EmbodiFlow文档
开始使用数据工作流与算法训练与部署开发者参考
  • 开发者参考

认证

  • 认证概览
  • API 密钥
  • 个人访问令牌
  • 浏览器会话

REST API

  • 使用 REST API
  • 全部端点
  • 项目
    • 录制
      • 上传
        • 任务片段与版本
          • 标注
            • 质量与任务
              • 检索
                • 工作流
                  • 算法
                    • 模型
                      • 产物与导出
                        • GET
                          List artifacts
                        • GET
                          Get a artifact
                        • GET
                          List compute profiles
                        • GET
                          List deployment packages
                        • POST
                          Create a deployment package
                        • GET
                          Get a deployment package
                        • GET
                          List download sessions
                        • POST
                          Create a download session
                        • GET
                          Get a download session
                        • POST
                          Presigned GET URLs for CLI / scripts; auth = download sessionToken only.
                        • GET
                          List embodiment profiles
                        • GET
                          Get a embodiment profile
                        • GET
                          List gpu class profiles
                        • POST
                          Create a completion request
                        • GET
                          List inference deployments
                        • POST
                          Create a inference deployment
                        • GET
                          Get a inference deployment
                        • POST
                          Connection details for a robot client (ADR 0145 §4). Read-tier because it hands out a dial target and the deployment's own token, not because it is harmless — the token is the credential guarding both wires, so the response carries no internal Kubernetes identifier.
                        • POST
                          Create a control request
                        • POST
                          Create a invocation
                        • GET
                          List model image definitions
                        • POST
                          Create a model image definition
                        • GET
                          Get a model image definition
                        • GET
                          List revisions for a model image definition
                        • POST
                          Create a revision
                        • POST
                          Create a publication request
                        • POST
                          Human review (must not be the revision author). Publish is a separate step.
                        • POST
                          Create a revocation
                        • POST
                          Create a image ref validation
                        • GET
                          List model versions
                        • POST
                          Create a model version
                        • DELETE
                          Delete a model version
                        • GET
                          Get a model version
                        • POST
                          Create a archival
                        • POST
                          Queue a model bundle build (ADR 0145 §5). Idempotent per reason.
                        • POST
                          Edge inference package (ADR 0072) — openpi_client + embodiment fail-closed.
                        • POST
                          Create a promotion
                        • POST
                          Create a import
                        • POST
                          Create a imported checkpoint
                        • GET
                          List exports of a version
                        • POST
                          Create or reuse a cached version export
                        • GET
                          Get a version export
                        • POST
                          Create a download session of presigned version-export files
                        • GET
                          Preview what an export of a version contains, with the unreviewed AI episodes and the privacy warnings
                        • GET
                          Recommend a training policy for a version
                        • POST
                          Start training from a version
                        • DELETE
                          Delete a training checkpoint
                        • GET
                          Get a training checkpoint
                        • POST
                          Create a download
                        • GET
                          List training jobs
                        • POST
                          Create a training job
                        • GET
                          Get a training job
                        • GET
                          List attempts for a training job
                        • GET
                          Get a attempt
                        • GET
                          List checkpoints for a attempt
                        • GET
                          List logs for a attempt
                        • GET
                          List metrics for a attempt
                        • POST
                          Create a cancellation request
                        • GET
                          List reproducibility bundle for a training job
                        • POST
                          Create a retry request
                      • 计费与用量
                        • 团队与治理
                          • 账号与通知
                            • Webhook 端点

                              事件

                              • 出站事件
                              • 签名校验

                              工具

                              • 命令行
                              • SDK
                              • MCP

                              更新

                              • 更新日志

                              © 2026 EmbodiFlow

                              llms.txtOpenAPI JSON
                              1. 开发者参考
                              2. REST API
                              3. 产物与导出
                              4. Edge inference package (ADR 0072) — openpi_client + embodiment fail-closed.

                              Edge inference package (ADR 0072) — openpi_client + embodiment fail-closed.

                              POST
                              /v1/model-versions/{modelVersionId}/edge-inference-package
                              curl -X POST 'https://api.embodiflow.io/v1/model-versions/{modelVersionId}/edge-inference-package' \
                                -H 'Authorization: Bearer $EMBODIFLOW_PAT' \
                                -H 'Content-Type: application/json' \
                                -d '{
                                  "embodimentProfileId": "00000000-0000-0000-0000-000000000000",
                                  "mode": "local",
                                  "strategyKey": "string",
                                  "remotePolicyHost": "string",
                                  "remotePolicyPort": 0,
                                  "steps": 0,
                                  "targetArch": "x86-cuda",
                                  "wireProtocols": "openpi-ws",
                                  "useFakeTransport": true
                                }'

                              认证

                              每个请求都需要携带个人访问令牌(PAT)。

                              • Authorization: Bearer PAT
                              认证指南

                              路径参数

                              • modelVersionIdstring必填

                              请求头参数

                              • Idempotency-Keystring

                                Optional idempotency key, at most 255 characters (UUIDv4 recommended). A repeated key returns the first response byte-for-byte, including a 5xx. The same key with a different body, or a second in-flight request, returns 409 IDEMPOTENCY_CONFLICT. Keys are unique per tenant and retained for 24 hours.

