It makes a structured decision model usable through an existing chat-completion API by translating schema fields into typed decisions, while retaining Jev probabilities and confidence in the original response metadata.
The library adapts Jev System One behind its OpenAI chat-completion interface for JSON-schema structured output. It converts the JSON schema into typed questions, maps messages into Jev state, and folds typed answers back into a JSON assistant message; it rejects tools, streaming, multiple completions, and unsupported message content. This is SDK functionality rather than one end-user Jev workflow.
Which team should handle this?
Support ticket text: customer has tried connecting a Stripe account for three days, it keeps failing, they are losing sales, and ask for help ASAP.
Does the message convey urgency or time-sensitivity?
Support ticket text: customer has tried connecting a Stripe account for three days, it keeps failing, they are losing sales, and ask for help ASAP.
Use the ticket text as the System One state.
Submit a Choice for department and a Noul for urgency in one systemOne call.
Read the selected department and its probabilities, then interpret the urgency probability against the client threshold.
It makes a structured decision model usable through an existing chat-completion API by translating schema fields into typed decisions, while retaining Jev probabilities and confidence in the original response metadata.
