占用少
+#
+# ==============================================================================
+
+apiVersion: sandbox.opensandbox.io/v1alpha1
+kind: Pool
+metadata:
+  name: kubectl-pool
+  namespace: default
+  labels:
+    app: opensandbox
+    component: kubectl-pool
+  annotations:
+    description: "纯 kubectl 场景 Pool，不包含 execd"
+spec:
+  template:
+    metadata:
+      labels:
+        pool: kubectl-pool
+        sdk-compatible: "false"
+    spec:
+      # ========================================
+      # 必需：共享进程命名空间（task-executor 需要）
+      # ========================================
+      shareProcessNamespace: true
+
+      # ========================================
+      # 主容器：Sandbox 环境
+      # ========================================
+      containers:
+      - name: sandbox-container
+        image: nginx:latest
+        imagePullPolicy: IfNotPresent
+
+        # 直接运行业务命令，不需要 bootstrap.sh
+        command: ["sleep", "infinity"]
+
+        # 资源配置
+        resources:
+          requests:
+            cpu: "100m"
+            memory: "128Mi"
+          limits:
+            cpu: "500m"
+            memory: "256Mi"
+
+      # ========================================
+      # Sidecar：Task Executor
+      # ========================================
+      - name: task-executor
+        image: opensandbox/task-executor:dev
+        imagePullPolicy: Never
+
+        ports:
+        - containerPort: 5758
+          name: task-executor
+          protocol: TCP
+
+        resources:
+          requests:
+            cpu: "100m"
+            memory: "128Mi"
+          limits:
+            cpu: "500m"
+            memory: "256Mi"
+
+        # 安全上下文
+        securityContext:
+          capabilities:
+            add: ["SYS_PTRACE"]
+
+  # ========================================
+  # Pool 容量配置
+  # ========================================
+  capacitySpec:
+    bufferMin: 2
+    bufferMax: 5
+    poolMin: 2
+    poolMax: 10

deploy/helm/opensandbox/examples/pool-sdk-compatible.yaml

+apiVersion: sandbox.opensandbox.io/v1alpha1
+kind: Pool
+metadata:
+  name: sdk-pool
+  namespace: default
+spec:
+  template:
+    spec:
+      # Init container: 安装execd
+      initContainers:
+      - name: execd-installer
+        image: opensandbox/execd:v1.0.5
+        command: ["/bin/sh", "-c"]
+        args:
+          - |
+            cp ./execd /opt/opensandbox/bin/execd && \
+            cp ./bootstrap.sh /opt/opensandbox/bin/bootstrap.sh && \
+            chmod +x /opt/opensandbox/bin/execd && \
+            chmod +x /opt/opensandbox/bin/bootstrap.sh
+        volumeMounts:
+        - name: opensandbox-bin
+          mountPath: /opt/opensandbox/bin
+
+      # 主容器：带execd
+      containers:
+      - name: sandbox-container
+        image: nginx:latest
+        command:
+        - /opt/opensandbox/bin/bootstrap.sh
+        - nginx
+        - -g
+        - daemon off;  # nginx前台运行
+        env:
+        - name: EXECD
+          value: /opt/opensandbox/bin/execd
+        ports:
+        - containerPort: 80
+          name: http
+        - containerPort: 44772
+          name: execd
+          protocol: TCP
+        resources:
+          requests:
+            cpu: "100m"
+            memory: "128Mi"
+          limits:
+            cpu: "500m"
+            memory: "256Mi"
+        volumeMounts:
+        - name: opensandbox-bin
+          mountPath: /opt/opensandbox/bin
+
+      # 共享卷
+      volumes:
+      - name: opensandbox-bin
+        emptyDir: {}
+
+  capacitySpec:
+    bufferMax: 10
+    bufferMin: 2
+    poolMax: 20
+    poolMin: 5

deploy/helm/opensandbox/examples/pool-sdk-with-tasks.yaml

+apiVersion: sandbox.opensandbox.io/v1alpha1
+kind: Pool
+metadata:
+  name: sdk-pool-with-tasks
+  namespace: default
+spec:
+  template:
+    spec:
+      shareProcessNamespace: true  # task-executor需要
+
+      # Init container: 安装execd
+      initContainers:
+      - name: execd-installer
+        image: opensandbox/execd:v1.0.5
+        command: ["/bin/sh", "-c"]
+        args:
+          - |
+            cp ./execd /opt/opensandbox/bin/execd && \
+            cp ./bootstrap.sh /opt/opensandbox/bin/bootstrap.sh && \
+            chmod +x /opt/opensandbox/bin/execd && \
+            chmod +x /opt/opensandbox/bin/bootstrap.sh
+        volumeMounts:
+        - name: opensandbox-bin
+          mountPath: /opt/opensandbox/bin
+
+      containers:
+      # 主容器：带execd
+      - name: sandbox-container
+        image: nginx:latest
+        command:
+        - /opt/opensandbox/bin/bootstrap.sh
+        - sleep
+        - infinity
+        env:
+        - name: EXECD
+          value: /opt/opensandbox/bin/execd
+        ports:
+        - containerPort: 44772
+          name: execd
+          protocol: TCP
+        resources:
+          requests:
+            cpu: "100m"
+            memory: "128Mi"
+          limits:
+            cpu: "500m"
+            memory: "256Mi"
+        volumeMounts:
+        - name: opensandbox-bin
+          mountPath: /opt/opensandbox/bin
+
+      # task-executor sidecar: 支持自定义entrypoint
+      - name: task-executor
+        image: opensandbox/task-executor:dev
+        imagePullPolicy: Never  # 使用本地镜像
+        securityContext:
+          capabilities:
+            add: ["SYS_PTRACE"]
+        resources:
+          requests:
+            cpu: "100m"
+            memory: "128Mi"
+          limits:
+            cpu: "500m"
+            memory: "256Mi"
+
+      volumes:
+      - name: opensandbox-bin
+        emptyDir: {}
+
+  capacitySpec:
+    bufferMax: 10
+    bufferMin: 2
+    poolMax: 20
+    poolMin: 5

deploy/helm/opensandbox/scripts/README.md

+# OpenSandbox Helm Chart Scripts
+
+This directory contains utility scripts for OpenSandbox Controller deployment and testing.
+
+## Script List
+
+### 1. install.sh - Installation Script
+
+Interactive installation of OpenSandbox Controller to Kubernetes cluster.
+
+**Features:**
+- Automatic detection of sudo privilege requirements
+- Validation of dependency tools (helm, kubectl)
+- Cluster connection verification
+- Support for multiple deployment environments:
+  - Default configuration (values.yaml)
+  - E2E testing (values-e2e.yaml)
+  - Custom configuration (via --set or custom values file)
+- Helm Chart validation
+- Display verification commands after deployment
+
+**Usage:**
+```bash
+cd scripts
+./install.sh
+```
+
+**Environment Variables:**
+- `IMAGE_REPO` - Override controller image repository
+- `IMAGE_TAG` - Override controller image tag
+- `SERVER_IMAGE_REPO` - Override server image repository
+- `SERVER_IMAGE_TAG` - Override server image tag
+
+**Example:**
+```bash
+# Using custom images
+IMAGE_REPO=myregistry.com/controller \
+IMAGE_TAG=v1.0.0 \
+SERVER_IMAGE_REPO=myregistry.com/server \
+SERVER_IMAGE_TAG=v0.1.0 \
+./install.sh
+```
+
+### 2. uninstall.sh - Uninstallation Script
+
+Uninstall OpenSandbox Controller and clean up related resources.
+
+**Features:**
+- Check running BatchSandbox and Pool resources
+- Display Controller and Server deployment status
+- Optional CRD deletion
+- Optional namespace deletion
+- Post-uninstall cleanup verification
+
+**Usage:**
+```bash
+cd scripts
+./uninstall.sh
+```
+
+**Environment Variables:**
+- `RELEASE_NAME` - Release name (default: opensandbox-controller)
+- `NAMESPACE` - Namespace (default: opensandbox)
+
+**Example:**
+```bash
+# Uninstall specific release
+RELEASE_NAME=my-release NAMESPACE=my-namespace ./uninstall.sh
+```
+
+### 3. e2e-test.sh - End-to-End Test Script
+
+Execute complete end-to-end test workflow.
+
+**Test Workflow:**
+1. Helm Install (using values-e2e.yaml)
+2. Verify Controller and Server deployment
+3. Verify Pool deployment
+4. Verify SDK calls
+5. Helm Uninstall
+
+**Features:**
+- Automatic Server port-forward setup
+- Server API health check validation
+- Pool Pod execd process verification
+- SDK integration test execution
+- Automatic resource cleanup (including port-forward processes)
+
+**Usage:**
+```bash
+cd scripts
+./e2e-test.sh [VALUES_FILE]
+
+# Using default values-e2e.yaml
+./e2e-test.sh
+
+# Using custom values file
+./e2e-test.sh custom-values.yaml
+```
+
+**Prerequisites:**
+- Required Docker images must be loaded:
+  - opensandbox/controller:dev
+  - opensandbox/server:v0.1.0
+  - opensandbox/task-executor:dev
+  - opensandbox/execd:v1.0.5
+  - nginx:latest
+- Python SDK installed (using uv)
+- Cluster has sufficient resources to run test Pods
+
+## General Instructions
+
+### Sudo Privileges
+
+All scripts automatically detect whether sudo privileges are required to execute kubectl and helm commands.
+
+### Script Paths
+
+Scripts use relative paths to locate the Chart directory and can be invoked from any location:
+```bash
+# From chart root directory
+./scripts/install.sh
+
+# From scripts directory
+cd scripts
+./install.sh
+
+# From other directory
+/path/to/opensandbox-controller/scripts/install.sh
+```
+
+### Colored Output
+
+Scripts use ANSI color codes to enhance readability:
+- 🟢 Green - Success messages
+- 🟡 Yellow - Warnings and step titles
+- 🔴 Red - Error messages
+
+### Error Handling
+
+All scripts use `set -e`, exiting immediately on errors. The e2e-test.sh uses trap to ensure cleanup functions execute on exit.
+
+## Troubleshooting
+
+### install.sh
+
+**Issue: Cannot connect to Kubernetes cluster**
+```bash
+# Check kubeconfig
+kubectl cluster-info
+
+# Check context
+kubectl config current-context
+```
+
+**Issue: Chart validation fails**
+```bash
+# Manual validation
+helm lint ../
+```
+
+### uninstall.sh
+
+**Issue: Resources are still running**
+```bash
+# View all BatchSandbox
+kubectl get batchsandboxes -A
+
+# View all Pool
+kubectl get pools -A
+
+# Delete all resources
+kubectl delete batchsandboxes --all -A
+kubectl delete pools --all -A
+```
+
+### e2e-test.sh
+
+**Issue: Port-forward fails**
+```bash
+# Check if any process is using port 8080
+lsof -i :8080
+
+# Manual port-forward test
+kubectl port-forward -n opensandbox svc/opensandbox-controller-server 8080:8080
+```
+
+**Issue: SDK test fails**
+```bash
+# Check Server logs
+kubectl logs -n opensandbox -l app.kubernetes.io/component=server
+
+# Check Pool Pod logs
+kubectl logs -n opensandbox -l pool=agent-pool
+
+# Test Server API
+curl http://localhost:8080/health
+```
+
+**Issue: Image not found**
+```bash
+# Check if images are loaded
+docker images | grep opensandbox
+
+# Reload images
+docker load -i /path/to/image.tar
+```
+
+## Development Guide
+
+### Modifying Scripts
+
+After modifying scripts, ensure:
+1. Maintain consistent error handling
+2. Update related documentation
+3. Use meaningful colored output
+4. Add appropriate validation steps
+
+### Adding New Scripts
+
+New scripts should follow these conventions:
+- Use `#!/bin/bash` shebang
+- Use `set -e` to enable exit-on-error
+- Implement automatic sudo detection
+- Add colored output for readability
+- Add documentation in this README
+
+## License
+
+Apache License 2.0

