ZStack 资源中心

API请求

URLs

POST zstack/v1/ai/model-services/{uuid}

Headers

Authorization: OAuth the-session-uuid

Body

{
  "params": {
    "name": "example",
    "description": "This is an example description.",
    "size": 0,
    "gpuComputeCapability": "7.5",
    "startCommand": "python start.py",
    "pythonVersion": "3.8",
    "system": false,
    "type": "Endpoint",
    "yaml": "example-yaml-configuration",
    "framework": "HuggingFace",
    "requestCpu": 4,
    "requestMemory": 8192,
    "cpuArchitectures": [
      "X86_64",
      "AARCH64"
    ],
    "architectureImages": [
      {
        "cpuArchitecture": "X86_64",
        "vmImageUuid": "21cb9f654e9d3feab962ce3f5eec5796",
        "dockerImage": "registry.example.com/x86_64/myimage:latest"
      },
      {
        "cpuArchitecture": "AARCH64",
        "vmImageUuid": "1d919fda666b3bdcb5f4925f1cf17c71",
        "dockerImage": "registry.example.com/aarch64/myimage:latest"
      }
    ]
  },
  "systemTags": [],
  "userTags": []
}

说明：上述示例中systemTags、userTags字段可以省略。列出是为了表示body中可以包含这两个字段。

Curl示例

curl -H "Content-Type: application/json;charset=UTF-8" \
-H "Authorization: OAuth b86c9016b4f24953a9edefb53ca0678c" \
-X POST -d '{"params":{"name":"example","description":"This is an example description.","size":0,"gpuComputeCapability":"7.5","startCommand":"python start.py","pythonVersion":"3.8","system":false,"type":"Endpoint","yaml":"example-yaml-configuration","framework":"HuggingFace","requestCpu":4,"requestMemory":8192,"cpuArchitectures":["X86_64","AARCH64"],"architectureImages":[{"cpuArchitecture":"X86_64","vmImageUuid":"21cb9f654e9d3feab962ce3f5eec5796","dockerImage":"registry.example.com/x86_64/myimage:latest"},{"cpuArchitecture":"AARCH64","vmImageUuid":"1d919fda666b3bdcb5f4925f1cf17c71","dockerImage":"registry.example.com/aarch64/myimage:latest"}]}}' \
http://localhost:8080/zstack/v1/ai/model-services/d76744cb53f9378abf42b907df3efb6e

参数列表


名字	类型	位置	描述	可选值	起始版本
uuid	String	url	资源的UUID，唯一标示该资源		5.3.20
name (可选)	String	body(包含在params结构中)	资源名称		5.3.20
description (可选)	String	body(包含在params结构中)	资源的详细描述		5.3.20
readme (可选)	String	body(包含在params结构中)	模型服务的说明文档		5.3.20
size (可选)	long	body(包含在params结构中)	模型服务的大小		5.3.20
gpuComputeCapability (可选)	String	body(包含在params结构中)	GPU计算能力		5.3.20
installPath (可选)	String	body(包含在params结构中)	模型服务存储的url		5.3.20
startCommand (可选)	String	body(包含在params结构中)	模型服务的启动命令		5.3.20
pythonVersion (可选)	String	body(包含在params结构中)	模型服务的python版本		5.3.20
condaVersion (可选)	String	body(包含在params结构中)	模型服务的conda版本		5.3.20
system (可选)	Boolean	body(包含在params结构中)	是否为系统默认模型服务		5.3.20
type (可选)	String	body(包含在params结构中)	模型服务的类型	Endpoint FineTune App ModelEval	5.3.20
yaml (可选)	String	body(包含在params结构中)	模型服务的yaml配置内容		5.3.20
source (可选)	String	body(包含在params结构中)	模型服务的源	Other Bentoml HuggingFace	5.3.20
framework (可选)	String	body(包含在params结构中)	模型服务的框架		5.3.20
requestCpu (可选)	Integer	body(包含在params结构中)	模型服务需求的CPU		5.3.20
requestMemory (可选)	Long	body(包含在params结构中)	模型服务需求的内存		5.3.20
resourceUuid (可选)	String	body(包含在params结构中)	资源UUID		5.3.20
tagUuids (可选)	List	body(包含在params结构中)	标签UUID列表		5.3.20
systemTags (可选)	List	body	系统标签		5.3.20
userTags (可选)	List	body	用户标签		5.3.20
cpuArchitectures (可选)	List	body(包含在params结构中)			5.3.40
architectureImages (可选)	List	body(包含在params结构中)	不同CPU架构对应的镜像配置。		5.3.40
supportDistributed (可选)	Boolean	body(包含在params结构中)	是否支持分布式推理		5.3.46
containerCommand (可选)	String	body(包含在params结构中)	容器启动命令。		5.5.22
containerArgs (可选)	String	body(包含在params结构中)	容器启动参数。		5.5.22
vendorToSpecUuidsMap (可选)	Map	body(包含在params结构中)	GPU厂商到GPU规格UUID列表的映射。		5.5.22

