databricks-routing.mdc

---
description: Routing for any Databricks task -- CLI, auth, profiles, data exploration, Jobs/Lakeflow, Spark Declarative Pipelines (formerly DLT), Apps/AppKit, Asset Bundles/DABs, Model Serving, Lakebase/Postgres, Vector Search/RAG, Genie, and classic-to-serverless migration. Apply when the request is Databricks-related so the Databricks skills are loaded instead of ad hoc commands.
alwaysApply: false
---

This request is Databricks-related. Handle it through the Databricks skills
rather than ad hoc commands. Load the `databricks-core` skill first (the parent:
CLI, auth, profile selection, data exploration), then load the product skill
that matches the request:

- Jobs / Lakeflow / workflows -> databricks-jobs
- Pipelines / Lakeflow Spark Declarative Pipelines (formerly DLT) -> databricks-pipelines
- Apps / AppKit -> databricks-apps
- Asset Bundles / DABs / databricks.yml -> databricks-dabs
- Model Serving / endpoints -> databricks-model-serving
- Lakebase / Postgres -> databricks-lakebase
- Vector Search / RAG -> databricks-vector-search
- Classic-to-serverless migration -> databricks-serverless-migration
- Genie / natural-language data Q&A -> databricks-core (Genie CLI support is experimental)
- Agent Bricks / Knowledge Assistants / Genie Spaces / Multi-Agent Supervisor -> databricks-agent-bricks
- AI Functions (ai_query, ai_classify, ai_extract, ai_parse_document) -> databricks-ai-functions
- AI/BI dashboards -> databricks-aibi-dashboards
- Python data apps (Streamlit, Dash, Gradio, Flask, FastAPI) -> databricks-apps-python
- Databricks SQL warehouses -> databricks-dbsql
- Databricks documentation lookup (llms.txt) -> databricks-docs
- Executing code on Databricks compute -> databricks-execution-compute
- Apache Iceberg tables / UniForm / Iceberg REST Catalog -> databricks-iceberg
- Lakeflow Connect managed ingestion connectors -> databricks-lakeflow-connect
- Metric Views / governed metrics -> databricks-metric-views
- ML model training -> databricks-ml-training
- MLflow agent / GenAI evaluation -> databricks-mlflow-evaluation
- Python SDK / Databricks Connect / REST API -> databricks-python-sdk
- Spark Structured Streaming -> databricks-spark-structured-streaming
- Synthetic / test data generation (Faker) -> databricks-synthetic-data-gen
- Unity Catalog governance, access control & system tables -> databricks-unity-catalog
- Synthetic PDF generation for RAG -> databricks-unstructured-pdf-generation
- Zerobus streaming ingest -> databricks-zerobus-ingest

Then follow the skill's guidance (it drives the `databricks` CLI). If no product
skill fits, databricks-core alone is enough.