deploy/helm/opensandbox/scripts/e2e-test.sh

+#!/bin/bash
+# OpenSandbox Helm Chart End-to-End Test Script
+set -x
+set -e
+
+# Get the parent directory of the script's directory (chart root directory)
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+CHART_DIR="$(dirname "$SCRIPT_DIR")"
+NAMESPACE="opensandbox"
+RELEASE_NAME="opensandbox-e2e-test"
+VALUES_FILE="${1:-values-e2e.yaml}"
+
+# Cleanup function: ensure temporary resources are cleaned up
+cleanup() {
+    # Clean up any test resources that might be left over
+    if command -v kubectl &> /dev/null || command -v minikube &> /dev/null; then
+        kubectl delete configmap sdk-test-script -n "$NAMESPACE" --ignore-not-found=true > /dev/null 2>&1 || true
+        kubectl delete job sdk-test-job -n "$NAMESPACE" --ignore-not-found=true > /dev/null 2>&1 || true
+        kubectl delete pod server-health-check -n "$NAMESPACE" --ignore-not-found=true > /dev/null 2>&1 || true
+    fi
+}
+
+# Register cleanup function to ensure execution on script exit
+trap cleanup EXIT INT TERM
+
+# Check kubectl availability and detect minikube
+USE_SUDO=false
+USE_MINIKUBE=false
+
+# First check if kubectl is available directly
+if kubectl get nodes &> /dev/null 2>&1; then
+    echo "Detected kubectl access to Kubernetes cluster"
+elif sudo kubectl get nodes &> /dev/null 2>&1; then
+    echo "Detected sudo privileges required, will use sudo for commands"
+    USE_SUDO=true
+# If kubectl is not available, check for minikube
+elif command -v minikube &> /dev/null && minikube kubectl -- get nodes &> /dev/null 2>&1; then
+    echo "Detected minikube cluster, will use 'minikube kubectl --' for commands"
+    USE_MINIKUBE=true
+elif command -v minikube &> /dev/null && sudo minikube kubectl -- get nodes &> /dev/null 2>&1; then
+    echo "Detected minikube cluster with sudo, will use 'sudo minikube kubectl --' for commands"
+    USE_MINIKUBE=true
+    USE_SUDO=true
+else
+    echo "Error: Unable to access Kubernetes cluster"
+    echo "Please ensure kubectl or minikube is properly configured"
+    exit 1
+fi
+
+# Define command functions
+kubectl_cmd() {
+    if [ "$USE_MINIKUBE" = true ]; then
+        if [ "$USE_SUDO" = true ]; then
+            sudo minikube kubectl -- "$@"
+        else
+            minikube kubectl -- "$@"
+        fi
+    else
+        if [ "$USE_SUDO" = true ]; then
+            sudo kubectl "$@"
+        else
+            kubectl "$@"
+        fi
+    fi
+}
+
+helm_cmd() {
+    if [ "$USE_SUDO" = true ]; then
+        sudo helm "$@"
+    else
+        helm "$@"
+    fi
+}
+
+GREEN='\033[0;32m'
+YELLOW='\033[1;33m'
+RED='\033[0;31m'
+NC='\033[0m'
+
+echo -e "${GREEN}==========================================${NC}"
+echo -e "${GREEN}OpenSandbox Helm Chart E2E Validation${NC}"
+echo -e "${GREEN}==========================================${NC}"
+echo ""
+echo "Test Coverage:"
+echo "  1. Helm Install (using ${VALUES_FILE})"
+echo "  2. Server Deployment Verification"
+echo "  3. Pool Deployment Verification"
+echo "  4. SDK Integration Verification"
+echo "  5. Helm Uninstall"
+echo ""
+echo "Environment Info:"
+echo "  Chart: ${CHART_DIR}"
+echo "  Values: ${VALUES_FILE}"
+echo "  Release: ${RELEASE_NAME}"
+echo "  Namespace: ${NAMESPACE}"
+echo ""
+
+# ==========================================
+# Stage 1: Helm Install
+# ==========================================
+echo -e "${GREEN}==========================================${NC}"
+echo -e "${GREEN}Stage 1: Helm Install${NC}"
+echo -e "${GREEN}==========================================${NC}"
+echo ""
+
+echo -e "${YELLOW}[1.1] Checking for existing release...${NC}"
+if helm_cmd list -n "$NAMESPACE" 2>/dev/null | grep -q "$RELEASE_NAME"; then
+    echo "  Release already exists, uninstalling first..."
+    helm_cmd uninstall "$RELEASE_NAME" -n "$NAMESPACE" 2>/dev/null || true
+    sleep 5
+fi
+echo -e "${GREEN}✓ Check completed${NC}"
+echo ""
+
+echo -e "${YELLOW}[1.2] Installing Helm chart (using ${VALUES_FILE})...${NC}"
+helm_cmd install "$RELEASE_NAME" "$CHART_DIR" \
+    --values "$CHART_DIR/$VALUES_FILE" \
+    --namespace "$NAMESPACE" \
+    --create-namespace \
+    --wait \
+    --timeout 3m 2>&1 | tail -5
+echo -e "${GREEN}✓ Helm chart installed successfully${NC}"
+echo ""
+
+echo -e "${YELLOW}[1.3] Waiting for Controller to be ready...${NC}"
+kubectl_cmd wait --for=condition=available \
+    deployment/opensandbox-controller-manager \
+    -n "$NAMESPACE" \
+    --timeout=120s 2>/dev/null
+echo -e "${GREEN}✓ Controller is ready${NC}"
+echo ""
+
+echo -e "${YELLOW}[1.4] Checking deployment status...${NC}"
+kubectl_cmd get deployment -n "$NAMESPACE"
+echo ""
+
+echo -e "${GREEN}✅ Stage 1 Complete: Helm Install Successful${NC}"
+echo ""
+
+# ==========================================
+# Stage 2: Server Deployment Verification
+# ==========================================
+echo -e "${GREEN}==========================================${NC}"
+echo -e "${GREEN}Stage 2: Server Deployment Verification${NC}"
+echo -e "${GREEN}==========================================${NC}"
+echo ""
+
+echo -e "${YELLOW}[2.1] Checking Server Service...${NC}"
+SERVER_SERVICE_NAME=$(kubectl_cmd get svc -n "$NAMESPACE" -l app.kubernetes.io/component=server -o jsonpath='{.items[0].metadata.name}' 2>/dev/null)
+if [ -z "$SERVER_SERVICE_NAME" ]; then
+    echo -e "${RED}❌ Server Service does not exist${NC}"
+    exit 1
+fi
+echo "  Server Service: ${SERVER_SERVICE_NAME}"
+kubectl_cmd get svc "$SERVER_SERVICE_NAME" -n "$NAMESPACE"
+echo ""
+
+echo -e "${YELLOW}[2.2] Waiting for Server Pod to be ready...${NC}"
+SERVER_DEPLOYMENT_NAME=$(kubectl_cmd get deployment -n "$NAMESPACE" -l app.kubernetes.io/component=server -o jsonpath='{.items[0].metadata.name}' 2>/dev/null)
+if [ -z "$SERVER_DEPLOYMENT_NAME" ]; then
+    echo -e "${RED}❌ Server Deployment does not exist${NC}"
+    exit 1
+fi
+echo "  Server Deployment: ${SERVER_DEPLOYMENT_NAME}"
+kubectl_cmd wait --for=condition=available \
+    deployment/"$SERVER_DEPLOYMENT_NAME" \
+    -n "$NAMESPACE" \
+    --timeout=120s 2>/dev/null
+echo -e "${GREEN}✓ Server Deployment is ready${NC}"
+echo ""
+
+echo -e "${YELLOW}[2.3] Checking Server Pod status...${NC}"
+kubectl_cmd get pods -n "$NAMESPACE" -l app.kubernetes.io/component=server
+echo ""
+
+echo -e "${YELLOW}[2.4] Testing Server API (from within cluster)...${NC}"
+# Create a simple test pod to check server health from inside the cluster
+cat <<EOF | kubectl_cmd apply -f - > /dev/null
+apiVersion: v1
+kind: Pod
+metadata:
+  name: server-health-check
+  namespace: $NAMESPACE
+spec:
+  restartPolicy: Never
+  containers:
+  - name: curl
+    image: curlimages/curl:latest
+    command: ['curl', '-s', 'http://${SERVER_SERVICE_NAME}:8080/health']
+EOF
+
+# Wait for pod to complete
+kubectl_cmd wait --for=condition=Ready pod/server-health-check -n "$NAMESPACE" --timeout=30s 2>/dev/null || true
+sleep 2
+HEALTH_RESPONSE=$(kubectl_cmd logs server-health-check -n "$NAMESPACE" 2>/dev/null || echo "")
+kubectl_cmd delete pod server-health-check -n "$NAMESPACE" --ignore-not-found=true > /dev/null 2>&1
+
+if [ -n "$HEALTH_RESPONSE" ]; then
+    echo -e "${GREEN}✓ Server API responding normally: $HEALTH_RESPONSE${NC}"
+else
+    echo -e "${RED}❌ Server API not responding${NC}"
+    exit 1
+fi
+echo ""
+
+echo -e "${GREEN}✅ Stage 2 Complete: Server Deployment Verified${NC}"
+echo ""
+
+# ==========================================
+# Stage 3: Pool Deployment Verification
+# ==========================================
+echo -e "${GREEN}==========================================${NC}"
+echo -e "${GREEN}Stage 3: Pool Deployment Verification${NC}"
+echo -e "${GREEN}==========================================${NC}"
+echo ""
+
+echo -e "${YELLOW}[3.1] Checking Pool resources...${NC}"
+POOL_COUNT=$(kubectl_cmd get pool -n "$NAMESPACE" --no-headers 2>/dev/null | wc -l)
+echo "  Pool count: ${POOL_COUNT}"
+if [ "$POOL_COUNT" -eq 0 ]; then
+    echo -e "${RED}❌ No Pool resources found${NC}"
+    exit 1
+fi
+kubectl_cmd get pool -n "$NAMESPACE"
+echo ""
+
+echo -e "${YELLOW}[3.2] Checking agent-pool status...${NC}"
+if ! kubectl_cmd get pool agent-pool -n "$NAMESPACE" --no-headers 2>/dev/null | grep -q agent-pool; then
+    echo -e "${RED}❌ agent-pool does not exist${NC}"
+    exit 1
+fi
+echo -e "${GREEN}✓ agent-pool exists${NC}"
+echo ""
+
+echo -e "${YELLOW}[3.3] Viewing Pool detailed status...${NC}"
+kubectl_cmd get pool agent-pool -n "$NAMESPACE" -o jsonpath='{.status}' 2>/dev/null | jq '.' 2>/dev/null || echo "  (jq not installed, skipping JSON formatting)"
+echo ""
+
+echo -e "${YELLOW}[3.4] Waiting for Pool Pods to be ready (up to 180 seconds)...${NC}"
+TIMEOUT=180
+ELAPSED=0
+READY=false
+while [ $ELAPSED -lt $TIMEOUT ]; do
+    AVAILABLE=$(kubectl_cmd get pool agent-pool -n "$NAMESPACE" -o jsonpath='{.status.available}' 2>/dev/null || echo "")
+    if [ -n "$AVAILABLE" ] && [ "$AVAILABLE" -gt 0 ]; then
+        echo -e "${GREEN}✓ Pool has ${AVAILABLE} available Pods${NC}"
+        READY=true
+        break
+    fi
+    echo "  Waiting... (${ELAPSED}s/${TIMEOUT}s)"
+    sleep 5
+    ELAPSED=$((ELAPSED + 5))
+done
+if [ "$READY" = false ]; then
+    echo -e "${RED}❌ Pool Pods not ready, timed out${NC}"
+    echo ""
+    echo "Viewing Pool events:"
+    kubectl_cmd describe pool agent-pool -n "$NAMESPACE" | tail -20
+    exit 1
+fi
+echo ""
+
+echo -e "${YELLOW}[3.5] Viewing Pool Pods...${NC}"
+kubectl_cmd get pods -l pool=agent-pool -n "$NAMESPACE"
+echo ""
+
+echo -e "${YELLOW}[3.6] Checking execd process in Pod...${NC}"
+POD_NAME=$(kubectl_cmd get pods -l pool=agent-pool -n "$NAMESPACE" -o jsonpath='{.items[0].metadata.name}' 2>/dev/null)
+if [ -n "$POD_NAME" ]; then
+    echo "  Checking Pod: ${POD_NAME}"
+    sleep 3
+    if kubectl_cmd exec -n "$NAMESPACE" "$POD_NAME" -c sandbox-container -- pgrep -f execd > /dev/null 2>&1; then
+        EXECD_PID=$(kubectl_cmd exec -n "$NAMESPACE" "$POD_NAME" -c sandbox-container -- pgrep -f execd 2>/dev/null)
+        echo -e "${GREEN}✓ execd process is running (PID: ${EXECD_PID})${NC}"
+    else
+        echo -e "${YELLOW}⚠️  execd process not found, checking container logs...${NC}"
+        kubectl_cmd logs -n "$NAMESPACE" "$POD_NAME" -c sandbox-container --tail=20 2>/dev/null || true
+    fi
+else
+    echo -e "${YELLOW}⚠️  No agent-pool Pod found${NC}"
+fi
+echo ""
+
+echo -e "${GREEN}✅ Stage 3 Complete: Pool Deployment Verified${NC}"
+echo ""
+
+# ==========================================
+# Stage 4: SDK Integration Verification
+# ==========================================
+echo -e "${GREEN}==========================================${NC}"
+echo -e "${GREEN}Stage 4: SDK Integration Verification${NC}"
+echo -e "${GREEN}==========================================${NC}"
+echo ""
+
+echo -e "${YELLOW}[4.1] Creating SDK test script ConfigMap...${NC}"
+SDK_TEST_SCRIPT=$(cat <<'EOF'
+import asyncio
+from datetime import timedelta
+from opensandbox import Sandbox
+from opensandbox.config import ConnectionConfig
+
+async def main():
+    print("=" * 60)
+    print("SDK End-to-End Test")
+    print("=" * 60)
+
+    # Use internal service name for cluster-internal communication
+    config = ConnectionConfig(domain="SERVER_SERVICE_PLACEHOLDER:8080")
+
+    print("\n[Test 1] Creating sandbox (using agent-pool)...")
+    try:
+        sandbox = await Sandbox.create(
+            "nginx:latest",
+            entrypoint=["/bin/sh", "-c", "sleep infinity"],
+            env={"TEST": "e2e"},
+            timeout=timedelta(minutes=10),
+            ready_timeout=timedelta(minutes=5),
+            connection_config=config,
+            extensions={"poolRef": "agent-pool"}
+        )
+        print(f"✅ Sandbox created successfully: {sandbox.id}")
+    except Exception as e:
+        print(f"❌ Sandbox creation failed: {e}")
+        import traceback
+        traceback.print_exc()
+        return False
+
+    try:
+        print("\n[Test 2] Executing command...")
+        execution = await sandbox.commands.run("echo 'Hello from E2E test'")
+        if execution.logs.stdout:
+            print(f"✅ Command executed successfully: {execution.logs.stdout[0].text}")
+        else:
+            print("⚠️  Command executed successfully but no output")
+
+        print("\n[Test 3] File operations...")
+        from opensandbox.models import WriteEntry
+        await sandbox.files.write_files([
+            WriteEntry(path="/tmp/e2e.txt", data="E2E Test", mode=644)
+        ])
+        print("✅ File written successfully")
+
+        content = await sandbox.files.read_file("/tmp/e2e.txt")
+        print(f"✅ File read successfully: {content}")
+
+        print("\n[Test 4] Cleaning up sandbox...")
+        await sandbox.kill()
+        print("✅ Sandbox cleaned up successfully")
+
+        print("\n" + "=" * 60)
+        print("✅ All SDK end-to-end tests passed!")
+        print("=" * 60)
+        return True
+
+    except Exception as e:
+        print(f"❌ Test failed: {e}")
+        import traceback
+        traceback.print_exc()
+        try:
+            await sandbox.kill()
+        except:
+            pass
+        return False
+
+if __name__ == "__main__":
+    success = asyncio.run(main())
+    exit(0 if success else 1)
+EOF
+)
+
+# Replace placeholder with actual service name
+SDK_TEST_SCRIPT="${SDK_TEST_SCRIPT//SERVER_SERVICE_PLACEHOLDER/${SERVER_SERVICE_NAME}}"
+
+# Create ConfigMap with test script
+cat <<EOF | kubectl_cmd apply -f - > /dev/null
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: sdk-test-script
+  namespace: $NAMESPACE
+data:
+  test.py: |
+$(echo "$SDK_TEST_SCRIPT" | sed 's/^/    /')
+EOF
+
+echo -e "${GREEN}✓ SDK test script ConfigMap created${NC}"
+echo ""
+
+echo -e "${YELLOW}[4.2] Creating SDK test Job...${NC}"
+cat <<EOF | kubectl_cmd apply -f - > /dev/null
+apiVersion: batch/v1
+kind: Job
+metadata:
+  name: sdk-test-job
+  namespace: $NAMESPACE
+spec:
+  ttlSecondsAfterFinished: 300
+  backoffLimit: 0
+  template:
+    spec:
+      restartPolicy: Never
+      containers:
+      - name: sdk-test
+        image: ghcr.io/astral-sh/uv:python3.12-bookworm
+        command:
+        - /bin/bash
+        - -c
+        - |
+          set -e
+          echo "Installing opensandbox SDK..."
+          uv pip install --system opensandbox
+          echo ""
+          echo "Running SDK tests..."
+          python /test/test.py
+        volumeMounts:
+        - name: test-script
+          mountPath: /test
+      volumes:
+      - name: test-script
+        configMap:
+          name: sdk-test-script
+EOF
+
+echo -e "${GREEN}✓ SDK test Job created${NC}"
+echo ""
+
+echo -e "${YELLOW}[4.3] Waiting for SDK test Job to complete (timeout: 5 minutes)...${NC}"
+# Wait for job to start
+sleep 5
+
+# Get pod name
+SDK_TEST_POD=""
+for i in $(seq 1 30); do
+    SDK_TEST_POD=$(kubectl_cmd get pods -n "$NAMESPACE" -l job-name=sdk-test-job -o jsonpath='{.items[0].metadata.name}' 2>/dev/null || echo "")
+    if [ -n "$SDK_TEST_POD" ]; then
+        echo "  SDK test pod: $SDK_TEST_POD"
+        break
+    fi
+    sleep 2
+done
+
+if [ -z "$SDK_TEST_POD" ]; then
+    echo -e "${RED}❌ SDK test pod not found${NC}"
+    kubectl_cmd get jobs -n "$NAMESPACE" sdk-test-job
+    exit 1
+fi
+
+# Stream logs and wait for completion
+echo ""
+echo "  Streaming test logs:"
+echo "  ---"
+kubectl_cmd logs -f "$SDK_TEST_POD" -n "$NAMESPACE" 2>/dev/null || true
+echo "  ---"
+echo ""
+
+# Wait a bit for Job to update status after pod completion
+echo "  Waiting for Job status to update..."
+sleep 5
+
+# Wait for job to complete (up to 60 seconds)
+for i in $(seq 1 60); do
+    JOB_STATUS=$(kubectl_cmd get job sdk-test-job -n "$NAMESPACE" -o jsonpath='{.status.conditions[?(@.type=="Complete")].status}' 2>/dev/null || echo "")
+    JOB_FAILED=$(kubectl_cmd get job sdk-test-job -n "$NAMESPACE" -o jsonpath='{.status.conditions[?(@.type=="Failed")].status}' 2>/dev/null || echo "")
+
+    if [ "$JOB_STATUS" = "True" ] || [ "$JOB_FAILED" = "True" ]; then
+        break
+    fi
+    sleep 1
+done
+
+# Check job status
+JOB_STATUS=$(kubectl_cmd get job sdk-test-job -n "$NAMESPACE" -o jsonpath='{.status.conditions[?(@.type=="Complete")].status}' 2>/dev/null || echo "")
+JOB_FAILED=$(kubectl_cmd get job sdk-test-job -n "$NAMESPACE" -o jsonpath='{.status.conditions[?(@.type=="Failed")].status}' 2>/dev/null || echo "")
+
+if [ "$JOB_STATUS" = "True" ]; then
+    echo -e "${GREEN}✅ Stage 4 Complete: SDK Integration Verified${NC}"
+elif [ "$JOB_FAILED" = "True" ]; then
+    echo -e "${RED}❌ Stage 4 Failed: SDK Integration Failed${NC}"
+    echo ""
+    echo -e "${YELLOW}Diagnostic Information:${NC}"
+    kubectl_cmd describe job sdk-test-job -n "$NAMESPACE" | tail -20
+    exit 1
+else
+    echo -e "${RED}❌ Stage 4 Failed: SDK test job did not complete${NC}"
+    kubectl_cmd get job sdk-test-job -n "$NAMESPACE"
+    exit 1
+fi
+echo ""
+
+# ==========================================
+# Stage 5: Helm Uninstall
+# ==========================================
+echo -e "${GREEN}==========================================${NC}"
+echo -e "${GREEN}Stage 5: Helm Uninstall${NC}"
+echo -e "${GREEN}==========================================${NC}"
+echo ""
+
+echo -e "${YELLOW}[5.1] Cleaning up test resources...${NC}"
+kubectl_cmd delete configmap sdk-test-script -n "$NAMESPACE" --ignore-not-found=true > /dev/null 2>&1
+kubectl_cmd delete job sdk-test-job -n "$NAMESPACE" --ignore-not-found=true > /dev/null 2>&1
+echo -e "${GREEN}✓ Test resources cleaned up${NC}"
+echo ""
+
+echo -e "${YELLOW}[5.2] Uninstalling Helm release...${NC}"
+helm_cmd uninstall "$RELEASE_NAME" -n "$NAMESPACE"
+echo -e "${GREEN}✓ Helm release uninstalled${NC}"
+echo ""
+
+echo "Waiting for resource cleanup..."
+sleep 10
+
+echo -e "${YELLOW}[5.3] Verifying resources are cleaned up...${NC}"
+REMAINING_PODS=$(kubectl_cmd get pods -n "$NAMESPACE" --no-headers 2>/dev/null | wc -l)
+REMAINING_POOLS=$(kubectl_cmd get pools -n "$NAMESPACE" --no-headers 2>/dev/null | wc -l)
+echo "  Remaining Pods: ${REMAINING_PODS}"
+echo "  Remaining Pools: ${REMAINING_POOLS}"
+
+if [ "$REMAINING_PODS" -eq 0 ] && [ "$REMAINING_POOLS" -eq 0 ]; then
+    echo -e "${GREEN}✓ All resources cleaned up${NC}"
+else
+    echo -e "${YELLOW}⚠️  Resources still remaining (Terminating)${NC}"
+    if [ "$REMAINING_PODS" -gt 0 ]; then
+        kubectl_cmd get pods -n "$NAMESPACE" 2>/dev/null || true
+    fi
+fi
+echo ""
+
+echo -e "${GREEN}✅ Stage 5 Complete: Helm Uninstall Successful${NC}"
+echo ""
+
+# ==========================================
+# Test Summary
+# ==========================================
+echo -e "${GREEN}==========================================${NC}"
+echo -e "${GREEN}End-to-End Test Complete!${NC}"
+echo -e "${GREEN}==========================================${NC}"
+echo ""
+echo "Test Results:"
+echo -e "  ${GREEN}✅ Stage 1: Helm Install - Success${NC}"
+echo -e "  ${GREEN}✅ Stage 2: Server Deployment Verification - Success${NC}"
+echo -e "  ${GREEN}✅ Stage 3: Pool Deployment Verification - Success${NC}"
+echo -e "  ${GREEN}✅ Stage 4: SDK Integration Verification - Success${NC}"
+echo -e "  ${GREEN}✅ Stage 5: Helm Uninstall - Success${NC}"
+echo ""
+echo -e "${GREEN}🎉 All tests passed!${NC}"
+echo ""

