INSPECT BEFORE EXPORT
Your records.
Nothing silently changed.
Spot unusual records with local AI. Inspect the evidence, keep every original value, and export a traceable review report.
Try the live synthetic model below, then explore a six-record review workflow. The demo uses fictional USD records. Inputs stay in this page; no API, account or paid service is used by the demo.
01 / TRY LOCAL INFERENCE
A model, running in your browser.
100 Isolation Forest trees trained in scikit-learn on 40 fictional USD records: quantities of 1 and amounts from 80 to 158. This small model illustrates the workflow; it cannot represent your business.
Choose an example or enter plain decimal numbers, then run the model.
Conservative policy: flag only when the forest score is negative and a feature lies outside reference support. This can miss interior anomalies. A score is not an error probability.
Start by reviewing the complete report. No export is available yet.
This digest identifies the demo report, not a raw source-file hash or a signature.
| Row | ID | Currency | Amount | Review hints |
|---|
Complete report, original values and model metadata
Review the report to load this snapshot.