                              请求体

                              Content-Typeapplication/json
                              • embodimentProfileIdstring (uuid)必填
                              • modestring必填
                                可选值
                                local
                                remote_ws
                              • strategyKeystring
                              • remotePolicyHoststring
                              • remotePolicyPortnumber
                              • stepsnumber
                              • targetArchstring

                                Image line (ADR 0145 §6). Jetson is a separate line, not a platform.

                                可选值
                                x86-cuda
                                jetson-jp6
                                cpu
                              • wireProtocolsarray<string>

                                Non-empty subset when provided; defaults to `openpi-ws` only.

                              • useFakeTransportboolean

                              响应

                              Success

                              • dataobject必填
                                • schemaVersionstring必填
                                  可选值
                                  embodiflow-edge-inference-package/v1
                                • modelVersionIdstring必填
                                • tenantIdstring必填
                                • projectIdstring必填
                                • embodimentProfileIdstring必填
                                • embodimentCodestring必填
                                • modestring必填
                                  可选值
                                  local
                                  remote_ws
                                • weightUrlsarray<string>必填
                                • runtimeImagestring | null必填
                                • edgeConfigobject必填
                                • runCommandstring必填
                                • docLinksobject必填
                                • generatedAtstring必填
                                • expiresAtstring必填
                                • targetArchstring必填
                                  可选值
                                  x86-cuda
                                  jetson-jp6
                                  cpu
                                • wiresarray<EdgeWireEndpointDto>必填
                                • hostRequirementsobject必填
                                • bundlePullobject必填
                                • securityNoticestring | null必填
                              响应示例
                              {
                                "data": {
                                  "schemaVersion": "embodiflow-edge-inference-package/v1",
                                  "modelVersionId": "string",
                                  "tenantId": "string",
                                  "projectId": "string",
                                  "embodimentProfileId": "string",
                                  "embodimentCode": "string",
                                  "mode": "local",
                                  "weightUrls": [
                                    "string"
                                  ],
                                  "runtimeImage": "string",
                                  "edgeConfig": {},
                                  "runCommand": "string",
                                  "docLinks": {},
                                  "generatedAt": "string",
                                  "expiresAt": "string",
                                  "targetArch": "x86-cuda",
                                  "wires": [
                                    {
                                      "protocol": "openpi-ws",
                                      "port": 0,
                                      "plaintext": false
                                    }
                                  ],
                                  "hostRequirements": {
                                    "targetArch": "x86-cuda",
                                    "gpuRuntime": "string",
                                    "dockerGpuFlag": "string",
                                    "minGpuMemoryGb": 0,
                                    "notes": [
                                      "string"
                                    ]
                                  },
                                  "bundlePull": {
                                    "bundleObjectPrefix": "string",
                                    "downloadSessionId": "string",
                                    "sessionToken": "string",
                                    "expiresAt": "string",
                                    "presignedUrls": [
                                      "string"
                                    ]
                                  },
                                  "securityNotice": "string"
                                }
                              }
                              服务器https://api.embodiflow.io/v1https://api.embodiflow.dev/v1
                              上一页

                              POST /model-versions/{modelVersionId}/bundle-builds

                              下一页

                              POST /model-versions/{modelVersionId}/promotions

                              curl -X POST 'https://api.embodiflow.io/v1/model-versions/{modelVersionId}/edge-inference-package' \
                                -H 'Authorization: Bearer $EMBODIFLOW_PAT' \
                                -H 'Content-Type: application/json' \
                                -d '{
                                  "embodimentProfileId": "00000000-0000-0000-0000-000000000000",
                                  "mode": "local",
                                  "strategyKey": "string",
                                  "remotePolicyHost": "string",
                                  "remotePolicyPort": 0,
                                  "steps": 0,
                                  "targetArch": "x86-cuda",
                                  "wireProtocols": "openpi-ws",
                                  "useFakeTransport": true
                                }'
                              响应示例
                              状态码
                              201
                              ​
                              {
                                "data": {
                                  "schemaVersion": "embodiflow-edge-inference-package/v1",
                                  "modelVersionId": "string",
                                  "tenantId": "string",
                                  "projectId": "string",
                                  "embodimentProfileId": "string",
                                  "embodimentCode": "string",
                                  "mode": "local",
                                  "weightUrls": [
                                    "string"
                                  ],
                                  "runtimeImage": "string",
                                  "edgeConfig": {},
                                  "runCommand": "string",
                                  "docLinks": {},
                                  "generatedAt": "string",
                                  "expiresAt": "string",
                                  "targetArch": "x86-cuda",
                                  "wires": [
                                    {
                                      "protocol": "openpi-ws",
                                      "port": 0,
                                      "plaintext": false
                                    }
                                  ],
                                  "hostRequirements": {
                                    "targetArch": "x86-cuda",
                                    "gpuRuntime": "string",
                                    "dockerGpuFlag": "string",
                                    "minGpuMemoryGb": 0,
                                    "notes": [
                                      "string"
                                    ]
                                  },
                                  "bundlePull": {
                                    "bundleObjectPrefix": "string",
                                    "downloadSessionId": "string",
                                    "sessionToken": "string",
                                    "expiresAt": "string",
                                    "presignedUrls": [
                                      "string"
                                    ]
                                  },
                                  "securityNotice": "string"
                                }
                              }