deploy/helm/opensandbox/scripts/install.sh

+#!/bin/bash
+# OpenSandbox Controller Deployment Script
+
+set -e
+
+# Check if sudo is required
+USE_SUDO=false
+if ! kubectl get nodes &> /dev/null 2>&1; then
+    if sudo kubectl get nodes &> /dev/null 2>&1; then
+        echo "Detected that sudo permissions are required, will use sudo to execute commands"
+        USE_SUDO=true
+    else
+        echo "Error: Unable to access Kubernetes cluster"
+        exit 1
+    fi
+fi
+
+# Define command functions
+kubectl_cmd() {
+    if [ "$USE_SUDO" = true ]; then
+        sudo kubectl "$@"
+    else
+        kubectl "$@"
+    fi
+}
+
+helm_cmd() {
+    if [ "$USE_SUDO" = true ]; then
+        sudo helm "$@"
+    else
+        helm "$@"
+    fi
+}
+
+# Color output
+GREEN='\033[0;32m'
+YELLOW='\033[1;33m'
+RED='\033[0;31m'
+NC='\033[0m' # No Color
+
+echo -e "${GREEN}======================================${NC}"
+echo -e "${GREEN}OpenSandbox Controller Helm Deployment${NC}"
+echo -e "${GREEN}======================================${NC}"
+echo ""
+
+# Check dependencies
+echo -e "${YELLOW}[1/6] Checking dependencies...${NC}"
+if ! command -v helm &> /dev/null; then
+    echo -e "${RED}Error: helm command not found${NC}"
+    echo "Please install Helm 3.0+: https://helm.sh/docs/intro/install/"
+    exit 1
+fi
+
+if ! command -v kubectl &> /dev/null; then
+    echo -e "${RED}Error: kubectl command not found${NC}"
+    echo "Please install kubectl: https://kubernetes.io/docs/tasks/tools/"
+    exit 1
+fi
+
+echo -e "${GREEN}✓ Helm version: $(helm version --short)${NC}"
+echo -e "${GREEN}✓ Kubectl version: $(kubectl version --client --short 2>/dev/null || kubectl version --client)${NC}"
+echo ""
+
+# Check cluster connection
+echo -e "${YELLOW}[2/6] Checking Kubernetes cluster connection...${NC}"
+if ! kubectl_cmd cluster-info &> /dev/null; then
+    echo -e "${RED}Error: Unable to connect to Kubernetes cluster${NC}"
+    echo "Please check your ~/.kube/config configuration"
+    exit 1
+fi
+echo -e "${GREEN}✓ Cluster connection successful${NC}"
+kubectl_cmd cluster-info | head -2
+echo ""
+
+# Configuration parameters
+# Get the parent directory of the script directory (chart root directory)
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+CHART_DIR="$(dirname "$SCRIPT_DIR")"
+RELEASE_NAME="opensandbox-controller"
+NAMESPACE="opensandbox"
+
+echo -e "${YELLOW}[3/6] Configuration parameters${NC}"
+echo "Chart directory: $CHART_DIR"
+echo "Release name: $RELEASE_NAME"
+echo "Namespace: $NAMESPACE"
+if [ -n "$IMAGE_REPO" ] || [ -n "$IMAGE_TAG" ]; then
+    echo "Image override: ${IMAGE_REPO:-<from values>}:${IMAGE_TAG:-<from values>}"
+else
+    echo "Image configuration: Using configuration from values file"
+fi
+echo ""
+
+# Select deployment environment
+echo -e "${YELLOW}[4/6] Select deployment environment${NC}"
+echo "1) Default configuration (values.yaml)"
+echo "2) End-to-end testing (values-e2e.yaml)"
+echo "3) Custom (custom values)"
+read -p "Please select [1-3]: " env_choice
+
+case $env_choice in
+    1)
+        VALUES_FILE="$CHART_DIR/values.yaml"
+        echo -e "${GREEN}✓ Using default configuration${NC}"
+        ;;
+    2)
+        VALUES_FILE="$CHART_DIR/values-e2e.yaml"
+        echo -e "${GREEN}✓ E2E test configuration selected${NC}"
+        ;;
+    3)
+        read -p "Please enter values file path: " custom_values
+        VALUES_FILE="$custom_values"
+        echo -e "${GREEN}✓ Using custom configuration: $VALUES_FILE${NC}"
+        ;;
+    *)
+        echo -e "${RED}Invalid selection, using default configuration${NC}"
+        VALUES_FILE="$CHART_DIR/values.yaml"
+        ;;
+esac
+echo ""
+
+# Validate Chart
+echo -e "${YELLOW}[5/6] Validating Helm Chart...${NC}"
+if ! helm lint "$CHART_DIR" &> /dev/null; then
+    echo -e "${RED}Error: Chart validation failed${NC}"
+    helm lint "$CHART_DIR"
+    exit 1
+fi
+echo -e "${GREEN}✓ Chart validation passed${NC}"
+echo ""
+
+# Confirm deployment
+echo -e "${YELLOW}[6/6] Preparing for deployment${NC}"
+echo "The following operations will be performed:"
+echo "  - Create namespace: $NAMESPACE"
+echo "  - Install CRDs: BatchSandbox, Pool"
+echo "  - Deploy Controller Manager"
+echo "  - Deploy Server (FastAPI control plane)"
+echo "  - Deploy RBAC resources"
+echo "  - Deploy Metrics service"
+echo "  - Create default Pool (agent-pool)"
+echo ""
+read -p "Confirm deployment? [y/N]: " confirm
+
+if [[ ! $confirm =~ ^[Yy]$ ]]; then
+    echo -e "${YELLOW}Deployment cancelled${NC}"
+    exit 0
+fi
+
+# Execute deployment
+echo ""
+echo -e "${GREEN}Starting deployment...${NC}"
+echo ""
+
+# If environment variables are set, override image configuration in values file
+EXTRA_ARGS=""
+if [ -n "$IMAGE_REPO" ]; then
+    EXTRA_ARGS="$EXTRA_ARGS --set controllerManager.image.repository=$IMAGE_REPO"
+fi
+if [ -n "$IMAGE_TAG" ]; then
+    EXTRA_ARGS="$EXTRA_ARGS --set controllerManager.image.tag=$IMAGE_TAG"
+fi
+if [ -n "$SERVER_IMAGE_REPO" ]; then
+    EXTRA_ARGS="$EXTRA_ARGS --set server.image.repository=$SERVER_IMAGE_REPO"
+fi
+if [ -n "$SERVER_IMAGE_TAG" ]; then
+    EXTRA_ARGS="$EXTRA_ARGS --set server.image.tag=$SERVER_IMAGE_TAG"
+fi
+
+helm_cmd upgrade --install "$RELEASE_NAME" "$CHART_DIR" \
+    --namespace "$NAMESPACE" \
+    --create-namespace \
+    -f "$VALUES_FILE" \
+    $EXTRA_ARGS \
+    --wait \
+    --timeout 5m
+
+echo ""
+echo -e "${GREEN}======================================${NC}"
+echo -e "${GREEN}✓ Deployment completed!${NC}"
+echo -e "${GREEN}======================================${NC}"
+echo ""
+
+# Display deployment information
+echo -e "${YELLOW}Deployment information:${NC}"
+helm_cmd status "$RELEASE_NAME" -n "$NAMESPACE"
+
+echo ""
+echo -e "${YELLOW}Verify deployment:${NC}"
+echo "1. Check Pod status:"
+echo "   $([ "$USE_SUDO" = true ] && echo "sudo ")kubectl get pods -n $NAMESPACE"
+echo ""
+echo "2. View Controller logs:"
+echo "   $([ "$USE_SUDO" = true ] && echo "sudo ")kubectl logs -n $NAMESPACE -l control-plane=controller-manager -f"
+echo ""
+echo "3. View Server logs:"
+echo "   $([ "$USE_SUDO" = true ] && echo "sudo ")kubectl logs -n $NAMESPACE -l app.kubernetes.io/component=server -f"
+echo ""
+echo "4. Access Server API (Port Forward):"
+echo "   $([ "$USE_SUDO" = true ] && echo "sudo ")kubectl port-forward -n $NAMESPACE svc/$RELEASE_NAME-server 8080:8080"
+echo "   curl http://localhost:8080/health"
+echo ""
+echo "5. View CRDs:"
+echo "   $([ "$USE_SUDO" = true ] && echo "sudo ")kubectl get crds | grep sandbox.opensandbox.io"
+echo ""
+echo "6. Check Pool status:"
+echo "   $([ "$USE_SUDO" = true ] && echo "sudo ")kubectl get pools -n $NAMESPACE"
+echo ""
+
+echo -e "${GREEN}Thank you for using OpenSandbox Controller!${NC}"

