hydrosdk.application module¶

class hydrosdk.application.Application(cluster, id, name, execution_graph, status, signature, kafka_streaming, metadata=None, message=None)¶

Bases: object

Applications are used to combine your ModelVersions into a linear graph and deploy it into production, exposing HTTP and gRPC interfaces for consumers.

Use ApplicationBuilder class to create a new Application in your Hydrosphere cluster or an Application.find method to get the existing Application.

Example

Parameters

cluster (hydrosdk.cluster.Cluster) –
id (int) –
name (str) –
execution_graph (ExecutionGraph) –
status (hydrosdk.application.ApplicationStatus) –
signature (hydro_serving_grpc.serving.contract.signature_pb2.ModelSignature) –
kafka_streaming (List[hydrosdk.application.StreamingParams]) –
metadata (Optional[Dict[str, str]]) –
message (Optional[str]) –

Return type

Application

List all applications created on the cluster.

>>> from hydrosdk.cluster import Cluster
>>> cluster = Cluster("http-cluster-endpoint")
>>> apps = Application.list(cluster)
>>> for app in apps:
>>>     print(app)

Find an application by name and perform a prediction from it.

>>> from hydrosdk.cluster import Cluster
>>> cluster = Cluster("http-cluster-endpoint", "grpc-cluster-endpoint")  # important to use a gRPC endpoint
>>> app = Application.find(cluster, "my-application")
>>> pred = app.predictor()
>>> resp = pred.predict({"my-input": 1})

Parameters

cluster (hydrosdk.cluster.Cluster) – active cluster
id (int) – unique application ID
name (str) – application Name
signature (hydro_serving_grpc.serving.contract.signature_pb2.ModelSignature) – signature, specifying input and output fields names, dtypes and shapes
execution_graph (ExecutionGraph) – linear graph which specifies ExecutionStages which sequentially transform input
status (hydrosdk.application.ApplicationStatus) – Application Status
kafka_streaming (List[hydrosdk.application.StreamingParams]) – list of Kafka parameters with input and output Kafka topics specified
metadata (Optional[Dict[str, str]]) – metadata with string keys and string values.
message (Optional[str]) – possible error message from the cluster

Return type

Application

static delete(cluster, application_name)¶

Delete an application by name.

Parameters

cluster (hydrosdk.cluster.Cluster) – active cluster
application_name (str) – application name

Returns

response from the cluster

Return type

dict

static find(cluster, application_name)¶

Search for an application by name.

Parameters

cluster (hydrosdk.cluster.Cluster) – active cluster
application_name (str) – application name

Returns

deserialized application object

Return type

hydrosdk.application.Application

static list(cluster)¶

List all available applications from the cluster.

Parameters: cluster (hydrosdk.cluster.Cluster) – active cluster
Returns: deserialized list of application objects
Return type: List[hydrosdk.application.Application]

lock_while_starting(timeout=120)¶

Wait for an application to become ready.

Parameters: timeout (int) –
Return type: hydrosdk.application.Application

predictor(return_type=PredictorDT.DICT_NP_ARRAY)¶

Return a predictor object which is used to transform your data into a proto message, pass it via gRPC to the cluster and decode the cluster output from proto to a dict with Python dtypes.

Parameters: return_type – Specifies into which format should predictor serialize model output. Numpy dtypes, Python dtypes or pd.DataFrame are supported.
Returns: PredictServiceClient with .predict() method which accepts your data
Return type: hydrosdk.predictor.PredictServiceClient

class hydrosdk.application.ApplicationBuilder(name)¶

Bases: object

ApplicationBuilder is used to create new Applications in your cluster.

Example
Parameters: name (str) –
Return type: ApplicationBuilder

Create an application from existing modelversions.

>>> from hydrosdk import Cluster, ModelVersion
>>> cluster = Cluster('http-cluster-endpoint')
>>> mv1 = ModelVersion.find(cluster, "my-model", 1)
>>> mv2 = ModelVersion.find(cluster, "my-model", 2)
>>> stage = ExecutionStageBuilder()                     .with_model_variant(mv1, 50)                     .with_model_variant(mv2, 50) \
                .build()
>>> app = ApplicationBuilder("my-application-ab-test")                 .with_stage(stage)                 .build(cluster)

Parameters

cluster – Hydrosphere cluster where you want to create an Application
name (str) – Future Application name

Return type

ApplicationBuilder

build(cluster)¶

Create an Application in your Hydrosphere cluster.

Returns: Application object
Parameters: cluster (hydrosdk.cluster.Cluster) –
Return type: hydrosdk.application.Application

with_kafka_params(source_topic, dest_topic)¶

Add a kafka parameters to your Application.

Parameters

source_topic (str) – source Kafka topic
dest_topic (str) – destination Kafka topic

Returns

Return type

hydrosdk.application.ApplicationBuilder

with_metadata(key, value)¶

Add a metadata value to your future Application.

Parameters

key (str) – string key under which value will be stored
value (str) – string value

Returns

Return type

hydrosdk.application.ApplicationBuilder

with_metadatas(metadata)¶

Add a metadata to your future Application.

Parameters: metadata (Dict[str, str]) – a dict containing metadata for your application
Returns
Return type: hydrosdk.application.ApplicationBuilder

with_stage(stage)¶

Add an ExecutionStage to your Application. See ExecutionStage for more information.

Parameters: stage (hydrosdk.application.ExecutionStage) –
Returns
Return type: hydrosdk.application.ApplicationBuilder

class hydrosdk.application.ApplicationStatus(value)¶

Bases: enum.Enum

An enumeration.