What it is
dbt-adobe-events is a plan for a free, open-source building block for data teams. dbt is a popular tool for turning raw warehouse data into tidy tables using version-controlled SQL. Adobe's event data arrives in warehouses in awkward shapes: deeply nested JSON, mirror tables, or files with about a thousand columns. This package would read any of those and produce a standard set of clean, documented, tested tables, so each team stops writing the same cleanup by hand.
How it works
As planned:
- Install the package into an existing dbt project.
- Set a few variables to say where the raw data lives and which format it is in.
- Run
dbt build. - The package flattens and types the raw events, removes duplicates, groups hits into sessions, labels event types and assigns marketing channels with a rules engine.
- Out come ready tables: events, page views, sessions, users, conversions, order items, and a daily table for comparing two Adobe products' numbers.
What exists today
- A detailed requirements document (about 2,500 words, 14 sections) with a definition of done for version 1.0
- A build plan for an AI agent team
- A list of open questions
What isn't there yet
- No code and no repository
- No sample data or test fixtures yet
- Warehouse support (Snowflake, BigQuery, Databricks) is planned, none tested
- No documentation site
- The channel rules engine is described but not designed in detail
Under the hood
- Planned as dbt SQL and Jinja models with macros as the public interface
- One shared staging layer for both the modern and the classic Adobe data formats
- dbt tests and contracts on every output table, plus synthetic fixtures
- Settings through dbt variables, so users never edit SQL
- Clean-room rule: only public documentation and invented data; Apache 2.0 licence