serving_creator_func.py

import mlrun 
import random

def deploy_serving_functions(context, number_of_servings, number_of_models):
    project = context.get_project_object()
    
    # Get a list of the logged models
    models = project.list_models()
    
    # Create servings
    for serving_num in range(number_of_servings):
        serving_fn = project.set_function("hub://v2_model_server", f"serving-func{str(serving_num)}", kind="serving", image="mlrun/mlrun")
        serving_fn.apply(mlrun.auto_mount())

        # Add the models to the serving function's routing spec
        for model_num in range(number_of_models):
            
            # Get a random model from the logged models and add it to the serving
            random_model = models[random.randint(0, len(models) -1 )]
            # serving_fn.add_model(key=random_model.key, model_path=f"store://models/{project.name}/{random_model.key}:latest")
            serving_fn.add_model(key=random_model.key, model_path=random_model.uri)


        # (OPTIONAL) Create a tracking policy 
        # tracking_policy = {'default_batch_intervals':"", "default_batch_image":"", "stream_image":"", "default_controller_image":""}

        # Enable model monitoring (If you specified tracking_policy, pass it to 'tracking_policy' param)
        serving_fn.set_tracking()
        
        serving_fn.deploy()