API返回

返回示例

{
  "inventory": {
    "uuid": "90d425a802ff44eeb530f21587633700",
    "name": "example",
    "description": "Example description for modelService",
    "yaml": "services:\n  - ports:\n      - 3000\n  name: qwen1.5-7b-chat:2b34xhrmqwhomjkd\n  livez: /livez\n  readyz: /readyz\n  serviceBootupTime: 30\nenv:\n  - key:value\n  - key:value\ndistro:\n  packages: vim,nfs-utils\npython:\n  requirements_txt: ./requirements.txt\n  index_url: https://pypi.tuna.tsinghua.edu.cn/simple\n  trusted_host: pypi.tuna.tsinghua.edu.cn\n",
    "requestCpu": 1,
    "requestMemory": 1024,
    "modelCenterUuid": "432c5fdd49374bb0a2fd7877f0a877cf",
    "type": "Endpoint",
    "system": true,
    "gpuComputeCapability": "3.7",
    "installPath": "/example/install/path",
    "pythonVersion": "3.8.10",
    "condaVersion": "23.7.4",
    "startCommand": "python3 app.py",
    "supportDistributed": true,
    "modelServiceImages": [
      {
        "uuid": "c13336388d574fec824752525ae05a21",
        "modelServiceUuid": "90d425a802ff44eeb530f21587633700",
        "cpuArchitecture": "x86_64",
        "vmImageUuid": "b31bc4574f22426a89902edac5d14a72",
        "dockerImage": "registry.example.com/x86_64/myimage:latest",
        "createDate": "Nov 25, 2025 11:51:50 AM",
        "lastOpDate": "Nov 25, 2025 11:51:50 AM"
      },
      {
        "uuid": "46c0e4bb46e849059e6d6bc8881d86f3",
        "modelServiceUuid": "90d425a802ff44eeb530f21587633700",
        "cpuArchitecture": "aarch64",
        "vmImageUuid": "b2085e2ac2d84029a7947cb09178daaa",
        "dockerImage": "registry.example.com/aarch64/myimage:latest",
        "createDate": "Nov 25, 2025 11:51:50 AM",
        "lastOpDate": "Nov 25, 2025 11:51:50 AM"
      }
    ],
    "createDate": "Nov 25, 2025 11:51:50 AM",
    "lastOpDate": "Nov 25, 2025 11:51:50 AM"
  }
}


名字	类型	描述	起始版本
success	boolean		5.3.20
error	ErrorCode	错误码，若不为null，则表示操作失败, 操作成功时该字段为null。详情参考error	5.3.20

#error


名字	类型	描述	起始版本
code	String	错误码号，错误的全局唯一标识，例如SYS.1000, HOST.1001	5.2.1
description	String	错误的概要描述	5.2.1
details	String	错误的详细信息	5.2.1
elaboration	String	保留字段，默认为null	5.2.1
opaque	LinkedHashMap	保留字段，默认为null	5.2.1
cause	ErrorCode	根错误，引发当前错误的源错误，若无原错误，该字段为null	5.2.1

SDK示例

Java SDK

CloneModelServiceAction action = new CloneModelServiceAction();
action.uuid = "d76744cb53f9378abf42b907df3efb6e";
action.name = "example";
action.description = "This is an example description.";
action.size = 0;
action.gpuComputeCapability = "7.5";
action.startCommand = "python start.py";
action.pythonVersion = "3.8";
action.system = false;
action.type = "Endpoint";
action.yaml = "example-yaml-configuration";
action.framework = "HuggingFace";
action.requestCpu = 4;
action.requestMemory = 8192;
action.cpuArchitectures = asList("X86_64","AARCH64");
action.architectureImages = asList([cpuArchitecture:X86_64, vmImageUuid:21cb9f654e9d3feab962ce3f5eec5796, dockerImage:registry.example.com/x86_64/myimage:latest],[cpuArchitecture:AARCH64, vmImageUuid:1d919fda666b3bdcb5f4925f1cf17c71, dockerImage:registry.example.com/aarch64/myimage:latest]);
action.sessionId = "b86c9016b4f24953a9edefb53ca0678c";
CloneModelServiceAction.Result res = action.call();

Python SDK

CloneModelServiceAction action = CloneModelServiceAction()
action.uuid = "d76744cb53f9378abf42b907df3efb6e"
action.name = "example"
action.description = "This is an example description."
action.size = 0
action.gpuComputeCapability = "7.5"
action.startCommand = "python start.py"
action.pythonVersion = "3.8"
action.system = false
action.type = "Endpoint"
action.yaml = "example-yaml-configuration"
action.framework = "HuggingFace"
action.requestCpu = 4
action.requestMemory = 8192
action.cpuArchitectures = [X86_64, AARCH64]
action.architectureImages = [[cpuArchitecture:X86_64, vmImageUuid:21cb9f654e9d3feab962ce3f5eec5796, dockerImage:registry.example.com/x86_64/myimage:latest], [cpuArchitecture:AARCH64, vmImageUuid:1d919fda666b3bdcb5f4925f1cf17c71, dockerImage:registry.example.com/aarch64/myimage:latest]]
action.sessionId = "b86c9016b4f24953a9edefb53ca0678c"
CloneModelServiceAction.Result res = action.call()

克隆模型服务 (CloneModelService)

API请求

API返回

SDK示例