deploy/helm/opensandbox/scripts/uninstall.sh

+#!/bin/bash
+# OpenSandbox Controller Uninstall Script
+
+set -e
+
+# Check if sudo is required
+USE_SUDO=false
+if ! kubectl get nodes &> /dev/null 2>&1; then
+    if sudo kubectl get nodes &> /dev/null 2>&1; then
+        echo "Detected sudo privileges required, will use sudo to execute commands"
+        USE_SUDO=true
+    else
+        echo "Error: Unable to access Kubernetes cluster"
+        exit 1
+    fi
+fi
+
+# Define command functions
+kubectl_cmd() {
+    if [ "$USE_SUDO" = true ]; then
+        sudo kubectl "$@"
+    else
+        kubectl "$@"
+    fi
+}
+
+helm_cmd() {
+    if [ "$USE_SUDO" = true ]; then
+        sudo helm "$@"
+    else
+        helm "$@"
+    fi
+}
+
+GREEN='\033[0;32m'
+YELLOW='\033[1;33m'
+RED='\033[0;31m'
+NC='\033[0m'
+
+echo -e "${YELLOW}======================================${NC}"
+echo -e "${YELLOW}OpenSandbox Controller Uninstall${NC}"
+echo -e "${YELLOW}======================================${NC}"
+echo ""
+
+RELEASE_NAME="${RELEASE_NAME:-opensandbox-controller}"
+NAMESPACE="${NAMESPACE:-opensandbox}"
+
+# Check if already installed
+if ! helm_cmd list -n "$NAMESPACE" | grep -q "$RELEASE_NAME"; then
+    echo -e "${RED}Release not found: $RELEASE_NAME${NC}"
+    echo "Currently installed releases:"
+    helm_cmd list -A
+    exit 1
+fi
+
+echo "About to uninstall:"
+echo "  Release: $RELEASE_NAME"
+echo "  Namespace: $NAMESPACE"
+echo ""
+
+# Show current resources
+echo -e "${YELLOW}Current resources:${NC}"
+echo "Controller:"
+kubectl_cmd get deployment -n "$NAMESPACE" -l control-plane=controller-manager 2>/dev/null || echo "  Controller not found"
+echo ""
+echo "Server:"
+kubectl_cmd get deployment -n "$NAMESPACE" -l app.kubernetes.io/component=server 2>/dev/null || echo "  Server not found"
+echo ""
+
+# Check if there are running resources
+echo -e "${YELLOW}Checking custom resources...${NC}"
+BATCHSANDBOXES=$(kubectl_cmd get batchsandboxes -A --no-headers 2>/dev/null | wc -l)
+POOLS=$(kubectl_cmd get pools -A --no-headers 2>/dev/null | wc -l)
+
+if [ "$BATCHSANDBOXES" -gt 0 ] || [ "$POOLS" -gt 0 ]; then
+    echo -e "${RED}Warning: Running resources detected!${NC}"
+    echo "  BatchSandboxes: $BATCHSANDBOXES"
+    echo "  Pools: $POOLS"
+    echo ""
+
+    # Show Pool details
+    if [ "$POOLS" -gt 0 ]; then
+        echo "Pool details:"
+        kubectl_cmd get pools -A
+        echo ""
+    fi
+
+    echo "Recommended to delete these resources first:"
+    echo "  $([ "$USE_SUDO" = true ] && echo "sudo ")kubectl delete batchsandboxes --all -A"
+    echo "  $([ "$USE_SUDO" = true ] && echo "sudo ")kubectl delete pools --all -A"
+    echo ""
+    read -p "Continue with uninstall? [y/N]: " force_continue
+    if [[ ! $force_continue =~ ^[Yy]$ ]]; then
+        echo -e "${YELLOW}Uninstall cancelled${NC}"
+        exit 0
+    fi
+fi
+
+# Confirm uninstall
+read -p "Confirm uninstall of $RELEASE_NAME (including Controller and Server)? [y/N]: " confirm
+if [[ ! $confirm =~ ^[Yy]$ ]]; then
+    echo -e "${YELLOW}Uninstall cancelled${NC}"
+    exit 0
+fi
+
+echo ""
+echo -e "${GREEN}Starting uninstall...${NC}"
+
+# Uninstall Helm release
+helm_cmd uninstall "$RELEASE_NAME" -n "$NAMESPACE"
+
+echo -e "${GREEN}✓ Helm release uninstalled${NC}"
+echo ""
+
+# Wait for Pod termination
+echo "Waiting for Pods to terminate..."
+sleep 5
+
+# Ask whether to delete CRDs
+read -p "Delete CRDs? (This will delete all BatchSandbox and Pool resources) [y/N]: " delete_crds
+if [[ $delete_crds =~ ^[Yy]$ ]]; then
+    echo "Deleting CRDs..."
+    kubectl_cmd delete crd batchsandboxes.sandbox.opensandbox.io 2>/dev/null || echo "  CRD batchsandboxes does not exist"
+    kubectl_cmd delete crd pools.sandbox.opensandbox.io 2>/dev/null || echo "  CRD pools does not exist"
+    echo -e "${GREEN}✓ CRDs deleted${NC}"
+else
+    echo -e "${YELLOW}⊗ CRDs retained${NC}"
+fi
+echo ""
+
+# Ask whether to delete namespace
+read -p "Delete namespace $NAMESPACE? [y/N]: " delete_ns
+if [[ $delete_ns =~ ^[Yy]$ ]]; then
+    echo "Deleting namespace..."
+    kubectl_cmd delete namespace "$NAMESPACE" 2>/dev/null || echo "  Namespace does not exist"
+    echo -e "${GREEN}✓ Namespace deleted${NC}"
+else
+    echo -e "${YELLOW}⊗ Namespace retained${NC}"
+fi
+
+echo ""
+echo -e "${GREEN}======================================${NC}"
+echo -e "${GREEN}✓ Uninstall completed${NC}"
+echo -e "${GREEN}======================================${NC}"
+echo ""
+
+# Verify uninstall
+echo -e "${YELLOW}Verifying uninstall:${NC}"
+echo "Checking Helm releases:"
+helm_cmd list -n "$NAMESPACE" 2>/dev/null || echo "  Namespace does not exist"
+echo ""
+echo "Checking CRDs:"
+kubectl_cmd get crds | grep sandbox.opensandbox.io || echo "  No OpenSandbox CRDs found"
+echo ""

deploy/helm/opensandbox/templates/NOTES.txt

+Thank you for installing {{ .Chart.Name }}!
+
+Your release is named {{ .Release.Name }}.
+
+To learn more about the release, try:
+
+  $ helm status {{ .Release.Name }}
+  $ helm get all {{ .Release.Name }}
+
+===============================================================================
+
+OpenSandbox Kubernetes Controller has been deployed to namespace: {{ include "opensandbox-controller.namespace" . }}
+
+Note: Both the controller and user resources (Pool, BatchSandbox) use the same namespace.
+
+1. Check the controller status:
+
+   kubectl get deployment -n {{ include "opensandbox-controller.namespace" . }} {{ .Values.namePrefix }}controller-manager
+
+2. View controller logs:
+
+   kubectl logs -n {{ include "opensandbox-controller.namespace" . }} -l control-plane=controller-manager -f
+
+3. Verify CRDs are installed:
+
+   kubectl get crds | grep sandbox.opensandbox.io
+
+   You should see:
+   - batchsandboxes.sandbox.opensandbox.io
+   - pools.sandbox.opensandbox.io
+
+4. Pool Resources:
+{{- if .Values.pools }}
+{{- $enabledPools := list }}
+{{- range .Values.pools }}
+{{- if .enabled }}
+{{- $enabledPools = append $enabledPools .name }}
+{{- end }}
+{{- end }}
+{{- if $enabledPools }}
+
+   ✅ The following Pools have been deployed:
+{{- range $enabledPools }}
+   - {{ . }}
+{{- end }}
+
+   Check Pool status:
+   kubectl get pools -n {{ include "opensandbox-controller.namespace" . }}
+
+   Check Pool details:
+   kubectl describe pool agent-pool -n {{ include "opensandbox-controller.namespace" . }}
+
+   Monitor Pool capacity:
+   kubectl get pool agent-pool -n {{ include "opensandbox-controller.namespace" . }} -o jsonpath='{.status}'
+{{- else }}
+
+   ⚠️  All Pools are disabled in values.yaml.
+   To enable the default agent-pool:
+   helm upgrade {{ .Release.Name }} opensandbox-controller --set pools[0].enabled=true
+{{- end }}
+{{- else }}
+
+   No Pools configured.
+{{- end }}
+
+   To disable Pool auto-deployment:
+   helm upgrade {{ .Release.Name }} opensandbox-controller --set pools[0].enabled=false
+
+   Create additional Pool manually:
+
+   # Basic Pool without task executor
+   cat <<EOF | kubectl apply -f -
+   apiVersion: sandbox.opensandbox.io/v1alpha1
+   kind: Pool
+   metadata:
+     name: my-first-pool
+   spec:
+     minBufferSize: 2
+     maxBufferSize: 5
+     capacity: 10
+     sandboxTemplate:
+       spec:
+         image: ubuntu:latest
+         command: ["sleep", "infinity"]
+   EOF
+
+   # Pool with task executor (for task execution features)
+   # Note: Requires task-executor image
+   cat <<EOF | kubectl apply -f -
+   apiVersion: sandbox.opensandbox.io/v1alpha1
+   kind: Pool
+   metadata:
+     name: task-enabled-pool
+   spec:
+     template:
+       spec:
+         shareProcessNamespace: true
+         containers:
+         - name: sandbox-container
+           image: ubuntu:latest
+           command: ["sleep", "infinity"]
+         - name: task-executor
+           image: {{ .Values.taskExecutor.image.repository }}:{{ .Values.taskExecutor.image.tag }}
+           securityContext:
+             capabilities:
+               add: ["SYS_PTRACE"]
+     capacitySpec:
+       bufferMax: 5
+       bufferMin: 2
+       poolMax: 10
+       poolMin: 2
+   EOF
+
+5. Create your first BatchSandbox:
+
+   cat <<EOF | kubectl apply -f -
+   apiVersion: sandbox.opensandbox.io/v1alpha1
+   kind: BatchSandbox
+   metadata:
+     name: my-first-batchsandbox
+   spec:
+     replicas: 3
+     ttlSecondsAfterFinished: 3600
+     sandboxTemplate:
+       spec:
+         image: ubuntu:latest
+         command: ["sleep", "infinity"]
+   EOF
+
+6. Check your resources:
+
+   kubectl get pools
+   kubectl get batchsandboxes
+   kubectl get pods
+
+{{- if .Values.metrics.enabled }}
+
+7. Access metrics (requires port-forward):
+
+   kubectl port-forward -n {{ include "opensandbox-controller.namespace" . }} svc/{{ .Values.namePrefix }}controller-manager-metrics-service {{ .Values.metrics.service.port }}:{{ .Values.metrics.service.port }}
+
+   Then visit: https://localhost:{{ .Values.metrics.service.port }}/metrics
+{{- end }}
+
+===============================================================================
+
+For more information:
+- Documentation: https://github.com/alibaba/OpenSandbox
+- Issues: https://github.com/alibaba/OpenSandbox/issues
+
+{{ if .Values.extraRoles.batchsandboxEditor.enabled }}
+Note: Additional ClusterRoles for user management have been created:
+- {{ .Values.namePrefix }}batchsandbox-editor-role
+- {{ .Values.namePrefix }}batchsandbox-viewer-role
+- {{ .Values.namePrefix }}pool-editor-role
+- {{ .Values.namePrefix }}pool-viewer-role
+
+You can bind these roles to users/groups as needed.
+{{- end }}

deploy/helm/opensandbox/templates/_helpers.tpl

+{{/*
+扩展chart名称
+*/}}
+{{- define "opensandbox-controller.name" -}}
+{{- default .Chart.Name .Values.nameOverride | trunc 63 | trimSuffix "-" }}
+{{- end }}
+
+{{/*
+创建完整的限定名称
+*/}}
+{{- define "opensandbox-controller.fullname" -}}
+{{- if .Values.fullnameOverride }}
+{{- .Values.fullnameOverride | trunc 63 | trimSuffix "-" }}
+{{- else }}
+{{- $name := default .Chart.Name .Values.nameOverride }}
+{{- if contains $name .Release.Name }}
+{{- .Release.Name | trunc 63 | trimSuffix "-" }}
+{{- else }}
+{{- printf "%s-%s" .Release.Name $name | trunc 63 | trimSuffix "-" }}
+{{- end }}
+{{- end }}
+{{- end }}
+
+{{/*
+Chart标签
+*/}}
+{{- define "opensandbox-controller.chart" -}}
+{{- printf "%s-%s" .Chart.Name .Chart.Version | replace "+" "_" | trunc 63 | trimSuffix "-" }}
+{{- end }}
+
+{{/*
+通用标签
+*/}}
+{{- define "opensandbox-controller.labels" -}}
+helm.sh/chart: {{ include "opensandbox-controller.chart" . }}
+{{ include "opensandbox-controller.selectorLabels" . }}
+{{- if .Chart.AppVersion }}
+app.kubernetes.io/version: {{ .Chart.AppVersion | quote }}
+{{- end }}
+app.kubernetes.io/managed-by: {{ .Release.Service }}
+{{- with .Values.labels }}
+{{ toYaml . }}
+{{- end }}
+{{- end }}
+
+{{/*
+选择器标签
+*/}}
+{{- define "opensandbox-controller.selectorLabels" -}}
+app.kubernetes.io/name: {{ include "opensandbox-controller.name" . }}
+app.kubernetes.io/instance: {{ .Release.Name }}
+control-plane: controller-manager
+{{- end }}
+
+{{/*
+创建ServiceAccount名称
+*/}}
+{{- define "opensandbox-controller.serviceAccountName" -}}
+{{- if .Values.rbac.serviceAccount.create }}
+{{- default (printf "%scontroller-manager" .Values.namePrefix) .Values.rbac.serviceAccount.name }}
+{{- else }}
+{{- default "default" .Values.rbac.serviceAccount.name }}
+{{- end }}
+{{- end }}
+
+{{/*
+命名空间
+*/}}
+{{- define "opensandbox-controller.namespace" -}}
+{{- if .Values.namespaceOverride }}
+{{- .Values.namespaceOverride }}
+{{- else }}
+{{- printf "%ssystem" .Values.namePrefix }}
+{{- end }}
+{{- end }}
+
+{{/*
+Controller镜像
+*/}}
+{{- define "opensandbox-controller.controllerImage" -}}
+{{- printf "%s:%s" .Values.controllerManager.image.repository (.Values.controllerManager.image.tag | default .Chart.AppVersion) }}
+{{- end }}
+
+{{/*
+Task Executor镜像
+*/}}
+{{- define "opensandbox-controller.taskExecutorImage" -}}
+{{- printf "%s:%s" .Values.taskExecutor.image.repository (.Values.taskExecutor.image.tag | default .Chart.AppVersion) }}
+{{- end }}

