Cloud

One of the main features of nextmv, the general Nextmv Python SDK, is the ability to interact with the Nextmv Cloud. This allows you to run, and experiment with your decision applications in Cloud.

Before starting:

1. Sign up for a Nextmv account.
2. Get your API key. Go to Team > API Key.

Make sure that you have your API key set as an environment variable:

export NEXTMV_API_KEY="<YOUR-API-KEY>"

Once you have your API key set up, you can create a cloud Client:

import os
import nextmv.cloud

client = nextmv.cloud.Client(api_key=os.getenv("NEXTMV_API_KEY"))

Application

An application represents a decision model that you can run in the Nextmv Cloud. The Application class defines the methods to interact with an app. You can create an app in multiple ways:

• Through the Python SDK (see next header).
• Through the Nextmv CLI.
• Through the Nextmv Cloud Console.

Once you have an app, and can identify its ID, you can interact with it.

Create an app

You can create an app by providing the Client instance and the app ID:

import os

import nextmv.cloud

client = nextmv.cloud.Client(api_key=os.getenv("NEXTMV_API_KEY"))
application = nextmv.cloud.Application.new(
    client=client,
    name="My Application",
    id="<YOUR_APP_ID>",
    description="My first application",
)

Delete an app

You can delete an application by using its ID.

import os

import nextmv.cloud

client = nextmv.cloud.Client(api_key=os.getenv("NEXTMV_API_KEY"))
application = nextmv.cloud.Application(client=client, id="<YOUR_APP_ID>")
application.delete()

Push to an app

There are two strategies to push an application to the Nextmv Cloud:

1. Specifying app_dir, which is the path to an app’s root directory. This acts as an external strategy, where the app is composed of files in a directory and those apps are packaged and pushed to Nextmv Cloud. This is language-agnostic and can work for an app written in any language.

   Place the following script in the root of your app directory and run it to push your app to the Nextmv Cloud. This is equivalent to using the Nextmv CLI and running nextmv app push.

   • Python

   import os
   
   from nextmv import cloud
   
   client = cloud.Client(api_key=os.getenv("NEXTMV_API_KEY"))
   app = cloud.Application(client=client, id="<YOUR_APP_ID>")
   app.push()  # Use verbose=True for step-by-step output.
   

   Copy

2. Specifying a model and model_configuration. This acts as an internal (or Python-native) strategy called "Apps from Models", where an app is created from a model. The model is encoded, some dependencies and accompanying files are packaged, and the app is pushed to Nextmv Cloud.

   To push a nextmv.Model to Nextmv Cloud, you need optional dependencies. You can install them by running:

   • Bash

   pip install "nextmv[all]"
   

   Copy

   Once all the optional dependencies are installed, you can push the app to Nextmv Cloud.

   • Python

   import os
   
   from nextmv import cloud
   
   class CustomDecisionModel(nextmv.Model):
       def solve(self, input: nextmv.Input) -> nextmv.Output:
           """Implement the logic to solve the decision problem here."""
           pass
   
   client = cloud.Client(api_key=os.getenv("NEXTMV_API_KEY"))
   app = cloud.Application(client=client, id="<YOUR_APP_ID>")
   
   model = CustomDecisionModel()
   options = nextmv.Options() # Define the options here.
   model_configuration = nextmv.ModelConfiguration(
       name="custom_decision_model",
       requirements=[ # Acts as a requirements.txt file.
           "nextmv==0.14.0",
           # Add any other dependencies here.
       ],
       options=options,
   )
   manifest = nextmv.cloud.Manifest.from_model_configuration(model_configuration)
   
   app.push( # Use verbose=True for step-by-step output.
       manifest=manifest,
       model=model,
       model_configuration=model_configuration,
   )
   

   Copy

Run an app

Make a run and get the results.

import os

from nextmv import cloud

input = {"foo": "bar"}

client = cloud.Client(api_key=os.getenv("NEXTMV_API_KEY"))
app = cloud.Application(client=client, id="<YOUR_APP_ID>")
result = app.new_run_with_result(
    input=input,
    instance_id="latest",
    run_options={"solve.duration": "1s"},
    polling_options=cloud.PollingOptions(),  # Customize the polling options.
)
print(result.to_dict())

Do more

Make sure to check out the latest version of nextmv to see all the available features of the Application. Here are some of the methods you can use that were not covered in this guide:

• acceptance_test: gets an acceptance test.
• batch_experiment: gets a batch experiment.
• cancel_run: cancels a run.
• delete_batch_experiment: deletes a batch experiment.
• delete_acceptance_test: deletes an acceptance test.
• input_set: gets an input set.
• instance: gets an instance.
• list_acceptance_tests: lists acceptance tests.
• list_batch_experiments: lists batch experiments.
• list_input_sets: lists input sets.
• list_instances: lists instances.
• list_versions: lists versions.
• new_acceptance_test: creates a new acceptance test.
• new_acceptance_test_with_result: creates a new acceptance test and gets the result.
• new_batch_experiment: creates a new batch experiment.
• new_input_set: creates a new input set.
• new_version: creates a new version.
• new_instance: creates a new instance.
• run_input: gets the input of a run.
• run_logs: gets the logs of a run.
• run_metadata: gets the metadata of a run.
• run_result: gets the result of a run.
• run_result_with_polling: gets the result of a run with polling.
• update_instance: updates an instance.
• upload_large_input: uploads a large input.
• upload_url: gets the URL for uploading.
• version: gets a version.