tour_of_artifacts.py

"""
# how to run this example on your local machine with uv (https://docs.astral.sh/uv/)
# !!! be sure to run against `prefect server start` or Prefect Cloud for a UI

GIST_URL="https://gist.githubusercontent.com/zzstoatzz/3eb9333625d465cb74381c5116be7aee/raw/556d8e1886e5725f383ff6e673681a42eec3f9cf/tour_of_artifacts.py"
curl -L $GIST_URL -o /tmp/artifacts.py
pip install -U uv
uv run --python 3.12 --with 'prefect>=3.0.0rc19' /tmp/artifacts.py
"""

import webbrowser
from typing import Any

from pydantic import BaseModel

from prefect import flow, task
from prefect.artifacts import create_markdown_artifact
from prefect.settings import PREFECT_UI_URL


class Data[T](BaseModel):
    value: T


@task(task_run_name="saving an {key} artifact")
def save(
    key: str,
    model: Data,
    description: str | None = None,
    add_a_duck: bool = True,
    show_me_now: bool = True,
):
    description = description or ""
    if add_a_duck:
        description += "\n\n![](https://random-d.uk/api/randomimg)"
    uuid = create_markdown_artifact(
        key=key,
        markdown=f"```json\n{model.model_dump_json(indent=2)}\n```",
        description=description or "No description provided.",
    )
    url = f"{PREFECT_UI_URL.value()}/artifacts/artifact/{uuid}"
    print("\n\n==========\n" f"\nSaved artifact {key!r}:\n\n{url}\n" "\n==========\n")
    if show_me_now:
        webbrowser.open(url)


@task(task_run_name="some_transform on a {data.__class__.__name__}")
def some_transform(data: Any) -> Any:
    # do something crazy
    return data.value


@flow(log_prints=True)
def main[T](data: Data[T]) -> Any:
    save("input", data, description="what our flow received", add_a_duck=False)
    results: list[T] = some_transform.map([data] * 10).result()
    save("output", Data[list[T]](value=results), description="what our flow produced")
    return results


if __name__ == "__main__":
    print(f"finished: {main(Data[int](value=42))}")

you can render markdown or table artifacts from runtime python variables (like a pd.DataFrame or BaseModel) and then store it under a key: str (input and output in this example)