deploy/helm/opensandbox/templates/clusterrole.yaml

+{{- if .Values.rbac.create }}
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRole
+metadata:
+  name: {{ .Values.namePrefix }}manager-role
+  labels:
+    {{- include "opensandbox-controller.labels" . | nindent 4 }}
+rules:
+- apiGroups:
+  - ""
+  resources:
+  - events
+  - pods
+  verbs:
+  - create
+  - delete
+  - get
+  - list
+  - patch
+  - update
+  - watch
+- apiGroups:
+  - ""
+  resources:
+  - pods/status
+  verbs:
+  - get
+  - patch
+  - update
+- apiGroups:
+  - sandbox.opensandbox.io
+  resources:
+  - batchsandboxes
+  - pools
+  verbs:
+  - create
+  - delete
+  - get
+  - list
+  - patch
+  - update
+  - watch
+- apiGroups:
+  - sandbox.opensandbox.io
+  resources:
+  - batchsandboxes/finalizers
+  - pools/finalizers
+  verbs:
+  - update
+- apiGroups:
+  - sandbox.opensandbox.io
+  resources:
+  - batchsandboxes/status
+  - pools/status
+  verbs:
+  - get
+  - patch
+  - update
+{{- end }}

deploy/helm/opensandbox/templates/clusterrolebinding.yaml

+{{- if .Values.rbac.create }}
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRoleBinding
+metadata:
+  name: {{ .Values.namePrefix }}manager-rolebinding
+  labels:
+    {{- include "opensandbox-controller.labels" . | nindent 4 }}
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: ClusterRole
+  name: {{ .Values.namePrefix }}manager-role
+subjects:
+- kind: ServiceAccount
+  name: {{ include "opensandbox-controller.serviceAccountName" . }}
+  namespace: {{ include "opensandbox-controller.namespace" . }}
+{{- end }}

deploy/helm/opensandbox/templates/deployment.yaml

+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: {{ .Values.namePrefix }}controller-manager
+  namespace: {{ include "opensandbox-controller.namespace" . }}
+  labels:
+    {{- include "opensandbox-controller.labels" . | nindent 4 }}
+spec:
+  replicas: {{ .Values.controllerManager.replicas }}
+  selector:
+    matchLabels:
+      {{- include "opensandbox-controller.selectorLabels" . | nindent 6 }}
+  template:
+    metadata:
+      annotations:
+        {{- with .Values.podAnnotations }}
+        {{- toYaml . | nindent 8 }}
+        {{- end }}
+      labels:
+        {{- include "opensandbox-controller.selectorLabels" . | nindent 8 }}
+        {{- with .Values.podLabels }}
+        {{- toYaml . | nindent 8 }}
+        {{- end }}
+    spec:
+      {{- with .Values.imagePullSecrets }}
+      imagePullSecrets:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      serviceAccountName: {{ include "opensandbox-controller.serviceAccountName" . }}
+      securityContext:
+        {{- toYaml .Values.controllerManager.securityContext | nindent 8 }}
+      terminationGracePeriodSeconds: {{ .Values.terminationGracePeriodSeconds }}
+      containers:
+      - name: manager
+        image: {{ include "opensandbox-controller.controllerImage" . }}
+        imagePullPolicy: {{ .Values.controllerManager.image.pullPolicy }}
+        command:
+        - /workspace/server
+        args:
+        {{- if .Values.controllerManager.leaderElect }}
+        - --leader-elect
+        {{- end }}
+        - --health-probe-bind-address={{ .Values.controllerManager.healthProbeBindAddress }}
+        - --zap-log-level={{ .Values.controllerManager.logLevel }}
+        {{- with .Values.controllerManager.extraArgs }}
+        {{- toYaml . | nindent 8 }}
+        {{- end }}
+        securityContext:
+          {{- toYaml .Values.controllerManager.containerSecurityContext | nindent 10 }}
+        {{- with .Values.controllerManager.env }}
+        env:
+          {{- toYaml . | nindent 10 }}
+        {{- end }}
+        livenessProbe:
+          httpGet:
+            path: /healthz
+            port: {{ .Values.controllerManager.healthProbePort | default 8081 }}
+          initialDelaySeconds: {{ .Values.healthProbes.liveness.initialDelaySeconds | default 15 }}
+          periodSeconds: {{ .Values.healthProbes.liveness.periodSeconds | default 20 }}
+          timeoutSeconds: {{ .Values.healthProbes.liveness.timeoutSeconds | default 1 }}
+          failureThreshold: {{ .Values.healthProbes.liveness.failureThreshold | default 3 }}
+        readinessProbe:
+          httpGet:
+            path: /readyz
+            port: {{ .Values.controllerManager.healthProbePort | default 8081 }}
+          initialDelaySeconds: {{ .Values.healthProbes.readiness.initialDelaySeconds | default 5 }}
+          periodSeconds: {{ .Values.healthProbes.readiness.periodSeconds | default 10 }}
+          timeoutSeconds: {{ .Values.healthProbes.readiness.timeoutSeconds | default 1 }}
+          failureThreshold: {{ .Values.healthProbes.readiness.failureThreshold | default 3 }}
+        resources:
+          {{- toYaml .Values.controllerManager.resources | nindent 10 }}
+      {{- with .Values.controllerManager.nodeSelector }}
+      nodeSelector:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.controllerManager.affinity }}
+      affinity:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.controllerManager.tolerations }}
+      tolerations:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}

deploy/helm/opensandbox/templates/extra-roles.yaml

+{{- if .Values.rbac.create }}
+{{- if .Values.extraRoles.batchsandboxEditor.enabled }}
+---
+# This role is provided to allow the cluster admin to help manage permissions for users.
+# Grants permissions to create, update, and delete BatchSandbox resources.
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRole
+metadata:
+  name: {{ .Values.namePrefix }}batchsandbox-editor-role
+  labels:
+    {{- include "opensandbox-controller.labels" . | nindent 4 }}
+rules:
+- apiGroups:
+  - sandbox.opensandbox.io
+  resources:
+  - batchsandboxes
+  verbs:
+  - create
+  - delete
+  - get
+  - list
+  - patch
+  - update
+  - watch
+- apiGroups:
+  - sandbox.opensandbox.io
+  resources:
+  - batchsandboxes/status
+  verbs:
+  - get
+{{- end }}
+
+{{- if .Values.extraRoles.batchsandboxViewer.enabled }}
+---
+# This role is provided to allow the cluster admin to help manage permissions for users.
+# Grants read-only permissions for BatchSandbox resources.
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRole
+metadata:
+  name: {{ .Values.namePrefix }}batchsandbox-viewer-role
+  labels:
+    {{- include "opensandbox-controller.labels" . | nindent 4 }}
+rules:
+- apiGroups:
+  - sandbox.opensandbox.io
+  resources:
+  - batchsandboxes
+  verbs:
+  - get
+  - list
+  - watch
+- apiGroups:
+  - sandbox.opensandbox.io
+  resources:
+  - batchsandboxes/status
+  verbs:
+  - get
+{{- end }}
+
+{{- if .Values.extraRoles.poolEditor.enabled }}
+---
+# This role is provided to allow the cluster admin to help manage permissions for users.
+# Grants permissions to create, update, and delete Pool resources.
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRole
+metadata:
+  name: {{ .Values.namePrefix }}pool-editor-role
+  labels:
+    {{- include "opensandbox-controller.labels" . | nindent 4 }}
+rules:
+- apiGroups:
+  - sandbox.opensandbox.io
+  resources:
+  - pools
+  verbs:
+  - create
+  - delete
+  - get
+  - list
+  - patch
+  - update
+  - watch
+- apiGroups:
+  - sandbox.opensandbox.io
+  resources:
+  - pools/status
+  verbs:
+  - get
+{{- end }}
+
+{{- if .Values.extraRoles.poolViewer.enabled }}
+---
+# This role is provided to allow the cluster admin to help manage permissions for users.
+# Grants read-only permissions for Pool resources.
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRole
+metadata:
+  name: {{ .Values.namePrefix }}pool-viewer-role
+  labels:
+    {{- include "opensandbox-controller.labels" . | nindent 4 }}
+rules:
+- apiGroups:
+  - sandbox.opensandbox.io
+  resources:
+  - pools
+  verbs:
+  - get
+  - list
+  - watch
+- apiGroups:
+  - sandbox.opensandbox.io
+  resources:
+  - pools/status
+  verbs:
+  - get
+{{- end }}
+{{- end }}

deploy/helm/opensandbox/templates/leader-election-role.yaml

+{{- if .Values.rbac.create }}
+# permissions to do leader election.
+apiVersion: rbac.authorization.k8s.io/v1
+kind: Role
+metadata:
+  name: {{ .Values.namePrefix }}leader-election-role
+  namespace: {{ include "opensandbox-controller.namespace" . }}
+  labels:
+    {{- include "opensandbox-controller.labels" . | nindent 4 }}
+rules:
+- apiGroups:
+  - ""
+  resources:
+  - configmaps
+  verbs:
+  - get
+  - list
+  - watch
+  - create
+  - update
+  - patch
+  - delete
+- apiGroups:
+  - coordination.k8s.io
+  resources:
+  - leases
+  verbs:
+  - get
+  - list
+  - watch
+  - create
+  - update
+  - patch
+  - delete
+- apiGroups:
+  - ""
+  resources:
+  - events
+  verbs:
+  - create
+  - patch
+{{- end }}

deploy/helm/opensandbox/templates/leader-election-rolebinding.yaml

+{{- if .Values.rbac.create }}
+apiVersion: rbac.authorization.k8s.io/v1
+kind: RoleBinding
+metadata:
+  name: {{ .Values.namePrefix }}leader-election-rolebinding
+  namespace: {{ include "opensandbox-controller.namespace" . }}
+  labels:
+    {{- include "opensandbox-controller.labels" . | nindent 4 }}
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: Role
+  name: {{ .Values.namePrefix }}leader-election-role
+subjects:
+- kind: ServiceAccount
+  name: {{ include "opensandbox-controller.serviceAccountName" . }}
+  namespace: {{ include "opensandbox-controller.namespace" . }}
+{{- end }}

deploy/helm/opensandbox/templates/metrics-rbac.yaml

+{{- if and .Values.rbac.create .Values.metrics.enabled }}
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRole
+metadata:
+  name: {{ .Values.namePrefix }}metrics-auth-role
+  labels:
+    {{- include "opensandbox-controller.labels" . | nindent 4 }}
+rules:
+- apiGroups:
+  - authentication.k8s.io
+  resources:
+  - tokenreviews
+  verbs:
+  - create
+- apiGroups:
+  - authorization.k8s.io
+  resources:
+  - subjectaccessreviews
+  verbs:
+  - create
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRoleBinding
+metadata:
+  name: {{ .Values.namePrefix }}metrics-auth-rolebinding
+  labels:
+    {{- include "opensandbox-controller.labels" . | nindent 4 }}
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: ClusterRole
+  name: {{ .Values.namePrefix }}metrics-auth-role
+subjects:
+- kind: ServiceAccount
+  name: {{ include "opensandbox-controller.serviceAccountName" . }}
+  namespace: {{ include "opensandbox-controller.namespace" . }}
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRole
+metadata:
+  name: {{ .Values.namePrefix }}metrics-reader
+  labels:
+    {{- include "opensandbox-controller.labels" . | nindent 4 }}
+rules:
+- nonResourceURLs:
+  - /metrics
+  verbs:
+  - get
+{{- end }}

deploy/helm/opensandbox/templates/metrics-service.yaml

+{{- if .Values.metrics.enabled }}
+apiVersion: v1
+kind: Service
+metadata:
+  name: {{ .Values.namePrefix }}controller-manager-metrics-service
+  namespace: {{ include "opensandbox-controller.namespace" . }}
+  labels:
+    {{- include "opensandbox-controller.labels" . | nindent 4 }}
+    control-plane: controller-manager
+spec:
+  type: {{ .Values.metrics.service.type }}
+  ports:
+  - name: https
+    port: {{ .Values.metrics.service.port }}
+    protocol: TCP
+    targetPort: {{ .Values.metrics.service.port }}
+  selector:
+    {{- include "opensandbox-controller.selectorLabels" . | nindent 4 }}
+{{- end }}

deploy/helm/opensandbox/templates/poddisruptionbudget.yaml

+{{- if .Values.podDisruptionBudget.enabled }}
+apiVersion: policy/v1
+kind: PodDisruptionBudget
+metadata:
+  name: {{ include "opensandbox-controller.fullname" . }}-controller
+  namespace: {{ include "opensandbox-controller.namespace" . }}
+  labels:
+    {{- include "opensandbox-controller.labels" . | nindent 4 }}
+spec:
+  {{- if .Values.podDisruptionBudget.minAvailable }}
+  minAvailable: {{ .Values.podDisruptionBudget.minAvailable }}
+  {{- end }}
+  {{- if .Values.podDisruptionBudget.maxUnavailable }}
+  maxUnavailable: {{ .Values.podDisruptionBudget.maxUnavailable }}
+  {{- end }}
+  selector:
+    matchLabels:
+      control-plane: controller-manager
+      {{- include "opensandbox-controller.selectorLabels" . | nindent 6 }}
+{{- end }}

