Airflow Sketcher

Airflow Sketcher is an Airflow 3.1+ plugin to build Airflow DAGs from Excelidraw drawings.

What is included

• src/airflow_sketcher/dag_factory.py: Excalidraw to Airflow DAG parser/importer.
• src/airflow_sketcher/plugins/builder.py: Airflow plugin exposing the Excalidraw builder UI.
• src/airflow_sketcher/sitecustomize.py: optional startup hook for importer registration.
• examples/airflow/: thin wrapper files that can be mounted into an Airflow dags/ and plugins/ folder.
• tests/: unit tests and real .excalidraw fixtures.

Local development

With uv (recommended)

uv sync --all-extras
uv run python -m unittest discover -s tests

With traditional venv

python3 -m venv .venv
source .venv/bin/activate
pip install -e '.[dev]'
python3 -m unittest discover -s tests

Plugin functionality

The AirflowSketcherPlugin adds an Excalidraw-based DAG builder page to the Airflow UI under /airflow-sketcher.

Key capabilities:

• Renders an interactive Excalidraw canvas in Airflow for creating and editing DAG diagrams.
• Saves diagrams as .excalidraw files in the Airflow DAGS_FOLDER.
• Enforces safe filenames (letters, numbers, ., _, -) and prevents path traversal.
• Auto-appends the .excalidraw extension when omitted.
• Loads existing scenes and preserves selected visual app state (for example grid and viewport settings).
• Derives DAG metadata (such as dag_id and schedule) from dag: text blocks in the drawing.
• Resolves a diagram for an existing DAG by checking DAG params (excalidraw_source_file) first, then scanning .excalidraw files.
• Registers the Excalidraw importer at startup so saved diagrams can be translated into Airflow DAG objects.

How to open the plugin

You can open the builder in two ways:

1. Dedicated icon/view: Open the Airflow top navigation entry named Airflow Sketcher (with the plugin icon). This opens the builder home at /airflow-sketcher/.
2. From a DAG created by this plugin: Open DAG details and click the Builder external view. It opens /airflow-sketcher/?dag_id=<your_dag_id> and tries to load the corresponding .excalidraw source file.

First DAG

Create a very simple DAG with two empty tasks in the builder UI:

1. Open Airflow and navigate to /airflow-sketcher.
2. Add one text element for DAG metadata with this content:

dag:
dag_id=first_two_tasks
schedule=@daily

3. Draw two rectangles (each rectangle is a task).
4. Add text inside the first rectangle:

task_a

5. Add text inside the second rectangle:

task_b

6. Draw an arrow from task_a to task_b.
7. Save as first_two_tasks.excalidraw.

Notes:

• You do not need to set class for this example. If omitted, tasks default to EmptyOperator.
• If no dag_id is provided in the dag: block, the file name (without extension) is used as the DAG ID.

Examples

1. The simplest DAG (two empty tasks)

Use these text blocks:

dag:
dag_id=1_simple_two_tasks
schedule=@daily

Task rectangle 1:

task_a

Task rectangle 2:

task_b

Then draw one arrow: task_a -> task_b.

In the Airflow UI, it should look like:

2. DAG with BashOperator and PythonOperator

Use these text blocks:

dag:
dag_id=2_bash_and_python
schedule=@daily

vars:
def greet(**kwargs):
   print("hello from python task")

Task rectangle 1:

bash_task
class = operators.bash.BashOperator
bash_command = 'echo "hello from bash"'

Task rectangle 2:

python_task
class = operators.python.PythonOperator
python_callable = greet

Then draw one arrow: bash_task -> python_task.

In the Airflow UI, it should look like:

Example files

Ready-to-use .excalidraw files are available in examples/excalidraw/:

1. examples/excalidraw/1_simple_two_tasks.excalidraw
2. examples/excalidraw/2_bash_and_python.excalidraw

Supported sections and task attributes

The parser reads specific text blocks from the Excalidraw scene.

dag: section

Use a text element starting with dag: and put one key=value pair per line.

Supported keys:

• dag_id: explicit DAG identifier.
• schedule: DAG schedule string (for example @daily or cron expression).

If dag_id is omitted, the importer uses the .excalidraw filename (without extension). If schedule is omitted, it defaults to None.

Example:

dag:
dag_id=my_first_dag
schedule=@daily

imports: section

Use a text element starting with imports: to declare Python imports used by tasks.

• Only Python import statements are supported in this block.
• import ... and from ... import ... are both supported.
• Imported names are available when evaluating task attributes (for example class, python_callable, op_args, op_kwargs).
• You can use multiple imports: blocks in the same diagram.

Default import behavior:

• No custom project modules are imported automatically. Add them in imports: when needed.
• Python builtins are available when evaluating attribute values.
• For the class attribute only, the operators. prefix is treated as shorthand for airflow.providers.standard.operators..
  • Example: class = operators.bash.BashOperator resolves to airflow.providers.standard.operators.bash.BashOperator.
• If class is omitted, the task defaults to airflow.providers.standard.operators.empty.EmptyOperator.

Example:

imports:
from airflow.providers.standard.operators.bash import BashOperator
import shared_funcs

vars: section

Use a text element starting with vars: to define Python variables/functions reused in task attributes.

• Contents are executed as Python code.
• Names defined here are available when evaluating task attributes (for example python_callable=my_func).

Example:

vars:
def my_func(**kwargs):
      print("hello")

Task text format (inside each rectangle)

Each task rectangle is configured by its associated text content.

Supported attributes:

• First non-empty line: task id (required), for example task_a.
• class (optional): operator class path or imported symbol.
  • Default when omitted: operators.empty.EmptyOperator.
• Any other key=value lines are passed as operator keyword arguments.

Attribute values support:

• Python literals (strings, numbers, booleans, lists, dicts).
• Names imported or defined in vars:.
• Multi-line Python snippets. If a snippet is statements plus a final expression, that expression becomes the value. You can also assign to result and that value is used.

Example:

task_python
class = operators.python.PythonOperator
python_callable = my_func
op_kwargs = {"name": "world"}

Airflow integration

Install the package into your Airflow image or environment, then expose the thin wrapper files from examples/airflow/ inside the Airflow plugins/ and dags/ directories.

Installation

From PyPI:

pip install airflow-sketcher

From sources (editable/development mode):

# With uv
uv pip install -e .

# With pip
pip install -e .

Typical setup

1. Install the package (see above).
2. If you are not installing with pip, copy the whole airflow_sketcher package directory into Airflow's plugins/ folder (for example as plugins/airflow_sketcher/).
3. Copy examples/airflow/plugins/excalidraw_builder_plugin.py into Airflow's plugins/ folder.
4. Copy examples/airflow/plugins/sitecustomize.py into Airflow's plugins/ folder if you want automatic importer registration at process startup.
5. Copy examples/airflow/dags/airflow_dag_generator.py into Airflow's dags/ folder if you want a Python entrypoint that re-exports the factory symbols.

Testing strategy

The unit tests install lightweight Airflow stubs when Apache Airflow is not installed. That keeps parser and factory coverage runnable in a plain Python environment while still allowing the same package to run under real Airflow.