kazeflow

kazeflow is a lightweight task-flow library for ordinary Python scripts. Turn functions into assets, inspect their dependency plan, explicitly approve execution, and review the result—without a daemon, service, or required runtime dependency.

Install

pip install kazeflow

For the optional Rich progress display, quote the extra so shells such as zsh do not expand the brackets:

pip install "kazeflow[tui]"

Your first reviewed flow

# daily.py
from kazeflow import Flow, asset


@asset
def fetch() -> str:
    return "report input"


@asset
def publish(fetch: str) -> None:
    print(fetch)


flow = Flow(["publish"])

kazeflow plan daily.py
kazeflow run daily.py       # review the preflight, then answer y
kazeflow run daily.py --yes # explicit approval for CI or a pipe

The complete user documentation lives on GitHub Pages:

• Get started
• Python API reference
• CLI and graph output
• Partitions
• Results, SQLite storage, and history

Trust boundary

kazeflow loads the Python entry before it can inspect a flow, so module-level code and imports can run. plan does not invoke decorated asset bodies after loading, but it is not a sandbox or a safety proof. Read and trust a script before loading it, especially when it was generated by another person or an AI.

Project information

Maintainer documentation stays in the repository: project goals, roadmap, and release notes. The runnable examples are in examples/.

License: MIT.