deploy/helm/opensandbox/templates/pools.yaml

+{{- if .Values.pools }}
+{{- range .Values.pools }}
+{{- if .enabled }}
+---
+apiVersion: sandbox.opensandbox.io/v1alpha1
+kind: Pool
+metadata:
+  name: {{ .name }}
+  namespace: {{ include "opensandbox-controller.namespace" $ }}
+  labels:
+    {{- include "opensandbox-controller.labels" $ | nindent 4 }}
+    {{- with .labels }}
+    {{- toYaml . | nindent 4 }}
+    {{- end }}
+  {{- with .annotations }}
+  annotations:
+    {{- toYaml . | nindent 4 }}
+  {{- end }}
+spec:
+  {{- if .template }}
+  template:
+    {{- with .template.metadata }}
+    metadata:
+      {{- toYaml . | nindent 6 }}
+    {{- end }}
+    spec:
+      {{- if .template.spec.shareProcessNamespace }}
+      shareProcessNamespace: {{ .template.spec.shareProcessNamespace }}
+      {{- end }}
+
+      {{- if .template.spec.initContainers }}
+      initContainers:
+      {{- range .template.spec.initContainers }}
+      - name: {{ .name }}
+        image: {{ .image }}
+        {{- if .imagePullPolicy }}
+        imagePullPolicy: {{ .imagePullPolicy }}
+        {{- end }}
+        {{- if .command }}
+        command:
+          {{- toYaml .command | nindent 10 }}
+        {{- end }}
+        {{- if .args }}
+        args:
+          {{- toYaml .args | nindent 10 }}
+        {{- end }}
+        {{- if .env }}
+        env:
+          {{- toYaml .env | nindent 10 }}
+        {{- end }}
+        {{- if .volumeMounts }}
+        volumeMounts:
+          {{- toYaml .volumeMounts | nindent 10 }}
+        {{- end }}
+        {{- if .resources }}
+        resources:
+          {{- toYaml .resources | nindent 10 }}
+        {{- end }}
+        {{- if .securityContext }}
+        securityContext:
+          {{- toYaml .securityContext | nindent 10 }}
+        {{- end }}
+      {{- end }}
+      {{- end }}
+
+      containers:
+      {{- range .template.spec.containers }}
+      - name: {{ .name }}
+        image: {{ .image }}
+        {{- if .imagePullPolicy }}
+        imagePullPolicy: {{ .imagePullPolicy }}
+        {{- end }}
+        {{- if .command }}
+        command:
+          {{- toYaml .command | nindent 10 }}
+        {{- end }}
+        {{- if .args }}
+        args:
+          {{- toYaml .args | nindent 10 }}
+        {{- end }}
+        {{- if .env }}
+        env:
+          {{- toYaml .env | nindent 10 }}
+        {{- end }}
+        {{- if .ports }}
+        ports:
+          {{- toYaml .ports | nindent 10 }}
+        {{- end }}
+        {{- if .volumeMounts }}
+        volumeMounts:
+          {{- toYaml .volumeMounts | nindent 10 }}
+        {{- end }}
+        {{- if .resources }}
+        resources:
+          {{- toYaml .resources | nindent 10 }}
+        {{- end }}
+        {{- if .securityContext }}
+        securityContext:
+          {{- toYaml .securityContext | nindent 10 }}
+        {{- end }}
+        {{- if .livenessProbe }}
+        livenessProbe:
+          {{- toYaml .livenessProbe | nindent 10 }}
+        {{- end }}
+        {{- if .readinessProbe }}
+        readinessProbe:
+          {{- toYaml .readinessProbe | nindent 10 }}
+        {{- end }}
+      {{- end }}
+
+      {{- if .template.spec.volumes }}
+      volumes:
+        {{- toYaml .template.spec.volumes | nindent 6 }}
+      {{- end }}
+
+      {{- if .template.spec.nodeSelector }}
+      nodeSelector:
+        {{- toYaml .template.spec.nodeSelector | nindent 8 }}
+      {{- end }}
+
+      {{- if .template.spec.tolerations }}
+      tolerations:
+        {{- toYaml .template.spec.tolerations | nindent 8 }}
+      {{- end }}
+
+      {{- if .template.spec.affinity }}
+      affinity:
+        {{- toYaml .template.spec.affinity | nindent 8 }}
+      {{- end }}
+  {{- end }}
+
+  capacitySpec:
+    {{- if .capacitySpec.bufferMin }}
+    bufferMin: {{ .capacitySpec.bufferMin }}
+    {{- end }}
+    {{- if .capacitySpec.bufferMax }}
+    bufferMax: {{ .capacitySpec.bufferMax }}
+    {{- end }}
+    {{- if .capacitySpec.poolMin }}
+    poolMin: {{ .capacitySpec.poolMin }}
+    {{- end }}
+    {{- if .capacitySpec.poolMax }}
+    poolMax: {{ .capacitySpec.poolMax }}
+    {{- end }}
+{{- end }}
+{{- end }}
+{{- end }}

deploy/helm/opensandbox/templates/server-configmap.yaml

+{{- if .Values.server.enabled }}
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: {{ include "opensandbox-controller.fullname" . }}-server-config
+  namespace: {{ include "opensandbox-controller.namespace" . }}
+  labels:
+    {{- include "opensandbox-controller.labels" . | nindent 4 }}
+    app.kubernetes.io/component: server
+data:
+  config.toml: |
+    # OpenSandbox Server Configuration
+    # Generated by Helm Chart
+
+    [server]
+    host = "{{ .Values.server.config.server.host }}"
+    port = {{ .Values.server.config.server.port }}
+    log_level = "{{ .Values.server.config.server.logLevel }}"
+    {{- if .Values.server.config.server.apiKey }}
+    api_key = "{{ .Values.server.config.server.apiKey }}"
+    {{- end }}
+
+    [runtime]
+    type = "{{ .Values.server.config.runtime.type }}"
+    execd_image = "{{ .Values.server.config.runtime.execdImage }}"
+
+    [kubernetes]
+    # Use in-cluster configuration (empty kubeconfig_path)
+    kubeconfig_path = ""
+    namespace = "{{ include "opensandbox-controller.namespace" . }}"
+    # 优先使用 Pool 模式（batchsandbox provider + Pool）
+    workload_provider = "{{ .Values.server.config.kubernetes.workloadProvider }}"
+    # BatchSandbox 模板文件（非 Pool 模式需要）
+    batchsandbox_template_file = "/etc/opensandbox/batchsandbox-template.yaml"
+
+  # BatchSandbox 模板（用于非 Pool 模式）
+  batchsandbox-template.yaml: |
+    # BatchSandbox CR template for non-pooled mode
+    # Pool mode does not use this template
+    metadata:
+      # Metadata will be merged with runtime-generated values
+
+    spec:
+      replicas: 1
+      template:
+        spec:
+          restartPolicy: Never
+          {{- if .Values.server.config.kubernetes.batchsandboxTemplate.tolerations }}
+          tolerations:
+            {{- toYaml .Values.server.config.kubernetes.batchsandboxTemplate.tolerations | nindent 12 }}
+          {{- else }}
+          tolerations:
+            - operator: "Exists"
+          {{- end }}
+          {{- if .Values.server.config.kubernetes.batchsandboxTemplate.nodeSelector }}
+          nodeSelector:
+            {{- toYaml .Values.server.config.kubernetes.batchsandboxTemplate.nodeSelector | nindent 12 }}
+          {{- end }}
+{{- end }}

deploy/helm/opensandbox/templates/server-deployment.yaml

+{{- if .Values.server.enabled }}
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: {{ include "opensandbox-controller.fullname" . }}-server
+  namespace: {{ include "opensandbox-controller.namespace" . }}
+  labels:
+    {{- include "opensandbox-controller.labels" . | nindent 4 }}
+    app.kubernetes.io/component: server
+spec:
+  replicas: {{ .Values.server.replicas }}
+  selector:
+    matchLabels:
+      {{- include "opensandbox-controller.selectorLabels" . | nindent 6 }}
+      app.kubernetes.io/component: server
+  template:
+    metadata:
+      annotations:
+        checksum/config: {{ include (print $.Template.BasePath "/server-configmap.yaml") . | sha256sum }}
+        {{- with .Values.server.podAnnotations }}
+        {{- toYaml . | nindent 8 }}
+        {{- end }}
+      labels:
+        {{- include "opensandbox-controller.selectorLabels" . | nindent 8 }}
+        app.kubernetes.io/component: server
+        {{- with .Values.server.podLabels }}
+        {{- toYaml . | nindent 8 }}
+        {{- end }}
+    spec:
+      {{- with .Values.imagePullSecrets }}
+      imagePullSecrets:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      serviceAccountName: {{ include "opensandbox-controller.serviceAccountName" . }}
+      securityContext:
+        {{- toYaml .Values.server.securityContext | nindent 8 }}
+      terminationGracePeriodSeconds: {{ .Values.server.terminationGracePeriodSeconds | default 10 }}
+      containers:
+      - name: server
+        image: "{{ .Values.server.image.repository }}:{{ .Values.server.image.tag }}"
+        imagePullPolicy: {{ .Values.server.image.pullPolicy }}
+        args:
+        - --config
+        - /etc/opensandbox/config.toml
+        {{- if .Values.server.reload }}
+        - --reload
+        {{- end }}
+        {{- with .Values.server.extraArgs }}
+        {{- toYaml . | nindent 8 }}
+        {{- end }}
+        ports:
+        - name: http
+          containerPort: {{ .Values.server.config.server.port }}
+          protocol: TCP
+        {{- with .Values.server.env }}
+        env:
+          {{- toYaml . | nindent 10 }}
+        {{- end }}
+        livenessProbe:
+          httpGet:
+            path: {{ .Values.server.healthProbes.liveness.path }}
+            port: http
+          initialDelaySeconds: {{ .Values.server.healthProbes.liveness.initialDelaySeconds }}
+          periodSeconds: {{ .Values.server.healthProbes.liveness.periodSeconds }}
+          timeoutSeconds: {{ .Values.server.healthProbes.liveness.timeoutSeconds }}
+          failureThreshold: {{ .Values.server.healthProbes.liveness.failureThreshold }}
+        readinessProbe:
+          httpGet:
+            path: {{ .Values.server.healthProbes.readiness.path }}
+            port: http
+          initialDelaySeconds: {{ .Values.server.healthProbes.readiness.initialDelaySeconds }}
+          periodSeconds: {{ .Values.server.healthProbes.readiness.periodSeconds }}
+          timeoutSeconds: {{ .Values.server.healthProbes.readiness.timeoutSeconds }}
+          failureThreshold: {{ .Values.server.healthProbes.readiness.failureThreshold }}
+        resources:
+          {{- toYaml .Values.server.resources | nindent 10 }}
+        volumeMounts:
+        - name: config
+          mountPath: /etc/opensandbox
+          readOnly: true
+        securityContext:
+          {{- toYaml .Values.server.containerSecurityContext | nindent 10 }}
+      volumes:
+      - name: config
+        configMap:
+          name: {{ include "opensandbox-controller.fullname" . }}-server-config
+      {{- with .Values.server.nodeSelector }}
+      nodeSelector:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.server.affinity }}
+      affinity:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.server.tolerations }}
+      tolerations:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+{{- end }}

deploy/helm/opensandbox/templates/server-ingress.yaml

+{{- if and .Values.server.enabled .Values.server.ingress.enabled }}
+apiVersion: networking.k8s.io/v1
+kind: Ingress
+metadata:
+  name: {{ include "opensandbox-controller.fullname" . }}-server
+  namespace: {{ include "opensandbox-controller.namespace" . }}
+  labels:
+    {{- include "opensandbox-controller.labels" . | nindent 4 }}
+    app.kubernetes.io/component: server
+  {{- with .Values.server.ingress.annotations }}
+  annotations:
+    {{- toYaml . | nindent 4 }}
+  {{- end }}
+spec:
+  {{- if .Values.server.ingress.className }}
+  ingressClassName: {{ .Values.server.ingress.className }}
+  {{- end }}
+  {{- if .Values.server.ingress.tls }}
+  tls:
+    {{- range .Values.server.ingress.tls }}
+    - hosts:
+        {{- range .hosts }}
+        - {{ . | quote }}
+        {{- end }}
+      secretName: {{ .secretName }}
+    {{- end }}
+  {{- end }}
+  rules:
+    {{- range .Values.server.ingress.hosts }}
+    - host: {{ .host | quote }}
+      http:
+        paths:
+          {{- range .paths }}
+          - path: {{ .path }}
+            pathType: {{ .pathType }}
+            backend:
+              service:
+                name: {{ include "opensandbox-controller.fullname" $ }}-server
+                port:
+                  number: {{ $.Values.server.service.port }}
+          {{- end }}
+    {{- end }}
+{{- end }}

deploy/helm/opensandbox/templates/server-service.yaml

+{{- if .Values.server.enabled }}
+apiVersion: v1
+kind: Service
+metadata:
+  name: {{ include "opensandbox-controller.fullname" . }}-server
+  namespace: {{ include "opensandbox-controller.namespace" . }}
+  labels:
+    {{- include "opensandbox-controller.labels" . | nindent 4 }}
+    app.kubernetes.io/component: server
+  {{- with .Values.server.service.annotations }}
+  annotations:
+    {{- toYaml . | nindent 4 }}
+  {{- end }}
+spec:
+  type: {{ .Values.server.service.type }}
+  {{- if and (eq .Values.server.service.type "NodePort") .Values.server.service.nodePort }}
+  ports:
+  - port: {{ .Values.server.service.port }}
+    targetPort: http
+    protocol: TCP
+    name: http
+    nodePort: {{ .Values.server.service.nodePort }}
+  {{- else }}
+  ports:
+  - port: {{ .Values.server.service.port }}
+    targetPort: http
+    protocol: TCP
+    name: http
+  {{- end }}
+  selector:
+    {{- include "opensandbox-controller.selectorLabels" . | nindent 4 }}
+    app.kubernetes.io/component: server
+{{- end }}

deploy/helm/opensandbox/templates/serviceaccount.yaml

+{{- if .Values.rbac.serviceAccount.create -}}
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: {{ include "opensandbox-controller.serviceAccountName" . }}
+  namespace: {{ include "opensandbox-controller.namespace" . }}
+  labels:
+    {{- include "opensandbox-controller.labels" . | nindent 4 }}
+  {{- with .Values.rbac.serviceAccount.annotations }}
+  annotations:
+    {{- toYaml . | nindent 4 }}
+  {{- end }}
+{{- end }}

deploy/helm/opensandbox/templates/servicemonitor.yaml

+{{- if and .Values.metrics.enabled .Values.metrics.serviceMonitor.enabled }}
+apiVersion: monitoring.coreos.com/v1
+kind: ServiceMonitor
+metadata:
+  name: {{ .Values.namePrefix }}controller-manager-metrics-monitor
+  namespace: {{ include "opensandbox-controller.namespace" . }}
+  labels:
+    {{- include "opensandbox-controller.labels" . | nindent 4 }}
+spec:
+  endpoints:
+  - path: /metrics
+    port: https
+    scheme: https
+    bearerTokenFile: /var/run/secrets/kubernetes.io/serviceaccount/token
+    tlsConfig:
+      insecureSkipVerify: true
+    interval: {{ .Values.metrics.serviceMonitor.interval }}
+    scrapeTimeout: {{ .Values.metrics.serviceMonitor.scrapeTimeout }}
+  selector:
+    matchLabels:
+      {{- include "opensandbox-controller.selectorLabels" . | nindent 6 }}
+      control-plane: controller-manager
+{{- end }}

