bastienboutonnet/elementary-lineage

Elementary was built out of the need to effortlessly and immediately gain visibility into the data stack, starting with tracing the actual upstream & downstream dependencies in the data warehouse, without any implementation efforts, security risks or compromises on accuracy.

Data lineage made simple, reliable, and automated.

(Some features are still under construction 🚧)

⭐ If you like what we are building, support us with a ⭐

Quick start

pip install elementary-lineage

# The tool is named edl (Elementary Data Lineage),
# run it to validate the installation:
edl --help

We need a connection file in a simple YAML called profiles.yml. Here is a template to create a Snowflake / BigQuery one.
For further instructions go to our quickstart page.

If you use dbt, you can start right away by running this command with the path to your profiles.yml and the relevant profile name:

edl -d ~/profiles_dir -p <profile_name>

Documentation

Our full documentation is available here.

Features

Lineage visualization: Visual map of data flow and dependencies in the data warehouse.
Dataset status: Present data about freshness and volume on the lineage graph.
Accuracy: Reflects the actual state in the DWH based on logs.
Plug-and-play: No need for code changes.
Graph filters: Filter the graph by dataset, dates, direction, and depth.

Coming soon:

Dataset monitoring: Continuously collect and monitor datasets' reliability metrics.
Lineage history: Store data about lineage versions and changes.
Column level lineage: Add column-level granularity.
Full lineage: Integrate with downstream and upstream tools to create a full graph.

Our plan is to eventually build a data observability platform that is open, transparent and powered by the community. A solution that data teams could easily integrate into their workflows, detect data incidents and prevent them from even happening in the first place.