Skip to main content

Getting started

Flyte is a highly extensible Python-native platform for authoring and executing distributed workflows. It allows you to define tasks with specific resource requirements and execution environments that can run locally or on a remote cluster.

Prerequisites​

  • Python: Version 3.10 or higher.
  • Docker: A running Local Runtime is required to build images and wheels.
  • uv: The recommended package manager for this project.

Install​

Clone the repository and use uv to set up the development environment and build the package:

git clone https://github.com/flyteorg/flyte
cd flyte
uv sync
make dist

This installs the package in editable mode and builds a wheel so that the default Image() can use your local changes.

Hello world​

Create a file named hello.py to define a simple task and environment:

import flyte
import asyncio

# Define an execution environment
env = flyte.TaskEnvironment(name="getting-started")

@env.task
async def greet(name: str) -> str:
    """A simple Flyte task."""
    return f"Hello, {name}!"

async def main():
    # Execute the task locally
    result = await greet(name="Developer")
    print(result)

if __name__ == "__main__":
    asyncio.run(main())

Run the script using uv:

uv run python hello.py

Verify​

To ensure your environment is correctly configured and the core components are functioning, run the unit tests:

make unit_test

You can also verify code quality and type safety:

make fmt
make mypy

Next steps​

  • Explore the CLI: Use the flyte command-line tool to interact with the system. 
    uv run flyte --help
  • Authoring Tasks: Learn about Task authoring experience to define complex logic.
  • Plugins: Extend functionality by exploring available Distributed Compute Plugins for Spark, Ray, or BigQuery.
  • Serving: Use flyte.serve() to deploy long-running applications like FastAPI.

Troubleshooting​

Docker Daemon​

If make dist fails, ensure your Docker daemon is running. Flyte requires Docker to build the local wheel into a container image for consistent execution.

Module Imports​

If you encounter import errors, ensure you are running commands via uv run to use the synchronized virtual environment. Avoid importing internal modules (prefixed with _); always use the public API exported in flyte.*.