deploy/helm/opensandbox/values-e2e.yaml

+# OpenSandbox Helm Chart - E2E测试配置
+# values-e2e.yaml
+#
+# 此配置文件专为端到端测试设计，使用最小资源配置
+
+# Controller配置
+controllerManager:
+  image:
+    repository: opensandbox/controller
+    tag: latest
+    pullPolicy: Never
+  replicas: 1
+  resources:
+    limits:
+      cpu: 1
+      memory: 1Gi
+    requests:
+      cpu: 200m
+      memory: 256Mi
+
+# Task Executor镜像配置
+taskExecutor:
+  image:
+    repository: opensandbox/task-executor
+    tag: latest
+    pullPolicy: Never
+
+# Server配置 - E2E测试
+server:
+  enabled: true
+  image:
+    repository: opensandbox/server
+    tag: latest
+    pullPolicy: Never
+  replicas: 1
+  resources:
+    limits:
+      cpu: 500m
+      memory: 512Mi
+    requests:
+      cpu: 100m
+      memory: 256Mi
+  # Server 安全上下文 - 镜像以 root 运行
+  securityContext: {}
+  containerSecurityContext:
+    allowPrivilegeEscalation: false
+    capabilities:
+      drop:
+      - "ALL"
+  # 使用 ClusterIP，通过 port-forward 访问
+  service:
+    type: ClusterIP
+    port: 8080
+
+# Pool配置 - 最小配置用于E2E测试
+pools:
+  - name: agent-pool
+    enabled: true
+    labels:
+      app: opensandbox
+      component: agent-pool
+    annotations:
+      description: "E2E Test Pool - Minimal Capacity"
+
+    template:
+      metadata:
+        labels:
+          pool: agent-pool
+          sdk-compatible: "true"
+      spec:
+        shareProcessNamespace: true
+        initContainers:
+        - name: execd-installer
+          image: opensandbox/execd:v1.0.5
+          imagePullPolicy: Never
+          command: ["/bin/sh", "-c"]
+          args:
+            - |
+              cp ./execd /opt/opensandbox/bin/execd && \
+              cp ./bootstrap.sh /opt/opensandbox/bin/bootstrap.sh && \
+              chmod +x /opt/opensandbox/bin/*
+          volumeMounts:
+          - name: opensandbox-bin
+            mountPath: /opt/opensandbox/bin
+        containers:
+        - name: sandbox-container
+          image: nginx:latest
+          imagePullPolicy: Never
+          command: ["/opt/opensandbox/bin/bootstrap.sh", "sleep", "infinity"]
+          env:
+          - name: EXECD
+            value: /opt/opensandbox/bin/execd
+          ports:
+          - containerPort: 44772
+            name: execd
+          resources:
+            requests:
+              cpu: "50m"
+              memory: "64Mi"
+            limits:
+              cpu: "200m"
+              memory: "128Mi"
+          volumeMounts:
+          - name: opensandbox-bin
+            mountPath: /opt/opensandbox/bin
+        - name: task-executor
+          image: opensandbox/task-executor:dev
+          imagePullPolicy: Never
+          resources:
+            requests:
+              cpu: "50m"
+              memory: "64Mi"
+            limits:
+              cpu: "200m"
+              memory: "128Mi"
+          securityContext:
+            capabilities:
+              add: ["SYS_PTRACE"]
+        volumes:
+        - name: opensandbox-bin
+          emptyDir: {}
+
+    # E2E测试容量配置：最小化以避免主机过载
+    capacitySpec:
+      bufferMin: 2      # 最小可用2个
+      bufferMax: 3      # 最多预热3个
+      poolMin: 2        # 最小总数2个
+      poolMax: 5        # 最大总数5个
+
+# Metrics监控
+metrics:
+  enabled: true
+  serviceMonitor:
+    enabled: false  # 禁用 ServiceMonitor

deploy/helm/opensandbox/values.yaml

+# Default values for opensandbox-controller.
+# This is a YAML-formatted file.
+# Declare variables to be passed into your templates.
+
+# Namespace configuration
+# Controller and user resources use this namespace uniformly
+namespaceOverride: "opensandbox"
+
+# Name override configuration (optional)
+# Used to customize the resource names generated by the Chart
+nameOverride: ""
+fullnameOverride: ""
+
+# Controller image configuration
+controllerManager:
+  image:
+    repository: opensandbox/controller
+    tag: "dev"
+    pullPolicy: Never
+
+  # Number of replicas
+  replicas: 1
+
+  # Resource limits
+  resources:
+    limits:
+      cpu: 500m
+      memory: 128Mi
+    requests:
+      cpu: 10m
+      memory: 64Mi
+
+  # Enable leader election (high availability)
+  leaderElect: true
+
+  # Health check address
+  healthProbeBindAddress: ":8081"
+
+  # Health check port (extracted from healthProbeBindAddress, or configured separately)
+  healthProbePort: 8081
+
+  # Log level
+  logLevel: 3
+
+  # Security context
+  securityContext:
+    runAsNonRoot: true
+    seccompProfile:
+      type: RuntimeDefault
+
+  containerSecurityContext:
+    allowPrivilegeEscalation: false
+    capabilities:
+      drop:
+      - "ALL"
+
+  # Node selector
+  nodeSelector: {}
+
+  # Tolerations
+  tolerations: []
+
+  # Affinity
+  affinity: {}
+
+  # Environment variables
+  env: []
+  # - name: EXAMPLE_ENV
+  #   value: "example-value"
+
+  # Additional container arguments
+  extraArgs: []
+  # - --some-arg=value
+
+# Task Executor image configuration
+# Note: task-executor is not an independently deployed service, but a sidecar container
+# injected into Pods created by Pool for executing tasks.
+# Only image information is configured here for users to use when creating Pool resources.
+taskExecutor:
+  image:
+    repository: opensandbox/task-executor
+    tag: "dev"
+    pullPolicy: Never
+
+  # Task Executor default resource configuration (for Pool template reference)
+  resources:
+    limits:
+      cpu: 500m
+      memory: 256Mi
+    requests:
+      cpu: 100m
+      memory: 128Mi
+
+  # Task Executor required security context
+  securityContext:
+    capabilities:
+      add: ["SYS_PTRACE"]  # Requires ptrace permission to inject processes
+
+# OpenSandbox Server configuration
+# Server is the FastAPI control plane that exposes REST API for SDK usage
+# SDK creates and manages Sandbox through Server
+server:
+  # Enable Server (required for SDK scenarios)
+  enabled: true
+
+  # Server image configuration
+  image:
+    repository: opensandbox/server
+    tag: "v0.1.0"
+    pullPolicy: Never
+
+  # Number of replicas
+  replicas: 1
+
+  # Resource limits
+  resources:
+    limits:
+      cpu: 1
+      memory: 512Mi
+    requests:
+      cpu: 100m
+      memory: 256Mi
+
+  # Security context
+  # ⚠️ Security Warning: Server image currently runs as root user (uid=0)
+  # This is a known security concern and does not comply with Kubernetes best practices
+  # Production recommendations:
+  #   1. Use NetworkPolicy to restrict Server's network access
+  #   2. Use PodSecurityPolicy/PodSecurityStandard to restrict Pod permissions
+  #   3. Regularly audit Server logs and behavior
+  #
+  # TODO: Update Server Dockerfile to support non-root user (Issue: #xxx)
+  # Planned to be fixed in next version
+  securityContext: {}  # Must be empty to allow root user execution
+
+  containerSecurityContext:
+    allowPrivilegeEscalation: false
+    capabilities:
+      drop:
+      - "ALL"
+
+  # Node selector
+  nodeSelector: {}
+
+  # Tolerations
+  tolerations: []
+
+  # Affinity
+  affinity: {}
+
+  # Environment variables
+  env: []
+
+  # Additional container arguments
+  extraArgs: []
+
+  # Development mode (enable hot reload)
+  reload: false
+
+  # Pod annotations
+  podAnnotations: {}
+
+  # Pod labels
+  podLabels: {}
+
+  # Termination grace period (seconds)
+  terminationGracePeriodSeconds: 10
+
+  # Server configuration (written to ConfigMap)
+  config:
+    server:
+      host: "0.0.0.0"
+      port: 8080
+      logLevel: "INFO"
+      # API Key authentication (optional)
+      apiKey: ""
+
+    runtime:
+      type: "kubernetes"
+      execdImage: "opensandbox/execd:v1.0.5"
+
+    kubernetes:
+      # workloadProvider: batchsandbox (use Pool) or agent-sandbox
+      workloadProvider: "batchsandbox"
+      # BatchSandbox template configuration (for non-Pool mode)
+      batchsandboxTemplate:
+        # BatchSandbox labels (optional)
+        labels: {}
+        # Pod labels (optional)
+        podLabels: {}
+        # Node tolerations
+        tolerations:
+          - operator: "Exists"
+        # Node selector (optional)
+        nodeSelector: {}
+
+  # Service configuration
+  service:
+    # Service type: ClusterIP, NodePort, LoadBalancer
+    type: ClusterIP
+    port: 8080
+    # NodePort (effective when type=NodePort)
+    # Set to specific port number (30000-32767) or leave empty for automatic allocation
+    # Example: 30080
+    nodePort: null
+    annotations: {}
+
+  # Ingress configuration (optional)
+  ingress:
+    enabled: false
+    className: ""
+    annotations: {}
+      # kubernetes.io/ingress.class: nginx
+      # cert-manager.io/cluster-issuer: letsencrypt-prod
+    hosts:
+      - host: opensandbox.example.com
+        paths:
+          - path: /
+            pathType: Prefix
+    tls: []
+    #  - secretName: opensandbox-tls
+    #    hosts:
+    #      - opensandbox.example.com
+
+  # Health probe configuration
+  healthProbes:
+    liveness:
+      path: /health
+      initialDelaySeconds: 15
+      periodSeconds: 20
+      timeoutSeconds: 5
+      failureThreshold: 3
+    readiness:
+      path: /health
+      initialDelaySeconds: 5
+      periodSeconds: 10
+      timeoutSeconds: 5
+      failureThreshold: 3
+
+# Metrics configuration
+metrics:
+  enabled: true
+  service:
+    type: ClusterIP
+    port: 8443
+
+  # Enable ServiceMonitor (requires Prometheus Operator)
+  serviceMonitor:
+    enabled: false
+    interval: 30s
+    scrapeTimeout: 10s
+
+# RBAC configuration
+rbac:
+  # Create RBAC resources
+  create: true
+
+  # ServiceAccount configuration
+  serviceAccount:
+    # Create ServiceAccount
+    create: true
+    # Use existing ServiceAccount
+    name: ""
+    # ServiceAccount annotations
+    annotations: {}
+
+# CRD configuration
+crds:
+  # Whether to install CRD (set to true for first installation, can be set to false for upgrades)
+  install: true
+  # CRD annotations
+  annotations: {}
+
+# Image pull secrets
+imagePullSecrets: []
+# - name: regcred
+
+# Global labels
+labels: {}
+
+# Global annotations
+annotations: {}
+
+# Name prefix, used to generate resource names and namespace (namePrefix + "system")
+namePrefix: "opensandbox-"
+
+# Pod annotations
+podAnnotations:
+  kubectl.kubernetes.io/default-container: manager
+
+# Pod labels
+podLabels: {}
+
+# Termination grace period (seconds)
+terminationGracePeriodSeconds: 10
+
+# Pod Disruption Budget (optional)
+# Used for high availability deployment to prevent too many Pods from being unavailable simultaneously
+podDisruptionBudget:
+  enabled: false
+  minAvailable: 1
+  # maxUnavailable: 1  # Mutually exclusive with minAvailable, choose one
+
+# Health probe configuration
+healthProbes:
+  liveness:
+    initialDelaySeconds: 15
+    periodSeconds: 20
+    timeoutSeconds: 1
+    failureThreshold: 3
+  readiness:
+    initialDelaySeconds: 5
+    periodSeconds: 10
+    timeoutSeconds: 1
+    failureThreshold: 3
+
+# Extra user roles (optional)
+extraRoles:
+  # Create BatchSandbox admin role
+  batchsandboxAdmin:
+    enabled: true
+  # Create BatchSandbox editor role
+  batchsandboxEditor:
+    enabled: true
+  # Create BatchSandbox viewer role
+  batchsandboxViewer:
+    enabled: true
+  # Create Pool admin role
+  poolAdmin:
+    enabled: true
+  # Create Pool editor role
+  poolEditor:
+    enabled: true
+  # Create Pool viewer role
+  poolViewer:
+    enabled: true
+
+# Pool resource configuration
+# Automatically create Pool resources during Helm deployment
+# Default: Enable a general Agent Pool suitable for most SDK scenarios
+pools:
+  # Default Agent Pool - SDK scenario (with execd + task-executor)
+  - name: agent-pool
+    enabled: true
+    labels:
+      app: opensandbox
+      component: agent-pool
+    annotations:
+      description: "Default Agent Pool with execd and task-executor for SDK usage"
+
+    template:
+      metadata:
+        labels:
+          pool: agent-pool
+          sdk-compatible: "true"
+      spec:
+        # Required: Share process namespace (required by task-executor)
+        shareProcessNamespace: true
+
+        # Init Container: Install execd (required for SDK scenario)
+        initContainers:
+        - name: execd-installer
+          image: opensandbox/execd:v1.0.5
+          imagePullPolicy: Never
+          command: ["/bin/sh", "-c"]
+          args:
+            - |
+              cp ./execd /opt/opensandbox/bin/execd && \
+              cp ./bootstrap.sh /opt/opensandbox/bin/bootstrap.sh && \
+              chmod +x /opt/opensandbox/bin/execd && \
+              chmod +x /opt/opensandbox/bin/bootstrap.sh
+          volumeMounts:
+          - name: opensandbox-bin
+            mountPath: /opt/opensandbox/bin
+
+        containers:
+        # Main container: Sandbox environment
+        - name: sandbox-container
+          image: nginx:latest
+          imagePullPolicy: Never
+          command: ["/opt/opensandbox/bin/bootstrap.sh", "sleep", "infinity"]
+          env:
+          - name: EXECD
+            value: /opt/opensandbox/bin/execd
+          ports:
+          - containerPort: 44772
+            name: execd
+            protocol: TCP
+          resources:
+            requests:
+              cpu: "100m"
+              memory: "128Mi"
+            limits:
+              cpu: "500m"
+              memory: "256Mi"
+          volumeMounts:
+          - name: opensandbox-bin
+            mountPath: /opt/opensandbox/bin
+
+        # Sidecar: Task Executor
+        - name: task-executor
+          image: opensandbox/task-executor:dev
+          imagePullPolicy: Never
+          ports:
+          - containerPort: 5758
+            name: task-executor
+            protocol: TCP
+          resources:
+            requests:
+              cpu: "100m"
+              memory: "128Mi"
+            limits:
+              cpu: "500m"
+              memory: "256Mi"
+          securityContext:
+            capabilities:
+              add: ["SYS_PTRACE"]
+
+        volumes:
+        - name: opensandbox-bin
+          emptyDir: {}
+
+    # Capacity configuration: Suitable for small to medium scale deployments (10-50 concurrent Agents)
+    capacitySpec:
+      bufferMin: 10    # Minimum available Pods
+      bufferMax: 20    # Maximum pre-warmed Pods
+      poolMin: 10      # Minimum total Pods
+      poolMax: 100     # Maximum total Pods
+
+# Notes:
+# - If Pool is not needed, set pools[0].enabled=false
+# - If custom configuration is needed, override or add more Pools in values
+# - Capacity configuration can be adjusted according to actual concurrency requirements:
+#   | Concurrent Agents | bufferMin | bufferMax | poolMin | poolMax |
+#   |------------------|-----------|-----------|---------|---------|
+#   | 1-10             | 2         | 5         | 2       | 20      |
+#   | 10-50            | 10        | 20        | 10      | 100     |
+#   | 50-200           | 50        | 100       | 50      | 300     |
+#   | 200+             | 100       | 200       | 100     | 500     |

