AI Core API

Lifecycle of AI scenarios.

Connect to AI Core API

class procaine.aicore.Client(ai_api_url, auth, rg='default')

Initializes the connection to AI Core and obtains a Bearer token.

Parameters:

  • ai_api_url – A URL to the AI API instance.

  • auth – Dictionary with creadentials to obtain a Bearer token. Should have url, clientid and clientsecret keys.

  • rg – (optional) A name of a resource group to use with every call.

Usage:

>>> from procaine import aicore

>>> auth = dict(url=AUTH_URL, clientid=CLIENT_ID, clientsecret=CLIENT_SECRET)
>>> api = aicore.Client(AI_API_URL, auth)

>>> api.healthz()
{'message': 'OK', 'status': 'READY'}

Scenarios & Templates

Client.list_scenarios()

Lists available scenarios.

Client.list_templates(scenario=None)

Lists templates (executables).

Parameters:

scenario – (optional) Scenario object or ID. Returns only templates belong to the scenario.

Client.template(name, scenario=None)

Returns information about a template.

Parameters:

  • name – Name of the template (executable ID).

  • scenario – (optional) Scenario object or ID.

Executions

Client.list_executions()

Returns all executions.

Client.create_execution(template, parameters=None, artifacts=None)

Starts a workflow execution.

Parameters:

  • template – Template name or object returned from Client.template() or Client.list_templates().

  • parameters – (optional) Dict of param-name -> value.

  • artifacts – (optional) Dict of binding-name -> S3 object URL or artifact or artifact ID.

Client.execution(execution)

Shows an execution status.

Parameters:

execution – Execution object or ID.

Client.execution_logs(execution)

Shows pods’ logs of an execution.

Parameters:

execution – Execution object or ID.

Client.stop_execution(execution)

Stops a running flow.

Parameters:

execution – Execution object or ID.

Artifacts

Client.list_artifacts()

Lists all registered artifacts.

Client.create_artifact(url, kind, scenario, name, description='')

Registers an artifact.

Parameters:

  • url – Object’s URL.

  • kind – One of the: model, dataset, resultset or other.

  • scenario – Scenario object or ID.

  • name – Name of an artifact.

  • description – (optional) Artifact’s description.

Client.artifact(id)

Returns artifact by ID.

Deployments

Client.list_deployments()

Returns all deployments.

Client.create_deployment(template)

Starts an inference server.

Parameters:

template – Template name or object returned from Client.template() or Client.list_templates().

Client.deployment(deployment)

Shows a deployment status.

Parameters:

deployment – Deployment object or ID.

Client.deployment_logs(deployment)

Shows pods’ logs of a deployment.

Parameters:

deployment – Deployment object or ID.

Client.stop_deployment(deployment)

Stops a model server.

Parameters:

deployment – Deployment object or ID.

Client.delete_deployment(deployment)

Deletes a model server.

Parameters:

deployment – Deployment object or ID.

Inference

Client.predict(deployment, path='', data=None)

Sends HTTP request to a model server.

Parameters:

  • deployment – Deployment object or ID.

  • path – (optional) Path attached to a deployment URL.

  • data – (optional) Data to be passed with a request. Sends a GET request when empty.

Configuration and secrets

Resource Groups

Client.list_resource_groups()

Lists resource groups.

Client.create_resource_group(name)

Creates a resource group.

Git repositories

Client.list_git_repositories()

Lists registered git repositories.

Client.create_git_repository(name, url, username=None, password=None)

Registers credentials to access a git repository.

Parameters:

  • name – The name of a secret.

  • url – A URL to the git repository.

  • username – (optional) Username to access the registry.

  • password – (optional) Password to access the registry.

Client.delete_git_repository(name)

Deletes a git repository secret.

Sync applications

Client.list_applications()

Lists created sync applications.

Client.application(app)

Returns application details.

Parameters:

app – Application object, name or ID

Client.create_application(repo_name, path='.', ref='HEAD')

Creates an application to sync git repository with AI Core.

Parameters:

  • repo_name – Name of a registered git repository.

  • path – (optional) Directory in a repo to sync.

  • ref – (optional) Git reference, i.e. branch, tag, etc.

Client.delete_application(name)

Deletes an application.

Docker registry secrets

Client.list_docker_registry_secrets()

Lists registered docker registry secrets.

Client.create_docker_registry_secret(name, registry, username, password)

Registers credentials to access a docker registry.

Parameters:

  • name – The name of a secret.

  • registry – A docker registry address.

  • username – Username to access the registry.

  • password – Password to access the registry.

Client.delete_docker_registry_secret(name)

Deletes a docker registry secret.

Object-store secrets

Client.list_s3_secrets()

Lists registered S3 secrets.

Client.create_s3_secret(name, access_key_id, secret_access_key, bucket=None, prefix=None, endpoint=None, region=None)

Registers credentials to access S3 bucket.

Parameters:

  • name – The name of a secret.

  • access_key_id – Access key ID to access the bucket.

  • secret_access_key – Secret access key to access the bucket.

  • bucket – (optional) A bucket name.

  • prefix – (optional) A key name prefix.

  • endpoint – (optional) S3 service endpoint.

  • region – (optional) A region of the bucket.

Client.delete_s3_secret(name)

Deletes an object-store secret.

Generic secrets

Client.list_generic_secrets()

Lists registered generic secrets.

Client.create_generic_secret(name, data)

Creates a generic secret.

Parameters:

  • name – The name of a secret.

  • data – Dict of key -> value.

Client.delete_generic_secret(name)

Deletes a generic secret.

Misc

Client.meta()

Returns AI API service metadata.

Client.kpi()

Provides usage statistics.

Client.healthz()

Checks AI API service status.