Skip to content

About

Govern and ship Databricks Genie agents to production as code: Unity Catalog classifies sensitive data, GenieRails enforces it, and a coverage gate blocks the release until it's covered.

Topics

Resources

Security policy

Stars

7 stars

Watchers

0 watching

Forks

Latest commit

 

History

1,756 Commits

Folders and files

Repository files navigation

GenieRails

GenieRails

GenieRails lets you give business users a Genie agent over your data without exposing sensitive columns.

Databricks Data Classification finds the sensitive columns, like emails and card numbers. GenieRails turns what it finds into masks and access rules, tests them in dev, then releases the same tested rules to prod. It also controls who can use each Genie agent, and only opens that access once the masks are proven to work. You don't write any Terraform.

Why use it

  • No access before protection. Nobody gets new access to an agent or its data until every sensitive column Data Classification finds in its tables has a mask. Only the groups you choose see raw values.
  • You prove it in dev first. Dev queries the data as each test group and checks that the masks hide it.
  • Every change is reviewed and recorded. Prod only takes changes that were committed and tested, you approve each step, and you can save an audit record of what's live.
  • A few commands replace hand-written Terraform and SQL.
  • Works on AWS and Azure.

How it works

  1. Classify. Databricks scans your tables and tags each sensitive column.
  2. Generate rules. GenieRails reads those tags and writes a mask for each sensitive column. It also writes who can read each table and who can use each Genie agent, using your existing groups.
  3. Rehearse in dev. It applies the rules to dev and queries as each group to prove the masks work.
  4. Release to prod. You copy the tested rules to prod and prod scans its own data. Then GenieRails applies the rules one approved step at a time. Masks go on before anyone gets access to the tables or the agent.

Get started

Start with the Dev-to-Prod walkthrough. It takes about 30 minutes and ships one Genie agent from dev to prod. The main commands, run from aws/ or azure/:

make setup ENV=dev
make generate ENV=dev
make rehearse ENV=dev
make setup ENV=prod
make promote-to ENV=prod
make release ENV=prod

Stuck? Run make next ENV=<env>. It prints the next command to run.

Other ways to use it

  • Govern tables only. Your data team owns the masks and table access for everyone, Genie included: govern tables only.
  • One team doing both. Change agents and table rules in one pull request: one team.
  • Pipelines and a deployment machine. Run checks in CI and release prod from one machine: pipelines and the deployment machine.

Docs

About

Govern and ship Databricks Genie agents to production as code: Unity Catalog classifies sensitive data, GenieRails enforces it, and a coverage gate blocks the release until it's covered.

Topics

Resources

Security policy

Stars

7 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages