Skip to main content

Getting Started

SDP-META supports several deployment paths. All lead to the same outcome: a running Bronze/Silver pipeline driven by onboarding metadata.

Prefer video?

Watch the SDP-Meta Deep-Dive: Building Data Pipelines at Scale on Databricks — a walkthrough of the framework with a Demo.

Migrating from DLT-META?

If you are upgrading an existing dlt-meta installation after the v0.1.0 rename, follow the migration guide first. Your onboarding files do not need to change, but package names, CLI commands, Python imports, and some compatibility shims have a deprecation timeline.

Prerequisites (all paths)

  • Python 3.10–3.12 — Python 3.10, 3.11, or 3.12 recommended. Python 3.13+ has known PySpark compatibility issues.
  • Databricks CLI v0.213 or laterInstall guide.
databricks labs install sdp-meta
databricks auth login --host YOUR_WORKSPACE_URL
note

To use the MCP server, install its optional dependencies with pip install 'databricks-labs-sdp-meta[mcp]'. For local development without MCP, install databricks-labs-sdp-meta.

Deployment paths

Git-tracked configuration, explicit dev/prod environments, CI/CD-friendly. Right for production workloads and team settings.

See Declarative Automation Bundles.

Interactive CLI

Onboard and deploy with interactive prompts, no bundle YAML required. Good for exploring SDP-META or running a one-off pipeline.

See Interactive CLI.

Databricks App

Browser-based GUI for onboarding and managing pipelines. Ideal for non-engineers who need to manage pipelines without a terminal.

See Databricks App.

MCP Server

AI-assisted scaffolding via Claude Code, Claude Desktop, or Cursor. Drive SDP-META configuration through natural language.

See MCP Server.

Agent Skill

A portable agent skill that teaches any skill-aware AI agent the SDP-META workflow, so it can guide you from raw input data to a running pipeline. Complements the MCP Server.

See Agent Skill.