ⓘ Read-only hosted preview!! no analysis or DataHub mutation is executed. Clone the repository to run the live workflow.

Provenance Logo
Production ML agent View repository Run locally with DataHub →

DataHub-grounded counterfactual testing

Test upstream data changes before they break production ML

Provenance follows DataHub lineage to the affected model, runs a bounded shadow experiment, applies deterministic safety policy, and writes verified evidence back.

  1. 1. Propose change →
  2. 2. Read DataHub context →
  3. 3. Run shadow inference →
  4. 4. Decide from policy →
  5. 5. Write evidence back →

Operational ML risk assessment

Analyze production change signals

transaction-velocity-surge RECORDED: WARNING

Account takeover velocity surge

A payment-fraud monitoring agent detected a sixfold transaction-velocity surge across eight percent of evaluated payments, consistent with an account-takeover-shaped stress condition.

Source
Monitoring agent
Column
transactions_last_10_minutes
Mutation
scale_numeric
Evaluation scope
8% of payments
View recorded analysis →
device-trust-range-change RECORDED: BLOCK

Device trust score contract change

The device-risk producer will change device_trust_score from a 0–100 range to a 0–1 range without updating downstream expectations.

Source
Monitoring agent
Column
device_trust_score
Mutation
scale_numeric
Evaluation scope
100% of payments
View recorded analysis →
new-digital-wallet-channel RECORDED: PASS

New digital-wallet channel rollout

Five percent of payments will begin using a new payment_channel category named digital_wallet.

Source
Monitoring agent
Column
payment_channel
Mutation
replace_category
Evaluation scope
5% of payments
View recorded analysis →

Optional AI planner

Analyze a custom change

Local mode only
Custom planning is intentionally unavailable in this hosted preview. Clone the repository to run validated Gemini planning against local DataHub OSS.