# DevOS Deployment & Infrastructure Runtime > Version: 1.0 > Status: Architecture Specification > Scope: Deployment / Infrastructure / Runtime Environment / Server / Container / CI/CD / Health / Rollback > Depends On: 01–33 > Project: `/home/dev/bun/devos` --- # 1. 文档目标 本文定义 DevOS Deployment & Infrastructure Runtime。 核心目标： > 让 DevOS 能够统一管理项目从“构建”到“部署”再到“运行、监控、升级、回滚”的完整生命周期。 最终链路： ```text Project ↓ Build ↓ Artifact ↓ Deployment Plan ↓ Environment ↓ Infrastructure Provider ↓ Deploy ↓ Runtime Instance ↓ Health ↓ Observability ↓ Upgrade / Rollback ``` --- # 2. Deployment Runtime 定位 Deployment Runtime 不是： - Docker CLI 的包装器 - SSH 脚本集合 - Kubernetes Dashboard - 云厂商 SDK 集合 - CI/CD 脚本目录 它是： > **DevOS 对“项目如何被部署、在哪里运行、如何升级和恢复”的统一抽象。** --- # 3. 核心模型 核心对象： ```text Project Environment Deployment DeploymentPlan DeploymentTarget Artifact RuntimeInstance Service Infrastructure Provider HealthCheck Release DeploymentStrategy Rollback ``` 关系： ```text Project ↓ Environment ↓ Deployment Plan ↓ Artifact ↓ Deployment Target ↓ Runtime Instance ↓ Health ``` --- # 4. Deployment 与 Infrastructure 的区别 必须明确： ```text Deployment = 把应用发布到某个运行环境 ``` 而： ```text Infrastructure = 提供运行应用所需要的基础资源 ``` 例如： ```text Infrastructure: Server Container Network Database Object Storage DNS Load Balancer Deployment: 把 my-app:v1.4.0 发布到 server-01 ``` --- # 5. Provider 基础设施不能写死。 例如： ```text Docker SSH AWS Cloudflare Kubernetes Hetzner DigitalOcean Vercel ``` 都应该通过： ```text Provider Runtime ``` 接入。 --- # 6. Provider Runtime 统一： ```ts interface InfrastructureProvider { id: string; capabilities(): ProviderCapability[]; validate( context: ProviderContext ): Promise; plan( request: DeploymentRequest ): Promise; deploy( plan: DeploymentPlan ): Promise; health( target: DeploymentTarget ): Promise; destroy( target: DeploymentTarget ): Promise; } ``` --- # 7. Provider 不拥有 Deployment Runtime Provider 只负责： ```text 实际基础设施操作 ``` 不能自己创建： ```text Task Runtime Event Runtime Secret Runtime Audit Runtime ``` 统一使用 DevOS Runtime。 --- # 8. Deployment Architecture ```text User / Agent / CI / CLI ↓ Security Context ↓ Deployment API ↓ Deployment Runtime ↓ Deployment Plan ↓ Provider Resolver ↓ Infrastructure Provider ↓ Infrastructure ↓ Runtime Instance ``` --- # 9. Deployment Action 部署本身是 Action： ```text deployment.create deployment.plan deployment.deploy deployment.cancel deployment.rollback deployment.restart deployment.scale deployment.destroy ``` 进入： ```text Action Runtime ``` --- # 10. Deployment 不直接执行 错误： ```text Controller ↓ ssh ↓ docker run ``` 正确： ```text API ↓ Action ↓ Action Runtime ↓ Deployment Runtime ↓ Provider ↓ Task ↓ Execution ``` --- # 11. Deployment Plan 部署必须先 Plan： ```text Project ↓ Artifact ↓ Environment ↓ Provider ↓ Deployment Plan ``` Plan 应包含： ```text Artifact Target Environment Services Config Secret References Network Health Checks Strategy Rollback Resource Requirements ``` --- # 12. Plan / Apply 必须支持： ```bash devos deploy plan ``` 以及： ```bash devos deploy apply ``` 原则： > **Plan 与 Apply 分离。** --- # 13. Dry Run 支持： ```bash devos deploy apply --dry-run ``` 输出： ```text Target: production/server-01 Artifact: my-app:1.4.0 Services: api worker web Changes: + api container ~ worker image ~ web image Database: migration pending Risk: HIGH ``` --- # 14. Environment Deployment 永远属于： ```text Environment ``` 例如： ```text development staging production ``` 环境来自： ```text Environment Runtime ``` 不是 Deployment 自己创建一套环境系统。 --- # 15. Environment 隔离 必须： ```text development ≠ staging ≠ production ``` 至少隔离： ```text Config Secret Credential Provider Database Storage Network Deployment Task ``` --- # 16. Production Guard Production 默认： ```text protected = true ``` 高风险操作： ```text Deploy Destroy Database Migration Rollback Credential Change Infrastructure Change ``` 可以要求： ```text Approval Runtime ``` --- # 17. Deployment Target Target 表示： > 项目最终运行在哪里。 例如： ```text server-01 docker-host-02 k8s-cluster-a aws-ecs-prod ``` --- # 18. Target Schema ```ts interface DeploymentTarget { id: string; organizationId: string; providerId: string; type: string; environmentId: string; region?: string; metadata: Record; } ``` --- # 19. Runtime Instance Deployment Target 是： ```text 目标 ``` Runtime Instance 是： ```text 实际运行实例 ``` 例如： ```text Target: server-01 Instances: api-01 worker-01 web-01 ``` --- # 20. Service 一个 Project 可以包含多个 Service： ```yaml services: api: type: web worker: type: worker agent: type: agent web: type: frontend ``` --- # 21. Service 与 Project Project： ```text 业务项目 ``` Service： ```text 运行单元 ``` 例如： ```text SEO Agent Project ├── web ├── api ├── worker └── agent ``` --- # 22. Service Definition ```ts interface ServiceDefinition { id: string; projectId: string; type: ServiceType; artifact: ArtifactReference; replicas: number; ports?: PortDefinition[]; resources?: ResourceRequirements[]; healthCheck?: HealthCheckDefinition; } ``` --- # 23. Service Types 初始： ```text web api worker agent scheduler mcp cron frontend ``` 不要过度细分。 --- # 24. Artifact Deployment 不应该直接从源码部署。 正确： ```text Source ↓ Build ↓ Artifact ↓ Deploy ``` Artifact 通过： ```text Artifact Runtime ``` 管理。 --- # 25. Artifact Examples ```text Docker Image npm Package Static Bundle Tarball Binary Server Bundle ``` Artifact 必须记录： ```text ID Version Checksum Build Project Source Revision CreatedAt ``` --- # 26. Build 与 Deployment Build： ```text 产生 Artifact ``` Deployment： ```text 发布 Artifact ``` 不要把两者混成： ```text deploy.ts ``` --- # 27. Release Release 是： > 某个 Project 在某个版本下的一组可部署 Artifact。 