prefecthq/prefect

Prefect is a workflow orchestration framework that helps Python developers turn ordinary scripts into reliable, production-grade data pipelines. The problem it tackles is that real data work is more than just running code: jobs need to be scheduled, retried when they fail, monitored, and orchestrated alongside other jobs that depend on them. Prefect describes itself as the simplest way to elevate a script into a production workflow that can react to the world around it and recover from unexpected changes. The way it works is through two decorators you add to your Python functions: @flow marks a function as a workflow, and @task marks the smaller steps inside it. With just those annotations, Prefect can schedule the flow on a cron expression, retry tasks that fail, cache results, log what happened, and react to external events. Workflow activity is tracked and can be inspected through either a self-hosted Prefect server with a local UI, or the managed Prefect Cloud dashboard. The README walks through a small example that fetches a repository's star count using the httpx HTTP library, then shows how the same script becomes a scheduled deployment by adding a serve call with a cron pattern. You would reach for Prefect when you have data jobs, ETL pipelines, machine-learning training runs, scheduled reports, infrastructure tasks, that are outgrowing a plain cron entry and need proper observability, retries, dependency management, and the ability to be triggered by events rather than just time. The README mentions use across organisations from Progressive Insurance to Cash App for orchestrating data workflows at scale. The framework is written in Python and requires Python 3.10 or newer. It is installed with pip or uv, and a companion package called prefect-client offers a lighter-weight option for environments that only need to talk to a remote Prefect server or Prefect Cloud. The full README is longer than what was provided.