Installation & Setup

Requirements

Component	Version
Apache Airflow	≥ 3.1 (uses the `react_apps` + `fastapi_apps` plugin APIs; no Airflow 2.x support)
Python	≥ 3.10
Metadata DB	PostgreSQL recommended (DagSmith stores drafts in its own `dagsmith_*` tables)
File access	to deploy, the Airflow api-server needs write access to the DAG bundle directory; working on drafts needs no file access at all

Install

pip install airflow-dagsmith   # in the environment running the Airflow api-server

The plugin registers itself via the airflow.plugins entry point — don't copy anything into $AIRFLOW_HOME/plugins. After restarting the api-server, DagSmith appears in the Airflow navigation. Its dagsmith_* tables are created automatically on startup (or run python -m dagsmith db upgrade manually).

Enabling Deploy

For safety, writing .py files is off by default — without it DagSmith works in read-and-draft mode only. Enable it consciously in airflow.cfg:

[dagsmith]
deploy_enabled = True
# optional hardening / team split:
# editors   = data-platform-team     # who can create and edit drafts
# deployers = admin                  # who can write .py files
# admins    = admin                  # who manages teams
# git_commit = True                  # commit to the bundle checkout on deploy

(or the equivalent AIRFLOW__DAGSMITH__DEPLOY_ENABLED=True environment variable, etc.)

Security note: deploying DAGs from the UI is equivalent to executing arbitrary code in your Airflow environment — exactly like shipping a DAG file any other way. Enable deploy_enabled only with proper access control on the Airflow UI. Import validation executes the file's top level in an isolated, time-limited subprocess. See Security & Permissionsfor the full model.