Levanto Sage2

Levanto’s Sage2 gives developers two ways to act on incoming language. The extraction API turns text into typed fields and links the fields back to the input. Streaming decisions update a score as a conversation evolves. Both are useful when an application must collect details before choosing what action to take.

How extraction handles a message

The developer defines the fields the application needs. Sage2 can return text, numbers, money, dates, times and durations, with the text and probability behind each returned value. Levanto’s example uses a customer describing a vehicle and budget. The API returns a null loan term until the customer supplies one. Developers can use the source position to check which part of a message produced a value before sending it onward.

Where streaming decisions fit

A decision can track whether the conversation has reached a condition, such as an explicit request for a quote. That lets an application use a narrower check in a flow that would otherwise ask a general assistant to decide when to call the next tool. Levanto’s demonstration marks replay timing as illustrative. A team should test its own stream and define what a low-confidence or missing result does before connecting the result to a real action.

Access and pricing

Levanto lists a free plan with $1 of usage per month and paid plans with larger included usage. Teams can start with one extraction schema and one representative decision. Use examples with corrections and incomplete answers to check whether the reported evidence still points to the right text after the input changes. That test is useful before extending the API to additional fields or adding automation.