agi-app-global-dag 2026.5.22

AGILAB cross-app DAG preview showing artifact handoffs between demo projects

agi-app-global-dag

agi-app-global-dag publishes the global_dag_project AGILAB app as a self-contained PyPI payload. It is a read-only workflow-contract example rather than a domain worker benchmark.

Purpose

Use this package to understand how AGILAB can connect several app projects with explicit artifact handoffs. The bundled DAG shows a flight stage producing a summary artifact that a weather stage can consume.

Installed Project

The distribution name is agi-app-global-dag; the AGILAB project name is global_dag_project. The package exposes both global_dag and global_dag_project through the agilab.apps entry point group, so AgiEnv(app="global_dag_project") resolves the project without a monorepo checkout.

Install

pip install agi-app-global-dag

Most users get this package through agi-apps, agilab[ui], or agilab[examples]; direct installation is useful when validating one app package in isolation.

Run In AGILAB

Select global_dag_project, open WORKFLOW, choose the multi-app DAG template, and inspect the runner-state preview. This package is primarily for workflow review; use flight_telemetry_project or weather_forecast_project when you want to execute concrete worker code.

Expected Inputs

The package ships a DAG template that names the producer and consumer projects. No private data, cluster, service mode, or live external service is required for the default preview.

Expected Outputs

The preview writes a runner-state JSON file under the AGILAB execution logs and renders the runnable versus blocked stages. It does not create a domain reducer artifact because it demonstrates cross-app orchestration, not a worker merge.

Change One Thing

Edit a handoff name in a copied DAG template and reload the preview. The blocked stage should make the missing artifact contract obvious before any downstream execution is attempted.

Scope

This package teaches global DAG contracts. It is not a scheduler replacement, an Airflow/Kubeflow clone, or a production workflow governance layer.