River API Client

This is an API client created for django-river-ml that is intended to make it easy to interact with an online ML server providing river models (learning, predicting, etc.). It currently does not provide a terminal or command line client and is intended to be used from Python, but if there is a good use case for a command line set of interactions this can be added.

Quick Start

Given that you have a server running that implements the same space as django-river-ml, you can do the following. Note that if your server requires authentication, you can generate a token and export to:

export RIVER_ML_USER=dinosaur
export RIVER_ML_TOKEN=xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

And then do the following example.

from river import datasets
from river import linear_model
from river import preprocessing

from riverapi.main import Client


def main():

    # This is the default, just to show how to customize
    cli = Client("http://localhost:8000")

    # Basic server info (usually no auth required)
    cli.info()

    # Upload a model
    model = preprocessing.StandardScaler() | linear_model.LinearRegression()

    # Save the model name for other endpoint interaction
    model_name = cli.upload_model(model, "regression")
    print("Created model %s" % model_name)

    # Train on some data
    for x, y in datasets.TrumpApproval().take(100):
        cli.learn(model_name, x=x, y=y)

    # Get the model (this is a json representation)
    model_json = cli.get_model_json(model_name)
    model_json

    # Saves to model-name>.pkl in pwd unless you provide a second arg, dest
    cli.download_model(model_name)

    # Make predictions
    for x, y in datasets.TrumpApproval().take(10):
        res = cli.predict(model_name, x=x)
        print(res)

    # By default the server will generate an identifier on predict that you can
    # later use to label it. Let's do that for the last predict call!
    identifier = res['identifier']

    # Let's pretend we now have a label Y for the data we didn't before.
    # The identifier is going to allow the server to find the features,
    # x, and we just need to do:
    res = cli.label(label=y, identifier=identifier, model_name=model_name)
    print(res)
    # Note that model_name is cached too, and we provide it here just 
    # to ensure the identifier is correctly associated.

    # Get stats and metrics for the model
    cli.stats(model_name)
    cli.metrics(model_name)

    # Get all models
    print(cli.models())

    # Stream events
    for event in cli.stream_events():
        print(event)

    # Stream metrics
    for event in cli.stream_metrics():
        print(event)

    # Delete the model
    cli.delete_model(model_name)

if __name__ == "__main__":
    main()