Overview

Metaflow 2.11, released on January 15, 2024, introduces the Deployer to simplify production workflow deployment.

Main Features

Deployer

The Deployer lets you deploy Metaflow flows to orchestrators (Argo, Step Functions) directly from Python.

python 
from metaflow import FlowSpec, step

class TrainFlow(FlowSpec):
    @step
    def start(self):
        self.data = [1, 2, 3]
        self.next(self.train)

    @step
    def train(self):
        self.model = sum(self.data)
        self.next(self.end)

    @step
    def end(self):
        print(f'Result: {self.model}')

Programmatic deployment

Deployment is done via the Python API, without the CLI, making CI/CD integration easier.

python 
from metaflow.runner import Deployer

# Programmatic deployment
deployer = Deployer('train_flow.py')
deployment = deployer.argo_workflows().create()

# Trigger a run
run = deployment.trigger()
run.wait_for_completion()
print(f'Status: {run.status}')

Sources