Cloud Client

What is Xplainable Cloud?

Xplainable Cloud is a hosted service that allows you to persist and load models and preprocessing pipelines and collaborate on them within teams and organisations. Persisted models are also able to be deployed as API endpoints in seconds. The cloud service is accessible via a web interface to manage organisations, teams, and users and provides an excellent interface for visualising model explainers and metrics. You can find more information about Xplainable Cloud at https://www.xplainable.io.

What is the Cloud Client?

The cloud client is built into the xplainable python package, allowing you to connect to Xplainable Cloud and query the API, enabling you to manage your account, models, and deployments within Python.

Initialising a session

To initialise a session, you first must generate an API key at https://app.xplainable.io.

Copyright Xplainable Pty Ltd, 2023

xplainable.client.init.initialise(hostname='https://api.xplainable.io', api_key=None)[source]

Initialise the client with an API Key.

API Keys can be generated from https://app.xplainable.io with a valid account.

Example

>>> xp.initialise()

Returns:: The users account information.
Return type:: dict

xplainable.client.init.reinitialise()[source]

Forgets the current API Key and reinitialises the client.

Example

>>> xp.reinitialise()

Querying the API

When you connect successfully to Xplainable Cloud, you can use the client to query the API. The client is accessible by running:

import xplainable as xp

# Initialise your session
xp.initialise()

# Query the API
xp.client.list_models()

class xplainable.client.client.Client(api_key, hostname='https://api.xplainable.io')[source]

Bases: object

A client for interfacing with the xplainable web api (xplainable cloud).

Access models, preprocessors and user data from xplainable cloud. API keys can be generated at https://app.xplainable.io.

Parameters:: api_key (str) – A valid api key.

create_model_id(model, model_name: str, model_description: str) → str[source]

Creates a new model and returns the model id.

Parameters:

model_name (str) – The name of the model
model_description (str) – The description of the model
model (XClassifier | XRegressor) – The model to create.

Returns:

The model id

Return type:

int

create_model_version(model, model_id: str, x: DataFrame, y: Series) → str[source]

Creates a new model version and returns the version id.

Parameters:

model_id (int) – The model id
partition_on (str) – The partition column name
ruleset (dict | str) – The feeature ruleset
health_info (dict) – Feature health information
versions (dict) – Versions of current environment

Returns:

The model version id

Return type:

int

create_preprocessor_id(preprocessor_name: str, preprocessor_description: str) → str[source]

Creates a new preprocessor and returns the preprocessor id.

Parameters:

preprocessor_name (str) – The name of the preprocessor
preprocessor_description (str) – The description of the preprocessor

Returns:

The preprocessor id

Return type:

int

create_preprocessor_version(preprocessor_id: str, preprocessor) → str[source]

Creates a new preprocessor version and returns the version id.

Parameters:

preprocessor_id (int) – The preprocessor id
stages (dict) – The preprocessor stages
deltas (dict) – The preprocessor deltas
versions (dict) – Versions of current environment

Returns:

The preprocessor version id

Return type:

int

deploy(hostname: None, model_id: str, version_id: str, partition_id: str, raw_output: bool = True) → dict[source]

Deploys a model partition to xplainable cloud.

The hostname should be the url of the inference server. For example: https://inference.xplainable.io

Parameters:

hostname (str) – The host name for the inference server
model_id (int) – The model id
version_id (int) – The version id
partition_id (int) – The partition id
raw_output (bool, optional) – returns a dictionary

Returns:

deployment status and details.

Return type:

dict

generate_deploy_key(description: str, deployment_id: str, days_until_expiry: float = 90, surpress_output: bool = False) → None[source]

Generates a deploy key for a model deployment.

Parameters:

description (str) – Description of the deploy key use case.
deployment_id (str) – The deployment id.
days_until_expiry (float) – The number of days until the key expires.
surpress_output (bool) – Surpress output. Defaults to False.

Returns:

No key is returned. The key is copied to the clipboard.

Return type:

None

get_user_data() → dict[source]

Retrieves the user data for the active user.

Returns:: User data
Return type:: dict

list_model_versions(model_id: int) → list[source]

Lists all versions of a model.

Parameters:: model_id (int) – The model id
Returns:: Dictionary of model versions.
Return type:: dict

list_models() → list[source]

Lists all models of the active user’s team.

Returns:: Dictionary of saved models.
Return type:: dict

list_preprocessor_versions(preprocessor_id: int) → list[source]

Lists all versions of a preprocessor.

Parameters:: preprocessor_id (int) – The preprocessor id
Returns:: Dictionary of preprocessor versions.
Return type:: dict

list_preprocessors() → list[source]

Lists all preprocessors of the active user’s team.

Returns:: Dictionary of preprocessors.
Return type:: dict

load_classifier(model_id: int, version_id: int, model=None)[source]

Loads a binary classification model by model_id

Parameters:

model_id (str) – A valid model_id
version_id (str) – A valid version_id
model (PartitionedClassifier) – An existing model to add partitions

Returns:

The loaded xplainable classifier

Return type:

xplainable.PartitionedClassifier

load_preprocessor(preprocessor_id: int, version_id: int, response_only: bool = False)[source]

Loads a preprocessor by preprocessor_id and version_id.

Parameters:

preprocessor_id (int) – The preprocessor id
version_id (int) – The version id
response_only (bool, optional) – Returns the preprocessor metadata.

Returns:

The loaded preprocessor

Return type:

xplainable.preprocessing.Preprocessor

load_regressor(model_id: int, version_id: int, model=None)[source]

Loads a regression model by model_id and version_id

Parameters:

model_id (str) – A valid model_id
version_id (str) – A valid version_id
model (PartitionedRegressor) – An existing model to add partitions to

Returns:

The loaded xplainable regressor

Return type:

xplainable.PartitionedRegressor