deploy/templates/docker-compose.prod.copy.yml

+# 用于部署的 docker-compose 文件:
+# - FastGPT 端口映射为 3000:3000
+# - FastGPT-mcp-server 端口映射 3005:3000
+# - 建议修改账密后再运行
+
+# plugin auth token
+x-plugin-auth-token: &x-plugin-auth-token 'token'
+# aiproxy token
+x-aiproxy-token: &x-aiproxy-token 'token'
+# 数据库连接相关配置
+x-share-db-config: &x-share-db-config
+  MONGODB_URI: mongodb://myusername:mypassword@mongo:27017/fastgpt?authSource=admin
+  DB_MAX_LINK: 100
+  REDIS_URL: redis://default:mypassword@redis:6379
+  # @see https://doc.fastgpt.cn/docs/self-host/config/object-storage
+  STORAGE_VENDOR: minio # minio | aws-s3 | cos | oss
+  STORAGE_REGION: us-east-1
+  STORAGE_ACCESS_KEY_ID: minioadmin
+  STORAGE_SECRET_ACCESS_KEY: minioadmin
+  STORAGE_PUBLIC_BUCKET: fastgpt-public
+  STORAGE_PRIVATE_BUCKET: fastgpt-private
+  STORAGE_EXTERNAL_ENDPOINT: http://192.168.0.2:9000 # 一个服务器和客户端均可访问到存储桶的地址，可以是固定的宿主机 IP 或者域名，注意不要填写成 127.0.0.1 或者 localhost 等本地回环地址(因为容器里无法使用)
+  STORAGE_S3_ENDPOINT: http://fastgpt-minio:9000 # 协议://域名(IP):端口
+  STORAGE_S3_FORCE_PATH_STYLE: true
+  STORAGE_S3_MAX_RETRIES: 3
+# Log 配置
+x-log-config: &x-log-config
+  LOG_ENABLE_CONSOLE: true
+  LOG_CONSOLE_LEVEL: debug
+  LOG_ENABLE_OTEL: false
+  LOG_OTEL_LEVEL: info
+  LOG_OTEL_URL: http://localhost:4318/v1/logs
+
+# 向量库相关配置
+x-vec-config: &x-vec-config
+${{vec.config}}
+
+version: '3.3'
+services:
+  # Vector DB
+${{vec.db}}
+  mongo:
+    image: ${{mongo.image}}:${{mongo.tag}} # cpu 不支持 AVX 时候使用 4.4.29
+    container_name: mongo
+    restart: always
+    networks:
+      - fastgpt
+    command: mongod --keyFile /data/mongodb.key --replSet rs0
+    environment:
+      - MONGO_INITDB_ROOT_USERNAME=myusername
+      - MONGO_INITDB_ROOT_PASSWORD=mypassword
+    volumes:
+      - ./mongo/data:/data/db
+    healthcheck:
+      test: ['CMD', 'mongo', '-u', 'myusername', '-p', 'mypassword', '--authenticationDatabase', 'admin', '--eval', "db.adminCommand('ping')"]
+      interval: 10s
+      timeout: 5s
+      retries: 5
+      start_period: 30s
+    entrypoint:
+      - bash
+      - -c
+      - |
+        openssl rand -base64 128 > /data/mongodb.key
+        chmod 400 /data/mongodb.key
+        chown 999:999 /data/mongodb.key
+        echo 'const isInited = rs.status().ok === 1
+        if(!isInited){
+          rs.initiate({
+              _id: "rs0",
+              members: [
+                  { _id: 0, host: "mongo:27017" }
+              ]
+          })
+        }' > /data/initReplicaSet.js
+        # 启动MongoDB服务
+        exec docker-entrypoint.sh "$$@" &
+
+        # 等待MongoDB服务启动
+        until mongo -u myusername -p mypassword --authenticationDatabase admin --eval "print('waited for connection')"; do
+          echo "Waiting for MongoDB to start..."
+          sleep 2
+        done
+
+        # 执行初始化副本集的脚本
+        mongo -u myusername -p mypassword --authenticationDatabase admin /data/initReplicaSet.js
+
+        # 等待docker-entrypoint.sh脚本执行的MongoDB服务进程
+        wait $$!
+  redis:
+    image: ${{redis.image}}:${{redis.tag}}
+    container_name: redis
+    networks:
+      - fastgpt
+    restart: always
+    command: |
+      redis-server --requirepass mypassword --loglevel warning --maxclients 10000 --appendonly yes --save 60 10 --maxmemory 4gb --maxmemory-policy noeviction
+    healthcheck:
+      test: ['CMD', 'redis-cli', '-a', 'mypassword', 'ping']
+      interval: 10s
+      timeout: 3s
+      retries: 3
+      start_period: 30s
+    volumes:
+      - ./redis/data:/data
+  fastgpt-minio:
+    image: ${{minio.image}}:${{minio.tag}}
+    container_name: fastgpt-minio
+    restart: always
+    ports:
+      - 9000:9000
+      - 9001:9001
+    networks:
+      - fastgpt
+    environment:
+      - MINIO_ROOT_USER=minioadmin
+      - MINIO_ROOT_PASSWORD=minioadmin
+    volumes:
+      - ./fastgpt-minio:/data
+    command: server /data --console-address ":9001"
+    healthcheck:
+      test: ['CMD', 'curl', '-f', 'http://localhost:9000/minio/health/live']
+      interval: 30s
+      timeout: 20s
+      retries: 3
+
+  opensandbox-server:
+    image: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/server:v0.1.7
+    container_name: opensandbox-server
+    restart: always
+    networks:
+      - fastgpt
+    extra_hosts:
+      - 'host.docker.internal:host-gateway' # Enable access to host machine
+    volumes:
+      - /var/run/docker.sock:/var/run/docker.sock
+    configs:
+      - source: opensandbox-config
+        target: /etc/opensandbox/config.toml
+    environment:
+      - SANDBOX_CONFIG_PATH=/etc/opensandbox/config.toml
+    healthcheck:
+      test: ['CMD', 'curl', '-f', 'http://localhost:8090/health']
+      interval: 10s
+      timeout: 5s
+      retries: 5
+
+  fastgpt:
+    container_name: fastgpt
+    image: ${{fastgpt.image}}:${{fastgpt.tag}} # git
+    ports:
+      - 3000:3000
+    networks:
+      - fastgpt
+    extra_hosts:
+      - 'host.docker.internal:host-gateway' # Enable access to host machine
+    depends_on:
+      - mongo
+      - sandbox
+      - vectorDB
+      - opensandbox-server
+    restart: always
+    environment:
+      <<: [*x-share-db-config, *x-vec-config, *x-log-config]
+      # 前端外部可访问的地址，用于自动补全文件资源路径。例如 https:fastgpt.cn，不能填 localhost。这个值可以不填，不填则发给模型的图片会是一个相对路径，而不是全路径，模型可能伪造Host。
+      FE_DOMAIN:
+      # root 密码，用户名为: root。如果需要修改 root 密码，直接修改这个环境变量，并重启即可。
+      DEFAULT_ROOT_PSW: 1234
+      # 登录凭证密钥
+      TOKEN_KEY: any
+      # root的密钥，常用于升级时候的初始化请求
+      ROOT_KEY: root_key
+      # 文件阅读加密
+      FILE_TOKEN_KEY: filetoken
+      # 密钥加密key
+      AES256_SECRET_KEY: fastgptkey
+
+      # plugin 地址
+      PLUGIN_BASE_URL: http://fastgpt-plugin:3000
+      PLUGIN_TOKEN: *x-plugin-auth-token
+      # sandbox 地址
+      CODE_SANDBOX_URL: http://sandbox:3000
+      # opensandbox server 地址
+      OPENSANDBOX_SERVER_URL: http://opensandbox-server:8090
+      # AI Proxy 的地址，如果配了该地址，优先使用
+      AIPROXY_API_ENDPOINT: http://aiproxy:3000
+      # AI Proxy 的 Admin Token，与 AI Proxy 中的环境变量 ADMIN_KEY
+      AIPROXY_API_TOKEN: *x-aiproxy-token
+
+      # 传递给 OTLP 收集器的服务名称
+      LOG_OTEL_SERVICE_NAME: fastgpt-client
+
+      # 工作流最大运行次数
+      WORKFLOW_MAX_RUN_TIMES: 1000
+      # 批量执行节点，最大输入长度
+      WORKFLOW_MAX_LOOP_TIMES: 100
+      # 对话文件过期天数
+      CHAT_FILE_EXPIRE_TIME: 7
+      # 服务器接收请求，最大大小，单位 MB
+      SERVICE_REQUEST_MAX_CONTENT_LENGTH: 10
+      # HTML 转换最大字符数
+      MAX_HTML_TRANSFORM_CHARS: 1000000
+    volumes:
+      - ./config.json:/app/data/config.json
+  sandbox:
+    container_name: sandbox
+    image: ${{fastgpt-sandbox.image}}:${{fastgpt-sandbox.tag}}
+    networks:
+      - fastgpt
+    restart: always
+    environment:
+      <<: [*x-log-config]
+      LOG_OTEL_SERVICE_NAME: fastgpt-code-sandbox
+    healthcheck:
+      test: ['CMD', 'curl', '-f', 'http://localhost:3000/health']
+      interval: 30s
+      timeout: 20s
+      retries: 3
+  fastgpt-mcp-server:
+    container_name: fastgpt-mcp-server
+    image: ${{fastgpt-mcp_server.image}}:${{fastgpt-mcp_server.tag}}
+    networks:
+      - fastgpt
+    ports:
+      - 3005:3000
+    restart: always
+    environment:
+      <<: [*x-log-config]
+      FASTGPT_ENDPOINT: http://fastgpt:3000
+  fastgpt-plugin:
+    image: ${{fastgpt-plugin.image}}:${{fastgpt-plugin.tag}}
+    container_name: fastgpt-plugin
+    restart: always
+    networks:
+      - fastgpt
+    environment:
+      <<: [*x-share-db-config, *x-log-config]
+      AUTH_TOKEN: *x-plugin-auth-token
+      # 工具网络请求，最大请求和响应体
+      SERVICE_REQUEST_MAX_CONTENT_LENGTH: 10
+      # 最大 API 请求体大小
+      MAX_API_SIZE: 10
+      # 传递给 OTLP 收集器的服务名称
+      LOG_OTEL_SERVICE_NAME: fastgpt-plugin
+    depends_on:
+      fastgpt-minio:
+        condition: service_healthy
+    healthcheck:
+      test: ['CMD', 'curl', '-f', 'http://localhost:3000/health']
+      interval: 30s
+      timeout: 20s
+      retries: 3
+  # AI Proxy
+  aiproxy:
+    image: ${{aiproxy.image}}:${{aiproxy.tag}}
+    container_name: aiproxy
+    restart: unless-stopped
+    depends_on:
+      aiproxy_pg:
+        condition: service_healthy
+    networks:
+      - fastgpt
+      - aiproxy
+    environment:
+      # 对应 fastgpt 里的AIPROXY_API_TOKEN
+      ADMIN_KEY: *x-aiproxy-token
+      # 错误日志详情保存时间（小时）
+      LOG_DETAIL_STORAGE_HOURS: 1
+      # 数据库连接地址
+      SQL_DSN: postgres://postgres:aiproxy@aiproxy_pg:5432/aiproxy
+      # 最大重试次数
+      RETRY_TIMES: 3
+      # 不需要计费
+      BILLING_ENABLED: false
+      # 不需要严格检测模型
+      DISABLE_MODEL_CONFIG: true
+    healthcheck:
+      test: ['CMD', 'curl', '-f', 'http://localhost:3000/api/status']
+      interval: 5s
+      timeout: 5s
+      retries: 10
+  aiproxy_pg:
+    image: ${{aiproxy-pg.image}}:${{aiproxy-pg.tag}} # docker hub
+    restart: unless-stopped
+    container_name: aiproxy_pg
+    volumes:
+      - ./aiproxy_pg:/var/lib/postgresql/data
+    networks:
+      - aiproxy
+    environment:
+      TZ: Asia/Shanghai
+      POSTGRES_USER: postgres
+      POSTGRES_DB: aiproxy
+      POSTGRES_PASSWORD: aiproxy
+    healthcheck:
+      test: ['CMD', 'pg_isready', '-U', 'postgres', '-d', 'aiproxy']
+      interval: 5s
+      timeout: 5s
+      retries: 10
+networks:
+  fastgpt:
+  aiproxy:
+  vector:
+${{vec.extra}}
+
+configs:
+  # opensandbox config
+  opensandbox-config:
+    content: |
+      [server]
+      host = "0.0.0.0"
+      port = 8090
+      log_level = "INFO"
+
+      [runtime]
+      type = "docker"
+      execd_image = "sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.7"
+
+      [egress]
+      image = "sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/egress:v1.0.1"
+
+      [docker]
+      network_mode = "bridge"
+      host_ip = "host.docker.internal"
+      drop_capabilities = ["AUDIT_WRITE", "MKNOD", "NET_ADMIN", "NET_RAW", "SYS_ADMIN", "SYS_MODULE", "SYS_PTRACE", "SYS_TIME", "SYS_TTY_CONFIG"]
+      no_new_privileges = true
+      pids_limit = 512
+
+      [ingress]
+      mode = "direct"