Agent tooling
rev turns labeled text into a local decision model, with data checks and a cutoff for sending uncertain answers to a person.
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
Its Python toolkit bundles training, score calibration and a System One-shaped server. It is an early, separate project from HCLTech's RevCT and TypeSafe's Jev. The README's performance comparisons are author claims, not independently reproduced results.
Begin with text whose correct decisions you already know. The data checks look for empty or conflicting examples before training. The toolkit then adjusts the point at which it asks a person, using the costs you assign to a wrong answer and to a review.