Skip to main content
Pre-release

This release is a pre-release and may not be stable for production use.

Flyte v2 SDK

The next-generation SDK for Flyte.

Publish Python Packages and Official Images

Quick start

  1. Run uv venv, and source .venv/bin/activate to create a new virtual environment.
  2. Install the latest version of the SDK by running the following:
uv pip install --no-cache --prerelease=allow --upgrade flyte
  1. Create the config and point it to your cluster by running the following:
flyte create config --endpoint <your-endpoint-url> --project <your-project> --domain <your-domain>

This will create a config.yaml file in the current directory which will be referenced ahead of any other config.yamls found in your system.

  1. Now you can run code with the CLI:
flyte run <path-to-your-script> <task-name>

Hello World Example

# hello_world.py

import flyte

env = flyte.TaskEnvironment(name="hello_world")


@env.task
async def say_hello(data: str) -> str:
    return f"Hello {data}"


@env.task
async def say_hello_nested(data: str) -> str:
    return await say_hello.override(resources=flyte.Resources(gpu="A100 80G:4")).execute(data)


if __name__ == "__main__":
    import asyncio

    # to run pure python - the SDK is not invoked at all
    asyncio.run(say_hello_nested("test"))

    # To run locally, but run through type system etc
    flyte.init()
    flyte.run(say_hello_nested, "World")

    # To run remote
    flyte.init(endpoint="dns:///localhost:8090", insecure=True)
    flyte.run(say_hello_nested, "World")
    # It is possible to switch local and remote, but keeping init to have and endpoint, but , changing context during run
    flyte.with_runcontext(mode="local").run(...)  # this will run locally only

    # To run remote with a config
    flyte.init_from_config("config.yaml")

CLI

All commands can be run from any root directory. For examples, it is not needed to have __init__.py in the directory. If you run from a directory, the code will automatically package and upload all modules that are imported. You can change the behavior by using --copy-style flag.

flyte run hello_world.py say_hello --data "World"

To follow the logs for the a0 action, you can use the --follow flag:

flyte run --follow hello_world.py say_hello --data "World"

Note that --follow has to be used with the run command.

Change copy style:

flyte run --copy-style all hello_world.py say_hello_nested --data "World"

Building Images

import flyte

env = flyte.TaskEnvironment(
    name="hello_world",
    image=flyte.Image.from_debian_base().with_apt_packages(...).with_pip_packages(...),
)

Deploying

flyte deploy hello_world.py say_hello_nested

Get information

Get all runs:

flyte get run

Get a specific run:

flyte get run "run-name"

Get all actions for a run:

flyte get actions "run-name"

Get a specific action for a run:

flyte get action "run-name" "action-name"

Get action logs:

flyte get logs "run-name" ["action-name"]

This defaults to root action if no action name is provided

Running workflows programmatically in Python

You can run any workflow programmatically within the script module using main:

if __name__ == "__main__":
    import flyte
    flyte.init()
    flyte.run(say_hello_nested, "World")

Running scripts with dependencies specified in metadata headers

You can also run a uv script with dependencies specified in metadata headers and build the task image automatically based on those dependencies:

# container_images.py

# /// script
# dependencies = [
#    "polars",
#    "flyte>=0.2.0b12"
# ]
# ///

import polars as pl

import flyte


env = flyte.TaskEnvironment(
    name="polars_image",
    image=flyte.Image.from_uv_script(
        __file__,
        name="flyte",
        registry="ghcr.io/<you-username>"
        arch=("linux/amd64", "linux/arm64"),
    ).with_apt_packages("ca-certificates"),
)


@env.task
async def create_dataframe() -> pl.DataFrame:
    return pl.DataFrame(
        {"name": ["Alice", "Bob", "Charlie"], "age": [25, 32, 37], "city": ["New York", "Paris", "Berlin"]}
    )


@env.task
async def print_dataframe(dataframe: pl.DataFrame):
    print(dataframe)


@env.task
async def workflow():
    df = await create_dataframe()
    await print_dataframe(df)


if __name__ == "__main__":
    flyte.init_from_config("config.yaml")
    run = flyte.run(workflow)
    print(run.name)
    print(run.url)
    run.wait(run)

When you execute

 uv run hello_world.py

uv will automatically update the local virtual environment with the dependencies specified in the metadata headers. Then, Flyte will build the task image using those dependencies and push it to the registry you specify. Flyte will then deploy the tasks to the cluster where the system will pull the image and run the tasks using it.

Metadata

Release files for flyte 0.2.0b23

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Built distribution (wheel)

Table of built distributions (wheels) for flyte 0.2.0b23
File Interpreter ABI Platform Actions
flyte-0.2.0b23-py3-none-any.whl Python 3 none any Details for flyte-0.2.0b23-py3-none-any.whl

Release files / flyte-0.2.0b23-py3-none-any.whl

Download URL flyte-0.2.0b23-py3-none-any.whl
Size 371.9 kB
Tags Python 3
SHA-256 checksum
How to use checksums
6e42bbeebac0fd1a3a56ee8182f38fb298b5d187f9f2b007ce423dd971f1c768
BLAKE2b-256 checksum
How to use checksums
379301a358df4eb25fcbc3dd5c5e01b27dd7f9bfdc06a9f5b983e306b8a32952
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.1.0 CPython/3.13.5

Security for flyte 0.2.0b23

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release
Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page