例如： ```text Release 1.4.0 api: image-a worker: image-b web: image-c ``` --- # 28. Release Schema ```ts interface Release { id: string; projectId: string; version: string; sourceRevision?: string; artifacts: ArtifactReference[]; createdAt: Date; } ``` --- # 29. Immutable Release Release 一旦发布： ```text 1.4.0 ``` 不应该修改成另一个内容。 如果代码变化： ```text 1.4.1 ``` 或者： ```text 1.4.0+build2 ``` 具体策略由版本 Runtime 决定。 --- # 30. Reproducible Deployment 生产部署必须尽量： ```text Release + Environment + Config Snapshot ``` 决定实际运行内容。 不能： ```text deploy latest ``` --- # 31. 禁止 latest 生产环境禁止： ```text image: latest ``` 应该： ```text image: my-app@sha256:... ``` 或者： ```text release: 1.4.0 artifactChecksum: ... ``` --- # 32. Deployment Lock Deployment 应记录： ```text DevOS Version Project Version Release Version Artifact Version Plugin Versions Module Versions Config Version Provider Version ``` 这样可以回答： > 这个生产实例到底运行了什么？ --- # 33. Configuration Deployment 配置来自： ```text Configuration Runtime ``` 例如： ```text DATABASE_URL LOG_LEVEL PUBLIC_URL MODEL WORKER_CONCURRENCY ``` 但业务代码不能： ```ts process.env.DATABASE_URL ``` 直接读取。 --- # 34. Secret Secret 来自： ```text Secret Runtime ``` Deployment 只能引用： ```text secret://production/database ``` 不能保存： ```text password: "xxxxx" ``` --- # 35. Credential Provider 认证使用： ```text Credential Runtime ``` 例如： ```text AWS Credential GitHub Credential Cloudflare Credential SSH Credential Docker Registry Credential ``` Deployment 不直接读取 Credential 数据库。 --- # 36. Network 网络配置属于： ```text Network Runtime ``` 至少抽象： ```text Port DNS Domain TLS Firewall Private Network Public Network ``` 第一阶段可以提供： ```text Docker Network SSH Host HTTP Endpoint DNS Provider ``` --- # 37. Domain Domain 例如： ```text api.example.com ``` 可以绑定： ```text Service Deployment Environment ``` 通过： ```text DNS Provider ``` 实现。 --- # 38. TLS TLS 可以由： ```text Provider ``` 提供。 例如： ```text Let's Encrypt Cloudflare Managed Certificate ``` Deployment Runtime 只描述： ```text TLS Required ``` --- # 39. Health Check 部署完成不代表成功。 必须： ```text Deploy ↓ Start ↓ Health Check ↓ Ready ``` --- # 40. Health Types 至少： ```text Process Health HTTP Health TCP Health Container Health Dependency Health Application Health ``` --- # 41. Readiness 服务： ```text running ``` 不代表： ```text ready ``` 例如： ```text Database Migration Loading Model Warming Cache ``` 期间： ```text running = true ready = false ``` --- # 42. Health Runtime 复用： ```text Health Runtime ``` 统一： ```text liveness readiness dependency health ``` --- # 43. Deployment State 建议： ```text pending planning approved deploying starting health_checking ready failed rolling_back rolled_back cancelled destroying destroyed ``` --- # 44. Deployment State Machine ```text pending ↓ planning ↓ approved ↓ deploying ↓ starting ↓ health_checking ↓ ready ``` 失败： ```text deploying ↓ failed ↓ rollback ``` --- # 45. Deployment Task 部署属于 Task： ```text Deployment ↓ Task ↓ Jobs ``` 例如： ```text Pull Artifact Prepare Stop Old Start New Health Check Switch Traffic Cleanup ``` --- # 46. Worker Deployment Job： ```text Queue ↓ Worker ``` Worker 执行： ```text Provider Action ``` 而不是： ```text Deployment Runtime ``` 自己创建 worker。 --- # 47. Deployment Timeout 每一步必须有： ```text timeout ``` 例如： ```text Pull Image: 5m Start Container: 2m Health Check: 3m Migration: 10m ``` 避免： ```text Deployment hanging forever ``` --- # 48. Deployment Cancellation 支持： ```bash devos deploy cancel ``` 但取消必须区分： ```text Safe Cancellation Unsafe Cancellation ``` 例如： ```text Pull Image ``` 可以取消。 而： ```text Database Migration ``` 可能只能： ```text Wait ``` 或者： ```text Rollback ``` --- # 49. Deployment Strategy 初始支持： ```text recreate rolling blue-green canary ``` 第一阶段： ```text recreate rolling ``` 优先。 --- # 50. Recreate ```text Old ↓ Stop ↓ New ``` 优点： ```text Simple ``` 缺点： ```text Downtime ``` --- # 51. Rolling ```text Old × N ↓ New × 1 ↓ Health ↓ New × 2 ↓ ... ``` 必须： ```text Health Check ``` 通过后才能继续。 --- # 52. Blue-Green ```text Blue = Current Green = New Deploy Green ↓ Health ↓ Switch Traffic ↓ Blue remains ``` 适合： ```text Production ``` 但第一阶段可以先作为抽象，不必立即完整实现。 --- # 53. Canary ```text 95% Old 5% New ``` 然后： ```text Metrics ↓ Health ↓ Increase ``` 这是后续增强能力。 --- # 54. Traffic Switch 流量切换通过： ```text Load Balancer Reverse Proxy DNS Service Mesh Provider ``` 实现。 Deployment Runtime 不应该自己实现所有网络层。 --- # 55. Rollback Rollback 不是： ```text git checkout ``` 而是： > 将 Deployment 恢复到一个已知的 Release。 例如： ```text Current: 1.5.0 Rollback: 1.4.2 ``` --- # 56. Rollback Preconditions 必须确认： ```text Artifact Exists Config Compatible Database Compatible Plugin Compatible Provider Available ``` 否则： ```text Rollback Blocked ``` --- # 57. Database Migration 与 Rollback 最危险的是： ```text App Upgrade + Database Migration ``` 例如： ```text v1 column A v2 column B ``` 直接回滚代码： ```text v2 → v1 ``` 可能失败。 所以 Migration 必须支持： ```text Expand Migrate Contract ``` 策略。 --- # 58. Expand / Contract 推荐： ```text Release N ↓ Add new field ↓ Release N+1 ↓ Use new field ↓ Release N+2 ↓ Remove old field ``` 避免： ```text One-step destructive migration ``` --- # 59. Deployment Rollback Plan 每次生产部署前生成： ```text Rollback Plan ``` 包括： ```text Previous Release Previous Config Database Compatibility Traffic Strategy Artifact ``` --- # 60. Deployment Approval 生产： ```text Deployment Plan ↓ Risk Evaluation ↓ Approval ↓ Apply ``` 高风险项： ```text Database Migration Permission Change Secret Change Infrastructure Destruction ``` 必须更加严格。 --- # 61. Risk Level 部署风险： ```text low medium high critical ``` 例如： ```text Frontend only: low API: medium Database Migration: high Production Destruction: critical ``` --- # 62. Deployment Policy 可以配置： ```yaml production: requireApproval: true highRisk: requireApproval: true critical: requireTwoPersonApproval: true ``` 最终由： ```text Policy Runtime ``` 执行。 --- # 63. CI/CD CI/CD 不是独立 Runtime。 它应该调用： ```text Project Factory Build Runtime Artifact Runtime Deployment Runtime Task Runtime ``` --- # 64. CI Pipeline 推荐： ```text Git Push ↓ CI ↓ Test ↓ Build ↓ Artifact ↓ Release ↓ Deployment Plan ↓ Approval ↓ Deploy ``` --- # 65. Git Provider GitHub/GitLab/Bitbucket： ```text Plugin ``` 负责： ```text Repository Webhook Commit Pull Request Release ``` --- # 66. CI Provider 第一阶段不要自己实现： ```text GitHub Actions GitLab CI Jenkins ``` 统一抽象： ```text CIRuntime ``` 后续通过 Provider 接入。 --- # 67. Deployment Trigger 支持： ```text Manual Git Push Tag Release Webhook Schedule Agent Workflow ``` 都进入： ```text Event Runtime ``` 然后： ```text Event ↓ Workflow / Deployment Action ``` --- # 68. Git Push 例如： ```text push main ``` 产生： ```text git.push ``` 然后： ```text Workflow ↓ Build ↓ Release ↓ Deploy Staging ``` 生产： ```text Approval ``` --- # 69. Scheduled Deployment 可以： ```text 每周日凌晨部署 ``` 但 Scheduler 仍然是： ```text Scheduler Runtime ``` Deployment Runtime 不创建第二套 Cron。 --- # 70. Agent Deployment Agent 可以提出： > 将 staging 上当前 release 部署到 production。 Agent： ```text Goal ↓ Capability Discovery ↓ deployment.plan ↓ Risk ↓ Approval ↓ deployment.deploy ``` Agent 不可以： ```text ssh root ``` 直接部署。 --- # 71. MCP Deployment MCP 可以提供： ```text deployment.plan deployment.status deployment.deploy deployment.rollback ``` 但： ```text MCP ↓ Action Runtime ``` 不能： ```text MCP ↓ Docker Socket ``` --- # 72. SSH Provider SSH 是基础设施 Provider。 可以： ```text Connect Execute Allowed Command Upload Download Check Health ``` 但必须： ```text Credential Runtime Security Runtime Audit Runtime ``` --- # 73. SSH Security 禁止默认： ```text root password auth unrestricted shell ``` 优先： ```text SSH Key Limited User Command Policy Network Policy Audit ``` --- # 74. Docker Provider Docker Provider 可以： ```text Image Pull Container Create Container Start Container Stop Container Remove Network Volume Health ``` 但： ```text Docker Provider ``` 不拥有： ```text Deployment State Task Queue Audit Secret ``` --- # 75. Docker Socket 生产环境： ```text Docker Socket ``` 属于高权限资源。 不能： ```text Plugin Agent MCP ``` 直接访问。 只有： ```text Infrastructure Provider ``` 在受控权限下使用。 --- # 76. Kubernetes Provider 后续支持： ```text Kubernetes ``` 但不要第一阶段直接把 Kubernetes API 暴露给： ```text Agent Plugin MCP ``` 必须： ```text Deployment Action ↓ Kubernetes Provider ``` --- # 77. Infrastructure Resource 未来可以统一抽象： ```text Server Container Volume Network Database Bucket DNS Certificate LoadBalancer ``` 这些属于： ```text Infrastructure Resource ``` --- # 78. Infrastructure Resource 与业务 Resource 不要混淆。 ```text seo.website ``` 是： ```text Business Resource ``` 而： ```text server-01 ``` 是： ```text Infrastructure Resource ``` 但两者都可以遵循： ```text Resource → Action → Runtime ``` --- # 79. Infrastructure Actions 例如： ```text infra.server.create infra.server.start infra.server.stop infra.server.restart infra.container.deploy infra.container.restart infra.dns.create infra.dns.delete ``` 统一进入： ```text Action Runtime ``` --- # 80. Infrastructure Query 查询： ```text server list container list deployment status health ``` 通过： ```text Query Runtime ``` 而不是： ```text Controller → Provider → random JSON ``` --- # 81. Infrastructure State Provider 返回的状态： ```text External State ``` DevOS 保存： ```text Desired State Observed State ``` --- # 82. Desired vs Observed 核心模型： ```text Desired ↓ Deployment Plan ↓ Provider ↓ Observed ``` 例如： ```text Desired replicas = 3 Observed replicas = 2 ``` 则： ```text Drift Detected ``` --- # 83. Drift Detection 必须支持： ```text deployment.drift.detect ``` 检测： ```text Image Replicas Ports Environment Network Domain TLS Resources ``` --- # 84. Reconciliation 后续可以： ```text Desired ↓ Observed ↓ Diff ↓ Reconcile ``` 但是第一阶段： > 只检测，不自动修复。 生产自动修复属于后续能力。 --- # 85. Infrastructure Discovery 可以： ```text devos infra discover ``` 发现： ```text Servers Containers Networks Volumes Domains ``` 然后建立： ```text Infrastructure Registry ``` --- # 86. Infrastructure Registry 记录： ```text Provider Resource External ID Environment Organization Project Status Metadata Last Observed ``` --- # 87. External ID 必须保留： ```text externalId ``` 例如： ```text AWS Instance ID Docker Container ID Kubernetes Pod ID ``` 不能只使用 DevOS UUID。 --- # 88. Import Existing Infrastructure DevOS 应支持： ```text Import ``` 例如： ```text 已有 Docker Server ↓ Discover ↓ Import ↓ Register ``` 避免强制用户重新创建所有基础设施。 --- # 89. Infrastructure Ownership 资源可以： ```text managed external imported orphaned ``` 只有： ```text managed ``` 默认允许自动修改。 --- # 90. Orphan Detection 例如： ```text DevOS Deployment 已删除 但 Docker Container 还存在 ``` 应该： ```text orphaned ``` 而不是自动删除。 --- # 91. Cost Infrastructure Provider 可以提供： ```text estimatedCost actualCost usage ``` 进入： ```text Usage / Cost Runtime ``` 例如： ```text Project ↓ Deployment ↓ Provider ↓ Cost ``` --- # 92. Resource Quota 组织可以限制： ```text CPU Memory Storage Deployments Servers Containers Network Cost ``` 通过： ```text Quota Runtime ``` 统一管理。 --- # 93. Deployment Budget Agent Deployment 还可以有： ```text Max Cost Max Instances Max Duration Max Network ``` 例如： ```text Agent: 不要让部署成本超过 $20/月。 ``` Resolver / Policy 必须验证。 --- # 94. Observability Deployment 必须自动接入： ```text Logging Metrics Tracing Audit Health ``` 所有日志包含： ```text deploymentId projectId environmentId releaseId providerId ``` --- # 95. Deployment Timeline UI 显示： ```text 12:00 Plan 12:01 Approval 12:02 Pull Artifact 12:03 Stop Old 12:04 Start New 12:05 Health Check 12:06 Traffic Switch 12:07 Completed ``` 数据来自： ```text Task Execution Event Audit ``` 而不是 UI 自己维护状态。 --- # 96. Deployment Logs Provider 日志： ```text stdout stderr command result provider response ``` 进入： ```text Logging / Artifact Runtime ``` 敏感内容必须： ```text Redact ``` --- # 97. Secret Leakage Protection 绝对禁止日志出现： ```text API_KEY PASSWORD TOKEN PRIVATE_KEY ``` 例如： ```text docker login -p XXXXX ``` 必须： ```text Redacted ``` --- # 98. Deployment Events 建议事件： ```text deployment.created deployment.planned deployment.approved deployment.started deployment.progressed deployment.ready deployment.failed deployment.rollback.started deployment.rollback.completed deployment.cancelled deployment.destroyed deployment.drift.detected ``` --- # 99. Notifications 事件可以触发： ```text Notification Runtime ``` 例如： ```text Production Deployment Failed ``` 通知： ```text In-App Email Chat ``` Deployment 不直接发 Email。 --- # 100. Deployment API 建议： ```text POST /v1/deployments POST /v1/deployments/:id/plan POST /v1/deployments/:id/apply POST /v1/deployments/:id/cancel POST /v1/deployments/:id/rollback GET /v1/deployments GET /v1/deployments/:id GET /v1/deployments/:id/tasks GET /v1/deployments/:id/logs GET /v1/deployments/:id/health ``` --- # 101. Infrastructure API ```text GET /v1/infrastructure/providers GET /v1/infrastructure/targets GET /v1/infrastructure/resources POST /v1/infrastructure/discover POST /v1/infrastructure/import ``` 所有 API： ```text Security Authorization Scope Audit ``` --- # 102. Deployment SDK 建议： ```text @devos/deployment-sdk ``` 提供： ```ts deployments.plan() deployments.apply() deployments.cancel() deployments.rollback() deployments.status() deployments.health() ``` --- # 103. CLI 最终： ```bash devos deploy plan devos deploy apply devos deploy status devos deploy logs devos deploy cancel devos deploy rollback devos deploy history ``` Infrastructure： ```bash devos infra list devos infra discover devos infra health devos infra doctor ``` --- # 104. Deployment UI 建议： ```text Deployments ├── Overview ├── Releases ├── Environments ├── Targets ├── Services ├── Health ├── Logs ├── Timeline └── Rollbacks ``` --- # 105. Project Detail Project 页面可以： ```text Project ↓ Deployments ``` 显示： ```text Development Staging Production ``` 例如： ```text Production Release: 1.4.0 Status: Healthy Target: server-01 Services: 4/4 ``` --- # 106. Deployment Detail 显示： ```text Release Artifact Environment Provider Target Services Health Task Logs Events Audit Rollback ``` --- # 107. Agent Deployment UI Agent 操作显示： ```text Goal: Deploy SEO Agent to production. Plan: 1. Validate release 2. Run migration 3. Deploy API 4. Deploy worker 5. Health check 6. Switch traffic Risk: HIGH Approval: Required ``` 而不是： ```text Agent is thinking... ``` 不展示隐藏 Chain-of-Thought。 --- # 108. Deployment History 必须保留： ```text Release Actor Time Environment Target Result Rollback ``` 可以回答： > 谁在什么时候把哪个版本部署到了生产？ --- # 109. Deployment Audit 至少记录： ```text deployment.plan deployment.approved deployment.started deployment.completed deployment.failed deployment.rollback deployment.destroy ``` --- # 110. Multi-Tenant Security 所有 Deployment： ```text organizationId projectId environmentId ``` 必须继承： ```text SecurityContext ``` 不能： ```text Project A ↓ Deploy ↓ Project B ``` --- # 111. Environment Permission 权限示例： ```text deployment.read deployment.plan deployment.deploy deployment.rollback deployment.destroy environment.production.deploy environment.production.destroy ``` --- # 112. Agent Permission Agent 权限： ```text deployment.plan ``` 不等于： ```text deployment.deploy ``` 更不等于： ```text production.destroy ``` --- # 113. Plugin Permission Plugin 例如 Docker Plugin： 只能获得： ```text docker.container.manage ``` 不能： ```text environment.production.admin ``` 除非明确授予。 --- # 114. Provider Network Security Provider 必须经过： ```text Network Policy ``` 控制： ```text Allowed Hosts Allowed Ports Regions Protocols ``` 避免任意外连。 --- # 115. SSRF 特别是： ```text HTTP Health Check Webhook Provider Metadata API ``` 必须防： ```text 169.254.169.254 localhost 127.0.0.1 private network ``` 除非显式允许。 --- # 116. Command Execution 基础设施操作涉及： ```text SSH Docker Shell ``` 必须： ```text Command Allowlist Argument Validation Timeout Resource Limit Audit ``` 禁止： ```text eval(userInput) ``` --- # 117. Shell Injection 错误： ```ts exec(`docker run ${image}`); ``` 正确： ```text Structured Command ↓ Argument Validation ↓ Command Runtime ``` --- # 118. Infrastructure Secrets 禁止： ```text SSH private key Cloud API token Registry password ``` 进入： ```text Project Manifest Git Artifact Logs ``` 只保存： ```text Credential Reference Secret Reference ``` --- # 119. Deployment Config Snapshot 部署时记录： ```text Config Version ``` 而不是： ```text 实时 Config ``` 否则： ```text 昨天部署的版本 ``` 可能因为今天 Config 改了而无法重现。 --- # 120. Secret Snapshot Secret 不保存 plaintext。 Deployment 记录： ```text secretRef secretVersion ``` 这样可以知道： > 使用的是哪个 Secret 版本。 --- # 121. Configuration Drift 如果： ```text Desired Config ≠ Observed Config ``` 报告： ```text Config Drift ``` --- # 122. Deployment Drift 统一： ```text Desired Observed ``` 检查： ```text Artifact Config Replicas Network Domain TLS Resources ``` --- # 123. Deployment Repair 第一阶段： ```text Detect Report ``` 第二阶段： ```text Plan Repair ``` 第三阶段： ```text Auto Reconcile ``` 不要第一阶段就自动修复生产。 --- # 124. Disaster Recovery Deployment Runtime 需要考虑： ```text Server Failure Provider Failure Database Failure Storage Failure Deployment Failure ``` 但： > Disaster Recovery 不应该重新创建一套 Storage/Backup Runtime。 复用： ```text Storage Runtime Artifact Runtime Backup Provider Task Runtime ``` --- # 125. Backup 生产部署前可以： ```text Database Backup Config Snapshot Artifact Retention ``` Backup 通过： ```text Backup Provider ``` 实现。 --- # 126. Restore 支持： ```bash devos backup list devos backup restore ``` 恢复必须： ```text Approval Audit Environment Protection ``` --- # 127. Rollback 与 Restore 两者必须区分： ```text Rollback = 恢复应用 Release ``` ```text Restore = 恢复数据 / 基础设施状态 ``` 不能混为一谈。 --- # 128. Disaster Recovery Plan Project 可以声明： ```yaml recovery: rpo: 1h rto: 30m ``` 最终交给： ```text Backup / Infrastructure / Deployment Runtime ``` 实现。 --- # 129. Availability 服务可以定义： ```yaml availability: replicas: 2 strategy: rolling ``` Provider 决定实际实现。 --- # 130. Resource Requirements Service 可以声明： ```yaml resources: cpu: "1" memory: "1Gi" storage: "10Gi" ``` Provider 转换为： ```text Docker Kubernetes VM Cloud ``` 各自配置。 --- # 131. Resource Abstraction 不要： ```yaml docker: memory: 1024 ``` 作为业务 Manifest。 优先： ```yaml resources: memory: "1Gi" ``` Provider 自己映射。 --- # 132. Deployment Profile Project Factory 可以声明： ```yaml deployment: profile: docker ``` 后续： ```text docker kubernetes serverless cloud bare-metal ``` --- # 133. Profile 与 Provider Profile： ```text 运行方式 ``` Provider： ```text 实际供应商 ``` 例如： ```text Profile: docker Provider: remote-docker ``` 或者： ```text Profile: kubernetes Provider: aws-eks ``` --- # 134. Deployment Environment Matrix 项目可能： ```text development → local docker staging → remote docker production → kubernetes ``` Manifest： ```yaml deployment: environments: development: profile: docker staging: profile: docker production: profile: kubernetes ``` --- # 135. Promote Release 推荐： ```text Build once Deploy many ``` 即： ```text Build ↓ Artifact ↓ Staging ↓ Production ``` 不要： ```text Build Staging Build Production ``` 两个不同 Artifact。 --- # 136. Promotion ```bash devos release promote 1.4.0 \ --from staging \ --to production ``` 实际： ```text Same Artifact ``` --- # 137. Release Approval Production Promotion： ```text Release ↓ Tests ↓ Staging Health ↓ Approval ↓ Production ``` --- # 138. Deployment Verification 部署成功必须至少验证： ```text Process Health Endpoint Dependencies Metrics Error Rate ``` --- # 139. Post-Deploy Checks 例如： ```text HTTP /health HTTP /ready Database Connection Queue Connection Storage Connection Provider Health ``` 失败： ```text Deployment Failed ``` 根据策略： ```text Rollback ``` --- # 140. Smoke Test 可以定义： ```yaml verification: smoke: - GET /health - GET /ready - GET /api/version ``` Smoke Test 通过后： ```text Deployment Ready ``` --- # 141. Deployment Verification 与 E2E Smoke： ```text 快速验证 ``` E2E： ```text 完整业务流程 ``` 两者都应该进入： ```text Task Runtime ``` --- # 142. Deployment Health Window 部署成功后可以观察： ```text 5 min 10 min 30 min ``` 检查： ```text Error Rate Latency CPU Memory Restart ``` 异常： ```text Auto Rollback ``` 这是后续增强功能。 --- # 143. Auto Rollback Policy 示例： ```yaml rollback: automatic: enabled: false errorRate: 0.1 window: 5m ``` 默认： ```text false ``` 生产自动回滚必须显式开启。 --- # 144. Deployment Notification 部署： ```text started success failed rollback ``` 通过 Notification Runtime： ```text In-App Email Slack Discord Webhook ``` --- # 145. Communication Provider Deployment 不应该直接： ```text sendSlack() sendEmail() ``` 而： ```text Event ↓ Notification ↓ Communication Runtime ``` --- # 146. Deployment Cost 每次 Deployment 可以记录： ```text Build Cost Storage Cost Provider Cost Network Cost Compute Cost ``` 进入： ```text Usage Runtime ``` --- # 147. Deployment Metrics 至少： ```text deployment.duration deployment.success deployment.failure deployment.rollback deployment.health_check deployment.drift ``` --- # 148. Infrastructure Metrics ```text infra.cpu infra.memory infra.disk infra.network infra.container.restart infra.provider.error ``` 具体采集由： ```text Observability Provider ``` 完成。 --- # 149. Deployment Tracing Trace： ```text deployment action ↓ task ↓ job ↓ provider ↓ external request ``` 所有步骤共享： ```text traceId correlationId deploymentId ``` --- # 150. Deployment Tables 建议： ```text projects environments deployment_targets deployment_plans deployments deployment_services releases release_artifacts runtime_instances health_checks deployment_events infrastructure_resources ``` --- # 151. Project 与 Deployment ```text Project 1 │ N ↓ Deployment ``` Deployment 必须属于： ```text Project Environment Release Target ``` --- # 152. Deployment 与 Release ```text Release 1 │ N ↓ Deployment ``` 同一个 Release 可以： ```text staging production ``` 分别部署。 --- # 153. Deployment 与 Artifact 一个 Release： ```text api artifact worker artifact web artifact ``` Deployment 使用： ```text exact artifact ``` --- # 154. Infrastructure Registry 建议包： ```text packages/infrastructure/ ``` 包含： ```text resource provider target discovery drift health ``` --- # 155. Deployment Packages 建议： ```text packages/deployment/ packages/release/ packages/infrastructure/ packages/deployment-provider/ packages/deployment-testing/ packages/backup/ ``` --- # 156. Provider Packages 例如： ```text plugins/ ├── docker/ ├── ssh/ ├── kubernetes/ ├── cloudflare/ └── aws/ ``` 它们通过： ```text Plugin Runtime ``` 注册 Provider。 --- # 157. Docker Plugin Docker 不应成为： ```text Core ``` 而应： ```text Official Plugin ``` 实现： ```text Infrastructure Provider ``` --- # 158. SSH Plugin 同样： ```text Official Plugin ``` 实现： ```text SSH Infrastructure Provider ``` --- # 159. Cloud Provider 例如 AWS： ```text Official / Verified Plugin ``` 提供： ```text Compute Storage Network DNS Container ``` 但每一项都通过： ```text Provider Capability ``` 声明。 --- # 160. Provider Capability ```ts interface ProviderCapability { id: string; version: string; resourceTypes: string[]; actions: string[]; regions?: string[]; } ``` 例如： ```text docker.container.deploy aws.ec2.create cloudflare.dns.create ``` --- # 161. Capability Discovery Agent 不应该知道： ```text AWS API Docker API ``` 它只发现： ```text deployment.deploy infra.container.deploy ``` 然后： ```text Provider Resolver ``` 选择具体实现。 --- # 162. Provider Health 每个 Provider： ```text healthy degraded unavailable unknown ``` Provider Runtime 应定期： ```text Health Check ``` --- # 163. Provider Failover 可以配置： ```text Primary Provider Secondary Provider ``` 但生产自动 Failover 必须： ```text Policy Approval Compatibility ``` 不能简单： ```text provider A failed → random provider B ``` --- # 164. Region Provider 支持： ```text region zone location ``` Project Manifest 不应该强绑定： ```text AWS us-east-1 ``` 除非用户明确要求。 --- # 165. Infrastructure Scheduling Deployment 可以选择： ```text region capacity cost latency compliance ``` 最终： ```text Placement Policy ``` 决定 Target。 --- # 166. Placement Policy 例如： ```yaml placement: region: ap-northeast-1 maxCost: 50 ``` Resolver： ```text Provider ↓ Available Targets ↓ Policy ↓ Selected Target ``` --- # 167. Compliance 未来可以支持： ```text data residency region restrictions network restrictions provider restrictions ``` 但这些属于： ```text Policy Runtime ``` 而不是 Deployment 自己硬编码。 --- # 168. Deployment Testing 必须覆盖： ```text Plan Apply Cancel Retry Health Rollback Drift Provider Failure Network Failure Artifact Failure Config Failure Secret Failure ``` --- # 169. Provider Contract Test 每个 Provider 必须通过： ```text Provider Contract Suite ``` 验证： ```text validate plan deploy health destroy ``` --- # 170. Fake Provider 测试环境提供： ```text FakeInfrastructureProvider ``` 可以模拟： ```text Success Timeout Failure Partial Failure Drift Health Failure ``` --- # 171. Deployment Integration Test 例如： ```text Project ↓ Build ↓ Artifact ↓ Deployment ↓ Fake Docker ↓ Health ↓ Ready ``` 完整测试。 --- # 172. E2E Test 至少： ```text Create Project ↓ Generate ↓ Build ↓ Release ↓ Deploy Staging ↓ Health ↓ Promote ↓ Deploy Production ↓ Rollback ``` --- # 173. Chaos Tests 后续增加： ```text Provider Timeout Network Disconnect Worker Crash Container Crash Health Failure Database Failure ``` 验证： ```text Recovery ``` --- # 174. Worker Crash Deployment Worker 在： ```text deploying ``` 时崩溃。 系统应该： ```text Lease Expire ↓ Job Retry ↓ Check Actual State ↓ Resume ``` 不能盲目重新执行破坏性操作。 --- # 175. Provider Timeout Provider： ```text timeout ``` 不一定代表： ```text operation failed ``` 可能： ```text external operation succeeded but response lost ``` 因此必须： ```text Query Observed State ``` 再决定 Retry。 --- # 176. Idempotency Deployment Actions 必须支持： ```text Idempotency-Key ``` 例如： ```text deploy(project=A, release=1.4.0) ``` 重复请求： ```text same operation ``` 不能创建两个 Deployment。 --- # 177. Deployment Concurrency 同一个： ```text Project + Environment ``` 默认只能有： ```text one active deployment ``` 否则： ```text Deploy A Deploy B ``` 互相覆盖。 --- # 178. Deployment Lock 使用： ```text Redis Lease ``` 协调： ```text projectId environmentId ``` 但真正状态： ```text PostgreSQL ``` 持久化。 --- # 179. Queue Deployment Queue 统一： ```text Task Job Queue Worker ``` 不要创建： ```text deploymentQueue ``` 第二套系统。 可以： ```text queue:deployment ``` 作为逻辑队列。 --- # 180. Backpressure 大量项目同时部署： ```text 1000 deployments ``` 必须： ```text Queue Concurrency Rate Limit Provider Limit ``` 控制。 --- # 181. Deployment Priority 例如： ```text critical high normal low ``` 但 Priority 不能绕过： ```text Authorization Approval Environment Guard ``` --- # 182. Deployment Maintenance Window 生产可以定义： ```yaml maintenance: allowed: - "02:00-04:00" ``` 由： ```text Scheduler / Policy Runtime ``` 判断。 --- # 183. Emergency Deployment 可以支持： ```text Break Glass ``` 但必须： ```text Short TTL Strong Audit Reason Actor Approval / notification ``` --- # 184. Production Protection 最终生产保护原则： ```text PRIVATE SCOPED AUDITED APPROVED IDEMPOTENT REVERSIBLE ``` --- # 185. Deployment Project Manifest Project Factory 可以生成： ```yaml deployment: profile: docker services: api: artifact: api replicas: 2 worker: artifact: worker replicas: 1 web: artifact: web replicas: 2 environments: staging: target: staging production: target: production approval: required ``` Manifest 描述： > Desired Deployment。 具体 Provider 实现由 Runtime 决定。 --- # 186. Deployment Profile Example ```yaml deployment: profile: docker services: api: type: api port: 3000 worker: type: worker ``` Docker Provider 可以转换： ```text Docker Network Container Port Health Check ``` --- # 187. Kubernetes Profile 未来： ```yaml deployment: profile: kubernetes ``` 同样的 Service： ```text api worker web ``` 转换成： ```text Deployment Service ConfigMap Secret Reference Ingress ``` 但： > Kubernetes-specific objects 应属于 Provider/Template，而不是业务 Manifest 核心模型。 --- # 188. Deployment Template 复杂部署可以使用： ```text Deployment Template ``` 例如： ```text single-server high-availability worker-cluster web-api-worker ``` Project Factory 可以选择。 --- # 189. Project Factory Integration 完整： ```text Project Manifest ↓ Deployment Profile ↓ Service Definition ↓ Build ↓ Artifact ↓ Release ↓ Deployment Plan ``` --- # 190. Deployment Upgrade 升级： ```text Release 1.4 ↓ Release 1.5 ``` 流程： ```text Compatibility ↓ Migration Plan ↓ Approval ↓ Deploy ↓ Health ↓ Promote ``` --- # 191. Deployment Version Compatibility 检查： ```text Project Version Runtime Version Module Version Plugin Version Artifact Version Database Version Config Version ``` 复用： ```text Compatibility Runtime ``` --- # 192. Deployment Rollback Compatibility Rollback 前： ```text Previous Release ``` 必须检查： ```text DB Schema Config Plugin Runtime Artifact ``` --- # 193. Deployment Garbage Collection 旧 Artifact / Container： ```text 保留 N 个 Release ``` 然后： ```text Garbage Collection ``` 但不能删除： ```text Current Release Rollback Release Protected Artifact ``` --- # 194. Artifact Retention 例如： ```yaml retention: releases: 10 artifacts: 30d ``` 通过： ```text Artifact Runtime ``` 执行。 --- # 195. Infrastructure Cleanup 删除 Project 后： 默认： ```text Do Not Destroy External Infrastructure ``` 必须明确： ```text destroy infrastructure ``` 并可能要求： ```text Approval ``` --- # 196. Project Delete 正确： ```text Delete Project ↓ Disable Deployment ↓ Keep Infrastructure ↓ Keep Artifacts ``` 除非用户明确： ```text Destroy ``` --- # 197. Infrastructure Destroy Destroy 是： ```text critical ``` 必须： ```text Explicit Action Strong Authorization Production Guard Audit Approval ``` --- # 198. Deployment Doctor CLI： ```bash devos deploy doctor ``` 检查： ```text Provider Credential Network Artifact Config Secret Target Health Task Queue Worker ``` --- # 199. Infrastructure Doctor ```bash devos infra doctor ``` 检查： ```text Provider Connectivity Credentials Targets Resource Drift Health Capacity Network ``` --- # 200. Deployment Architecture Rules 必须： ```text Deployment → Action Runtime Deployment → Task Runtime Deployment → Event Runtime Deployment → Provider Runtime Deployment → Config Runtime Deployment → Secret Runtime Deployment → Artifact Runtime Deployment → Health Runtime Deployment → Observability Runtime ``` 禁止： ```text Deployment → raw Docker Deployment → raw SSH Deployment → raw AWS SDK Deployment → raw DB ``` --- # 201. Package Layout 推荐： ```text packages/ ├── deployment/ │ ├── domain/ │ ├── runtime/ │ ├── plan/ │ ├── strategy/ │ ├── rollback/ │ └── verification/ │ ├── release/ ├── infrastructure/ │ ├── domain/ │ ├── runtime/ │ ├── provider/ │ ├── target/ │ ├── discovery/ │ └── drift/ │ ├── deployment-sdk/ └── deployment-testing/ ``` --- # 202. Official Providers 第一阶段建议： ```text Official Docker Provider Official SSH Provider Official Local Provider ``` 后续： ```text Kubernetes AWS Cloudflare GCP Azure ``` 通过 Plugin。 --- # 203. Local Provider 开发环境非常重要： ```text Local Provider ``` 支持： ```text Local Process Local Docker ``` 用于： ```text dev test ``` --- # 204. Mock Provider 测试： ```text Mock Provider ``` 不连接真实基础设施。 --- # 205. Provider Development 开发者可以通过： ```text @devos/plugin-sdk ``` 创建： ```text Infrastructure Provider ``` 而不是修改 DevOS Core。 --- # 206. Provider Certification Provider 需要： ```text Basic Runtime Safe Secure Verified Official ``` 与 Plugin Certification 对齐。 --- # 207. Provider Permission Manifest Provider Plugin 声明： ```yaml permissions: - infrastructure.read - infrastructure.deploy network: allowed: - "*.docker.local" ``` 高权限 Provider： ```text approval ``` --- # 208. Provider Isolation 未来支持： ```text in-process restricted isolated container ``` 与 Plugin Sandbox 共用。 --- # 209. Deployment Runtime 不拥有 Cloud SDK 例如 AWS： 禁止： ```text packages/deployment/aws.ts ``` 大量写： ```ts new EC2Client(...) ``` 应该： ```text AWS Plugin ↓ AWS Provider ↓ Deployment Runtime ``` --- # 210. Provider Adapter Provider 负责： ```text External API ↓ Normalize ``` 例如： ```text AWS response Docker response Kubernetes response ``` 统一为： ```text DeploymentResult HealthResult InfrastructureResource ``` --- # 211. Error Normalization Provider 错误： ```text DockerError AWSError SSHError ``` 统一为： ```text ProviderError DeploymentError InfrastructureError ``` 并记录： ```text code retryable safeMessage providerCode ``` --- # 212. Retry Deployment Retry： ```text Provider Retry Policy ``` 复用： ```text Task Retry Runtime ``` 不要： ```text for (let i = 0; i < 5; i++) ``` 散落在 Provider。 --- # 213. Circuit Breaker Provider 连续失败： ```text Provider ↓ Circuit Open ``` 防止： ```text 1000 Deployments ↓ 全部攻击 Provider ``` 复用： ```text Provider Runtime ``` --- # 214. Rate Limit 限制： ```text Provider Organization Project Environment Action ``` 复用： ```text Rate Limit Runtime ``` --- # 215. Deployment Observability 每个 Deployment： ```text deploymentId ``` 贯穿： ```text Action Task Job Provider Event Audit Logs Metrics Trace ``` 这样可以做到： > 一次部署全链路可追踪。 --- # 216. Deployment Cost Attribution 成本可以归属： ```text Organization Project Environment Deployment Release Agent Actor ``` 例如： ```text Agent-generated deployment ``` 可以统计： ```text agent cost infrastructure cost ``` --- # 217. Deployment Security Defaults 默认： ```text DENY PRIVATE SCOPED AUDITED APPROVED IDEMPOTENT ``` 生产： ```text PROTECTED ``` Secret： ```text SHORT-LIVED ``` Provider： ```text LEAST-PRIVILEGE ``` --- # 218. Zcode Implementation Protocol 实现本章必须按照： ```text 1. Deployment Domain 2. Release / Artifact Integration 3. Environment Integration 4. Provider Runtime 5. Mock Provider 6. Deployment Plan 7. Deployment Task 8. Health Verification 9. Docker Provider 10. SSH Provider 11. Rollback 12. Drift 13. UI 14. API / SDK / CLI 15. Security 16. Observability 17. Tests ``` --- # 219. 第一 Vertical Slice 实现： ```text Local Project ↓ Build ↓ Artifact ↓ Local Provider ↓ Deployment ↓ Health ``` 成功后再做： ```text Docker ``` --- # 220. 