Skip to content

Fabricks User Guide

Fabricks is a pragmatic framework to build Databricks Lakehouse pipelines using YAML for orchestration and SQL for transformations. It standardizes jobs, steps, schedules, CDC, and checks while keeping development SQL-first.

Steps Overview

Fabricks organizes your Lakehouse into clear layers. Each step has a dedicated reference with modes, options, and examples.

Bronze

Raw ingestion from source systems (files, streams, existing tables). Keep logic light; land data for downstream processing.

  • Typical modes: memory, append, register
  • Focus: lightweight parsing/landing; no business logic
  • Output: raw tables or temporary views

Read the full reference: Bronze Step

Silver

Standardize, clean, and enrich data; optionally apply CDC (SCD1/SCD2). Produces conformed datasets and convenience views.

  • Typical modes: memory, append, latest, update, combine
  • CDC: nocdc, scd1, scd2 with built-in helpers and views
  • Output: conformed tables and curated views

Read the full reference: Silver Step

Gold

Curated business models for analytics and reporting; dimensional or mart-style outputs. Can also invoke notebooks when needed.

  • Typical modes: memory, append, complete, update, invoke (notebooks)
  • Focus: dimensional models, marts, KPI-ready data
  • Output: business-consumption tables and views

Read the full reference: Gold Step

Where to Configure