Common Specifications

Organization and User

Resources (channels and datasources etc) in ABEJA Platform are managed in the unit called Organization. Users also belong to organizations and are assigned roles. A user can perform an operation according to the user role in the specific organization.

Organization and administrator user information will be issued from ABEJA.

Role and user privileges

User role and corresponding operations will be described in berrow. Users are assigned one of the following roles in their organization.

Role Description
Admin A user with administrator role can manage (create/update/delete) all resources belonging to the organization.
Developer A user with a developer role can view and change resources related to model development among the organization resources.
Regular Users with regular roles can only view resources belonging to organizations.

Operation authority list

Operation Item Admin Developer Regular
User Add / Delete ◯ ✕ ✕
User Change Permission ◯ ✕ ✕
Datalake Channel Create / Archive ◯ ◯ ✕
Datalake File Upload ◯ ◯ ✕
Datalake File Delete ◯ ◯ ✕
Dataset Create / Delete ◯ ◯ ✕
Training Job Definition Create / Archive ◯ ◯ ✕
Training Job Version Create / Archive ◯ ◯ ✕
Training Job Create / Archive ◯ ◯ ✕
Notebook Create / Modify / Delete ◯ ◯ ✕
Model Upload / Download / Delete ◯ ◯ ✕
Deployment Code Version Create / Delete ◯ ◯ ✕
Deployment Code Download ◯ ◯ ✕
HTTP Service Create / Delete ◯ ◯ ✕
Endpoint Create / Switch / Delete ◯ ◯ ✕
Trigger Create / Delete ◯ ◯ ✕
Custome Image Repository Create / Delete ◯ ◯ ✕
Usage Page View ◯ ◯ ◯
Access Key Create / Delete ◯ ◯ ✕

Authentication

In order to use the API, you need to have the following information. Please check User Name from the upper right of the management screen.

  • ORGANIZATION_ID of the organization.
  • USER_ID of the user with admin role.
  • PERSONAL_ACCESS_TOKEN of the user.

Authentication is done with basic authentication. Please specify the following for user name and password.

  • User Name user-{USER_ID}
  • Password {PERSONAL_ACCESS_TOKEN}
$ curl --user user-9876543210321:b9555a16ad6513d91af399807c0c22bff3484c64 ...

Request header

At the time of API request if the body payload is JSON, please specify application/json as Content-Type in the request header.

$ curl -H "Content-Type: application/json" ...

User-Specifiable Environment Variables

When creating a job definition version/learning job/code version/HTTP service/trigger/run, the user can specify runtime environment variables.

It is a so-called environment variable, where the key consists of alphanumeric characters and underscores, and no numbers are allowed in the first character. In Trigger/Run, if you specify an environment variable with an empty value, no environment variable will be set in the execution environment. In addition, the following keys are reserved by the system and cannot be specified by the user.

  • Names beginning with ABEJA_
  • CONTEXT
  • DATADOG_ENV
  • DATADOG_SERVICE_NAME
  • DATADOG_TRACE_AGENT_HOSTNAME
  • DATASETS
  • DISPLAY
  • HANDLER
  • INPUT
  • LD_LIBRARY_PATH
  • OUTPUT
  • PLATFORM_AUTH_TOKEN
  • PYTHONUNBUFFERED
  • SERVICE_TYPE
  • STOP_AFTER
  • TENSORBOARD_ID
  • TRAINING_CLUSTER_NAMESPACE
  • TRAINING_JOB_DATASET_IDS
  • TRAINING_JOB_DATASETS
  • TRAINING_JOB_IDS
  • TRAINING_JOB_DEFINITION_NAME
  • TRAINING_JOB_DEFINITION_VERSION
  • TRAINING_JOB_ID
  • TRAINING_NOTEBOOK_ID

Updated on 07 Oct 2019