gust-py 0.1.0

Python runtime and authoring SDK for Gust DAGs, including parser and task invoker

gust_py

Python runtime/package for Gust DAG authoring, parsing, and task execution.

What This Project Provides

• A Python authoring API:
  • Dag
  • @task(...)
  • log(...)
  • runtime calls: get_task_by_name_run(...), get_secret_by_name(...)
• A parser that converts Python DAG files into a stable JSON spec.
• A task invoker to execute one task from a DAG file.
• A run-completion invoker to execute a DAG callback (run done).

Requirements

• Python >=3.10
• uv

Install

uv sync --dev

DAG Authoring API

from gust import Dag, task, get_task_by_name_run, get_secret_by_name, log


class MyDag(Dag):
    def __init__(self):
        super().__init__(schedule="daily", on_finished_callback="done_my_run")

    @task(deps=("prepare",), save=True)
    def hello(self, ctx):
        run_id = ctx["run_id"]
        prepare_run = get_task_by_name_run("prepare", run_id)
        secret = get_secret_by_name("SUPER_SECRET")
        log(f"prepare={prepare_run}, secret_loaded={bool(secret)}")
        return {"ok": True}

    def done_my_run(self, status, run):
        log(f"run {run['run_id']} finished with status={status}", level="info")

Runtime Messaging Protocol

gust runtime messages use a framed protocol on stdio:

1. 4-byte big-endian frame length
2. UTF-8 JSON payload

Helpers:

• send_frame(payload) writes one frame to stdout.
• _read_frame() reads one frame from stdin.

Call helpers emit messages and wait for a response:

• get_task_by_name_run(name, run_id) emits:
  • {"type": "call", "op": "get_task_by_name_run", "run_id": run_id, "name": name}
• get_secret_by_name(name) emits:
  • {"type": "call", "op": "get_secret_by_name", "name": name}

Logging helper emits:

• log(msg, level="info"):
  • {"type": "log", "level": level, "msg": msg}

CLI

The package exposes a gust CLI.

Parse DAGs

gust parse --file path/to/dag_file.py

Output is JSON (printed to stdout), one list entry per discovered DAG class.

Run a Task

gust task run \
  --file path/to/dag_file.py \
  --dag MyDag \
  --task hello \
  --ctx-json '{"run_id":"123"}'

Output is a framed JSON payload:

• Success:
  • {"type":"result","ok":true,"data":{"value": ...}}
• Error:
  • {"type":"result","ok":false,"error":{"message":"...","type":"...","details":"..."}}

Run Completion Callback

gust run done \
  --file path/to/dag_file.py \
  --dag MyDag \
  --fn-name done_my_run \
  --status ok \
  --run-id 123

Behavior:

• On success: no payload output.
• On failure: process exits with code 1.

Development

Run tests:

uv run pytest -q

Run lint:

uv run ruff check .