1.3.2

First release.

Ask TypeSafe Jev questions about your house and get numbers back. Jev is a decision
model rather than a chat assistant: it answers a typed question with a probability, a
choice or a score, and this integration turns each answer into an entity you can
automate on.

Four actions. Three ask one question each and arrive with every field filled in:
a yes/no question returning a probability, a choice returning the winning option and
the whole distribution, and a rating against levels you describe. The fourth asks any
mix of them about the same text in a single request, which takes about as long as
asking one.

Point it at entities instead of writing a template. Every action takes the normal
Home Assistant target picker, so you select entities, devices, areas or labels and the
integration builds the state itself.

Tell it how to read the numbers. A background field carries standing facts about
what a reading means, attached to the question rather than to the state. Measured on
the same question, that took the separation between an idle washing machine and a
running one from 0.21 to 0.60.

It shows what it spends. Calls, input tokens and estimated cost are entities, and
a daily token budget stops evaluation when it trips, keeping the last answers and
saying why.

Needs Home Assistant 2026.9 or newer. Twelve worked examples ship in examples/, four
of them combining Jev with a language model through ai_task.generate_data.