Levanto

Levanto, built by Levanto Labs, offers Sage, which it describes as the first decision model. It lets a running program ask questions about meaning and continue based on the structured answer, rather than relying only on hand-written pattern matching.

What It Is

Sage is a runtime decision model used from application code. Instead of matching known patterns with regexes, the program passes a document and a question, such as "Contains personal information?", to Sage through a decide call. Sage returns a structured response containing an answer (yes, no, or null) and a probability.

How the Runtime Loop Works

The homepage example shows the flow: the application asks, Sage responds, and the program continues. A yes answer blocks a document, a no answer publishes it, and a null or uncertain answer routes it to a person for review. The homepage contrasts this with regex checks that miss names and addresses, for example in the sentence "Maya Chen lives at 14 Cedar Lane."

Where to Learn More

The site links to documentation, a model page, benchmarks, case studies and a thesis page. The homepage scene list references example domains such as refunds, security, trading, cattle, Mario and chess.