What it does
It does not trust the model's own confidence: input quality, the margin between options and entropy are checked separately, because Jev can be near-certain about junk input like a lone question mark.
Agent tooling
A drop-in routing layer that turns named routes into one Jev Choice, with a gate question for whether the request is even in domain.
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
It does not trust the model's own confidence: input quality, the margin between options and entropy are checked separately, because Jev can be near-certain about junk input like a lone question mark.
Start by listing the destinations you want incoming messages directed to, such as billing or product help. Add brief descriptions of what each team handles, plus an extra category for everything else. Your developer can configure these paths in code along with an access key that connects the app to TypeSafe.
The tool checks whether incoming text is clear enough before selecting a category. When a message is too short or the answer is uncertain, it pauses to let a person review it instead. Keep in mind that live routing sends incoming messages to the external TypeSafe service.