// for the people who find out on Monday
Riftless reads every pull request against the data your company runs on. When a change would break a report, or quietly change what a number means, it says so on the PR before merge, and shows which reports depend on it.
We're letting in a few teams at a time. Work email, please.
Does orders.amount still include tax?
# services/orders/totals.py- amount = subtotal + tax_total+ amount = subtotal
“Order total in cents, including tax.”
Types unchanged. Every other check on this PR is green.
What we're building
A dropped column or a narrowed type is flagged in the pull request that causes it, along with the downstream models and dashboards that read that field.
The amount field stops including tax. Same column, same type, CI stays green, and finance finds out at month end. Riftless asks the question CI can't: does this change what the data means?
Findings are grouped into issues: one per field and kind of change, with an owner, a history and a resolve button. Fixed ones reopen if they come back.
You'll never put a contract on every table, and you don't need to. Most of the damage comes through a handful of boundaries: the producer repos behind the numbers leadership actually looks at. Protect those with a few lines of YAML in the producer's repo, then add more as new ones start to hurt. That works for the whole data team at a 50-person company, or for one team inside an enterprise. No six-month rollout, no governance council.
# riftless.yamlcontracts: - asset: orders owner: Checkout team source_paths: [services/orders/] fields: - name: amount type: INT64 meaning: Order total in cents, including tax