Getting started with Flyte SDK for Python

This guide will help you get started with flytekit, the Python SDK for Flyte. Flyte is a workflow automation platform that helps you build reliable, scalable, and maintainable data and ML workflows.

Prerequisites

Before you begin, ensure you have the following installed:

Python: flytekit requires Python versions >=3.9,<3.13. We recommend using Python 3.9, 3.10, 3.11, or 3.12.
Operating System: flytekit is primarily supported on Linux.
System Dependencies: You might need build-essential, vim, libmagic1, and git for certain functionalities or if building from source.

Install

We recommend using uv for a fast and efficient installation.

Install uv: If you don't have uv installed, you can get it via pip:
```
pip install uv
```
Create a Virtual Environment: It's good practice to install Python packages in a virtual environment to avoid conflicts with system-wide packages.
```
python -m venv .venv
source .venv/bin/activate
```
Install Flytekit: Use uv to install flytekit. You can also include optional dependencies like agent or connector if needed.
```
uv pip install flytekit[agent,connector]
```

Hello World

Let's create a simple Flyte workflow and run it locally.

Create a Python file (e.g., my_flyte_app.py):

# my_flyte_app.py
from flytekit import task, workflow

@task
def greet(name: str) -> str:
    """A simple task that returns a greeting."""
    return f"Hello, {name}!"

@workflow
def my_workflow(name: str) -> str:
    """A workflow that calls the greet task."""
    greeting = greet(name=name)
    return greeting

if __name__ == "__main__":
    # Run the workflow locally using Python directly
    print(f"Running my_workflow locally: {my_workflow(name='Flyte User')}")

    # You can also run it using the pyflyte CLI for a more complete local simulation
    # To do this, save the file and then run from your terminal:
    # pyflyte run my_flyte_app.py my_workflow --name "Flyte CLI User"

Run the workflow locally:

python my_flyte_app.py

You should see output similar to:

Running my_workflow locally: Hello, Flyte User!

Verify Installation

To verify that flytekit and its CLI tools are correctly installed, check the version of pyflyte:

pyflyte --version

You should see the installed flytekit version.

Other Install Options

Using `pip`

If you prefer pip over uv, you can install flytekit as follows:

python -m venv .venv
source .venv/bin/activate
pip install flytekit[agent,connector]

Using Docker

You can also use Docker to run flytekit in a containerized environment. The project provides Dockerfile examples.

To build a Docker image:

docker build -f Dockerfile.dev --build-arg PYTHON_VERSION=3.10 -t localhost:30000/flytekittest:someversion .

Then you can run your Flyte workflows within this image. For example, to run a workflow with a specific image:

pyflyte run --image localhost:30000/flytekittest:someversion my_flyte_app.py my_workflow --name "Docker User"

Troubleshooting

Python Version Mismatch: If you encounter issues, ensure your active Python environment matches the required versions (>=3.9,<3.13). You can check your Python version with python --version.
Missing Dependencies: If you see ModuleNotFoundError or similar, ensure all necessary dependencies are installed. Re-run uv pip install flytekit[agent,connector] or pip install flytekit[agent,connector] to catch any missing packages.
Rich Tracebacks: If you need more detailed error messages for debugging, you can enable rich tracebacks by setting an environment variable:
```
export FLYTE_SDK_RICH_TRACEBACKS=1
```
PYTHONPATH Issues: If pyflyte commands or imports fail, check your PYTHONPATH environment variable. It might need to include your project root or the flytekit installation path.
```
echo $PYTHONPATH
# Example: export PYTHONPATH="/path/to/your/project:$PYTHONPATH"
```

Next Steps

Now that you have flytekit installed and can run a basic workflow, explore the official Flyte documentation and examples to learn more about:

Defining more complex tasks and workflows.
Running workflows on a remote Flyte cluster.
Using advanced features like caching, scheduling, and data types.
Interacting with the Flyte CLI (pyflyte and flyte-cli).

Prerequisites​

Install​

Hello World​

Verify Installation​

Other Install Options​

Using pip​

Using Docker​

Troubleshooting​

Next Steps​