domino-flows

Domino Flows Skill

This skill provides comprehensive knowledge for orchestrating ML workflows using Domino Flows, built on the Flyte platform.

Key Concepts

What are Domino Flows?

Domino Flows enable:

• DAG-based orchestration: Define workflows as directed acyclic graphs
• Typed interfaces: Strong typing for inputs and outputs
• Heterogeneous environments: Different environments per task
• Automatic lineage: Track data and model provenance
• Reproducibility: Version-controlled workflows
• Scalability: Distributed execution across compute resources

Core Components

Component	Description
Task	Single unit of work (runs as a Domino Job)
Workflow	DAG connecting tasks
Artifact	Typed input/output passed between tasks
Launch Plan	Configured workflow execution

Related Documentation

• FLOW-BASICS.md - DAG concepts, task definitions
• EXAMPLES.md - Common flow patterns

Quick Start

⚠️ Critical: Domino Flows does NOT support native Flyte @task decorators. Tasks must use DominoJobTask + DominoJobConfig. Only @workflow is unchanged.

Basic Flow

Each task runs as a Domino Job. Stage scripts read from /workflow/inputs/<name> and write to /workflow/outputs/o0. Pass PYTHONPATH=/mnt/code in the command.

from flytekit import workflow
from flytekitplugins.domino.task import DominoJobConfig, DominoJobTask

preprocess_task = DominoJobTask(
    name="Preprocess Data",
    domino_job_config=DominoJobConfig(
        Command="bash -c 'PYTHONPATH=/mnt/code python /mnt/code/stages/preprocess.py'",
    ),
    inputs={"input_path": str},
    outputs={"o0": str},
    use_latest=True,
)

train_task = DominoJobTask(
    name="Train Model",
    domino_job_config=DominoJobConfig(
        Command="bash -c 'PYTHONPATH=/mnt/code python /mnt/code/stages/train.py'",
    ),
    inputs={"preprocess_output": str},
    outputs={"o0": str},
    use_latest=True,
)

@workflow
def training_pipeline(input_path: str = "/mnt/data/raw.csv") -> str:
    preprocess_output = preprocess_task(input_path=input_path)
    result = train_task(preprocess_output=preprocess_output)
    return result