CLI - Manage access tokens

Learn how to create application tokens for AI Training.

Requirements

• a working ovhai CLI (learn more here)

Instructions

Token creation help

If you need any help while creating a new token, run ovhai token create --help:

Usage: ovhai token create [OPTIONS] --role <ROLE> <NAME>

Arguments:
  <NAME>  Token's name

Options:
  -r, --role <ROLE>                  Operator role gives all access, read role only allows access to a job's APIs [possible values: read, operator]
      --token <TOKEN>                Authentication using Token rather than OAuth
  -l, --label-selector <name=value>  Only jobs with this label's name and value will be accessible using this token
  -o, --output <OUTPUT>              Command output format [possible values: json, yaml, description]
      --no-color                     Remove colors from output
  -h, --help                         Print help

Create an app token

The creation of application tokens is pretty straightforward; you need to define what roles to assign to the token and its scope.

There are two roles available:

• read: A token with this role will allow access to the services exposed by the job on the job_url.
• operator: A token with this role will allow interaction with AI Training API and CLI.

The tokens can be scoped to apply only to jobs matching a specific label. Upon creating a token simply provide a name=value pair and when running a new job add the label to the job to be accessible ovhai job run -l name=value. If you create a new job using an app token the label-selector is automatically added to the created job so that it falls within the token scope.

Let us create a new operator token scoped to model=base and named basetoken:

ovhai token create basetoken --label-selector model=base --role operator

Use the app token

If your token has a role read and you wish to access the job_url or if it is an operator token and you wish to use the AI Training API, simply add the following header to your HTTP requests:

-H 'Authorization: Bearer <token>'

You can also use the operator token directly with the ovhai CLI. For all commands, simply add the flag --app-token to override the default authentication method. Running a simple Hello World with the app token :

ovhai job run --app-token <my-token> ubuntu -- echo `Hello World`

Token lifecycle

Application tokens do not have an expiration date. We need an additional mechanism to revoke those tokens and remove access to unwanted applications.

There are two ways to revoke a token.

token delete

You can delete any token using its ID. List the tokens and find the ID associated with your token. Then simply run:

ovhai token delete <token-id>

token renew

You may wish to revoke a token while keeping its name and label-selector, a typical use case is if your token leaked or is compromised in any way. Rather than deleting and recreating the token, you can renew it. Simply run:

ovhai token renew <token-id>

Go further

For more information and tutorials, please see our other AI & Machine Learning support guides or explore the guides for other OVHcloud products and services.

If you need training or technical assistance to implement our solutions, contact your sales representative or click on this link to get a quote and ask our Professional Services experts for a custom analysis of your project.