第二 Vertical Slice ```text Project ↓ Docker Artifact ↓ Docker Provider ↓ Deploy ↓ Health ↓ Logs ``` --- # 221. 第三 Vertical Slice 增加： ```text Task Job Queue Worker ``` 验证： ```text Deployment ``` 不会阻塞 API。 --- # 222. 第四 Vertical Slice 增加： ```text Release Promotion Staging Production Approval ``` 验证： ```text Build Once Deploy Many ``` --- # 223. 第五 Vertical Slice 增加： ```text Rollback ``` 验证： ```text Release 1 ↓ Release 2 ↓ Failure ↓ Rollback ↓ Release 1 ``` --- # 224. 第六 Vertical Slice 增加： ```text SSH Provider ``` 验证： ```text Remote Deployment ``` 但： ```text Credential Security Audit ``` 必须完整接入。 --- # 225. 第七 Vertical Slice Agent： ```text Deploy staging ``` Agent： ```text Plan ↓ Risk ↓ Approval ↓ Deploy ↓ Observe ``` --- # 226. 第八 Vertical Slice 自然语言： > 把当前 staging 版本发布到生产，如果健康检查失败就停止。 转换： ```text Goal ↓ Deployment Plan ↓ Policy ↓ Approval ↓ Deploy ↓ Health ``` Agent 不直接执行 Shell。 --- # 227. Deployment Definition of Done 第一阶段： ```text [ ] Deployment Domain [ ] Release [ ] Artifact Integration [ ] Environment Integration [ ] Deployment Plan [ ] Plan / Apply [ ] Task Integration [ ] Queue Integration [ ] Worker Integration [ ] Mock Provider [ ] Local Provider [ ] Docker Provider [ ] Health Check [ ] Logs [ ] Deployment Events [ ] Audit [ ] Rollback [ ] Deployment Lock [ ] Idempotency [ ] Security [ ] Production Guard [ ] API [ ] SDK [ ] CLI [ ] UI [ ] Contract Tests [ ] Integration Tests [ ] E2E Tests ``` --- # 228. Infrastructure Definition of Done ```text [ ] Infrastructure Resource [ ] Provider Runtime [ ] Target [ ] Discovery [ ] Import [ ] External ID [ ] Health [ ] Drift Detection [ ] Docker Provider [ ] SSH Provider [ ] Credential Integration [ ] Network Policy [ ] Provider Health [ ] Provider Contract Tests [ ] Security Tests ``` --- # 229. Zcode 禁止事项 禁止： ```text Controller → Docker ``` 禁止： ```text Controller → SSH ``` 禁止： ```text Agent → SSH ``` 禁止： ```text MCP → Docker Socket ``` 禁止： ```text Plugin → Production Infrastructure ``` 禁止： ```text Deployment → Secret plaintext ``` 禁止： ```text Deployment → custom queue ``` 禁止： ```text Deployment → custom scheduler ``` 禁止： ```text Project → infrastructure implementation ``` --- # 230. Zcode Deployment Master Prompt ```text You are implementing the DevOS Deployment & Infrastructure Runtime. Deployment is an Action Runtime capability. The correct flow is: Project → Release → Artifact → Environment → Deployment Plan → Approval → Deployment Task → Provider → Runtime Instance → Health → Event → Observability Never let business code call: Docker SSH AWS SDK Kubernetes API Cloud APIs directly. All infrastructure operations must go through: Infrastructure Provider Provider Runtime Security Runtime Credential Runtime Action Runtime Task Runtime Deployment must use exact immutable artifacts. Never deploy "latest" in production. Plan and Apply must be separate. Production deployments must respect: Authorization Environment Guard Risk Policy Approval Rollback must restore a known Release, not simply change Git code. Provider failures and timeouts must check observed state before retrying. Deployment must be idempotent. Do not create a second queue, scheduler, event bus, secret system, logging system, or configuration system. Reuse existing DevOS Runtime components. The Infrastructure Provider is an adapter to external infrastructure, not a second DevOS Runtime. ``` --- # 231. 最终 Deployment Architecture ```text USER │ AGENT │ CI │ CLI │ ↓ Security Context │ ↓ Deployment API │ ↓ Action Runtime │ ↓ Deployment Runtime │ ┌───────────┴────────────┐ ↓ ↓ Release / Artifact Environment │ │ └───────────┬────────────┘ ↓ Deployment Plan │ ↓ Policy / Approval │ ↓ Task / Job / Queue │ ↓ Worker │ ↓ Provider Runtime │ ┌────────────┼────────────┐ ↓ ↓ ↓ Docker SSH Kubernetes │ │ │ └────────────┼────────────┘ ↓ Runtime Instance │ ↓ Health │ ┌─────────┼──────────┐ ↓ ↓ ↓ Event Logs Metrics │ │ │ └─────────┼──────────┘ ↓ Observability │ ↓ Notification / UI ``` --- # 232. DevOS 完整生产闭环 经过 33 + 34，DevOS 已经形成： ```text Requirement ↓ Agent ↓ Project Manifest ↓ Project Factory ↓ Capability Resolver ↓ Module / Plugin ↓ Generated Project ↓ Build ↓ Artifact ↓ Release ↓ Deployment Plan ↓ Environment ↓ Provider ↓ Infrastructure ↓ Runtime Instance ↓ Health ↓ Observability ↓ Event ↓ Notification ``` 发生故障： ```text Health Failure ↓ Event ↓ Agent / Workflow ↓ Diagnosis ↓ Rollback Plan ↓ Approval ↓ Rollback ``` --- # 233. DevOS 最终职责边界 ```text Project Factory ↓ “我要什么项目？” Build Runtime ↓ “如何产生 Artifact？” Release Runtime ↓ “这个版本包含什么？” Deployment Runtime ↓ “这个版本部署到哪里？” Provider Runtime ↓ “如何操作外部基础设施？” Environment Runtime ↓ “在哪个隔离环境运行？” Health Runtime ↓ “它是否正常？” Observability Runtime ↓ “发生了什么？” Event Runtime ↓ “发生之后通知谁/触发什么？” Agent Runtime ↓ “下一步应该做什么？” ``` --- # 234. 最终 Golden Rule > **Deployment Runtime 负责描述和执行“把哪个 Release 部署到哪个 Environment”，Infrastructure Provider 负责“如何操作实际基础设施”；任何 Docker、SSH、Cloud、Kubernetes 操作都必须经过统一的 Security + Credential + Provider + Action + Task Runtime。** 最终链路： ```text PROJECT ↓ RELEASE ↓ ARTIFACT ↓ DEPLOYMENT PLAN ↓ ENVIRONMENT ↓ APPROVAL / POLICY ↓ TASK / JOB ↓ PROVIDER ↓ INFRASTRUCTURE ↓ RUNTIME INSTANCE ↓ HEALTH ↓ OBSERVABILITY ↓ EVENT ↓ NOTIFICATION / AGENT / ROLLBACK ``` **DevOS 不应该成为另一个 Kubernetes，也不应该成为另一个 Terraform。** 它真正应该做的是： > **把 Project、Artifact、Environment、Provider、Deployment、Task、Security、Observability 统一到一个可被 Human、Agent、Plugin、MCP、Workflow、CLI、API 共同使用的 Runtime 中。**