User Endpoints

After logic for your application is defined with Malevich flow, you can easily integrate it to your app without a need of deploying your script anywhere. Endpoints are dedicated URLs that invoke specified logic on our clouds. Explore the feature in action:

Define a logic

Let us start with a simple example. We will define a flow that takes some textual input and returns a rephrased version of it. Define a flow that utilizes OpenAI’s GPT-3 to rephrase a given text.

Install an app to utilize it in your code:

malevich install openai

And then define a flow:

from malevich import flow, collection, CoreInterpreter
from malevich.openai import prompt_completion

@flow
def rewrite_product_description():
    product_description = collection('product-description')
    return prompt_completion(
        product_description,
        model='gpt-3.5-turbo',
        openai_api_key=os.getenv('OPENAI_API_KEY'),
        system_prompt="""
            You are SEO specialist.
        """,
        user_prompt="""
            Rewrite a raw product
            description for it to inscrease amount of sales.

            {raw_description}
        """
    )

Publish an endpoint

Now, let’s publish an endpoint that will utilize the flow we have just defined.

logic = rewrite_product_description()
logic.interpret(CoreInterpreter(core_auth=('example', 'Welcome to Malevich!')))
endpoint = logic.publish()

Note

Endpoints are defined by a unique generated hash. They are invoked by sending a POST request to the following URL:

https://core.malevich.ai/api/v1/endpoints/run/<ENDPOINT_HASH>

Run an endpoint

They can be invoked from any application that can send HTTP requests as well from the object directly.

print('Endpoint hash': endpoint.hash)
print('Endpoint URL': endpoint.get_url())

print(endpoint.run(
    endpoint_override=EndpointOverride(
        cfg=UserConfig(rawMapCollections={
            'product-description': [
                {'raw_description': 'A beautiful ink pen with a comfortable grip and a smooth writing experience.'}
            ]
        })
    )
).results)

There is an example of running an endpoint from JavaScript:

const myHeaders = new Headers();
myHeaders.append("Content-Type", "application/json");

const raw = JSON.stringify({
"cfg": {
    "rawMapCollections": {
    "product-description": [
        {
        "raw_description": "A beautiful ink pen with a comfortable grip and a smooth writing experience."
        }
    ]
    }
}
});

const requestOptions = {
method: "POST",
headers: myHeaders,
body: raw,
redirect: "follow"
};

fetch("https://core.malevich.ai/api/v1/endpoints/run/fae7e96f288fcab9b0ff38ebdda57b87f83d66f1a2d2acd2ac39adaf54d2af91", requestOptions)
.then((response) => response.text())
.then((result) => console.log(result))
.catch((error) => console.error(error));

Update an endpoint

You can update already running endpoint. It can be some minor adjustments or a complete change of the logic. To update, you can run publish method with hash= argument. It will update the endpoint under specified hash.

@flow
def new_logic():
    ...


task = new_logic()
task.interpret(CoreInterpreter(core_auth=('example', 'Welcome to Malevich!')))
endpoint = task.publish(hash='...')