cosmos-dbt-fusion_skill

Overview

This skill guides implementation of a dbt Fusion project using Astronomer Cosmos (Cosmos ≥1.11.0). It focuses on Fusion-specific constraints for Snowflake and Databricks when running dbt with ExecutionMode.LOCAL. Use it to validate environment, install the Fusion binary, and configure Cosmos Project/Profile/Execution settings.

How this skill works

The skill inspects Cosmos version, verifies the dbt engine is Fusion (not Core), and confirms the warehouse is Snowflake or Databricks. It enforces LOCAL execution, requires a Fusion binary path in ExecutionConfig, and walks through RenderConfig, ProfileConfig, ProjectConfig, and DAG or TaskGroup assembly. Finally, it lists validation and runtime checks to confirm a successful run.

When to use it

• Running a dbt Fusion project with Astronomer Cosmos (Cosmos ≥1.11)
• Target warehouse is Snowflake or Databricks (public beta)
• You can install the Fusion binary into the Airflow runtime/image
• You must run dbt locally (ExecutionMode.LOCAL only)
• You need guidance wiring ProjectConfig, ProfileConfig, and ExecutionConfig together

Best practices

• Confirm Cosmos version is ≥1.11.0 before implementing Fusion
• Install the dbt Fusion binary into the Airflow image/runtime and validate with dbt --version
• Use ProfileMapping classes: SnowflakeUserPasswordProfileMapping or DatabricksTokenProfileMapping
• Select a single parsing strategy (dbt_manifest, dbt_ls, or automatic) based on project size and selector needs
• Keep secrets in Airflow connections or environment variables, never in plaintext

Example use cases

• Create a standalone DbtDag for a Fusion project with dbt_executable_path pointing to the installed binary
• Embed DbtTaskGroup inside an existing DAG to run Fusion models as part of a larger workflow
• Use dbt_manifest load mode for large projects to speed parsing when a manifest.json is available
• Use dbt_ls load mode when you need dbt-native selectors and the Fusion binary is accessible to the scheduler