jevracket — calibrated probabilities for Racket, now at full API parity

jevracket, our Racket client for the TypeSafe System One API, just landed full parity with the typesafe 0.7.5 API, a few weeks after its September debut.

Jev answers typed questions about a blob of state and returns calibrated probabilities. You don’t get prose back and you don’t parse anything: a yes/no question returns a number between 0 and 1, a multiple-choice question returns the winner plus the whole distribution, and a rating question returns a probability-weighted value plus a confidence. Several questions about the same state cost one request — jev reads the state once and evaluates every question against it.

(define r
  (ask c "Payouts have been failing for three days and I am losing sales."
       (hash "urgent" (noul "The message conveys time pressure.")
             "team"   (choice "Which team should handle this"
                              (hash "billing" "Payments, invoicing, refunds"
                                    "technical" "Bugs, outages, integrations")))))

(noul-answer-value   (response-ref r "urgent"))   ; 0.99
(choice-answer-value (response-ref r "team"))     ; "technical"

Built from the spec, not ported

The client is written against the published HTTP reference at docs.typesafe.ai/api, not translated from the vendor’s pre-1.0 Python SDK — the wire protocol is two endpoints and a handful of JSON shapes, and the SDK’s bulk is ergonomics that wouldn’t survive a move to Racket anyway. The tests pin the documented request and response examples verbatim, so a change on the vendor’s side fails the suite rather than surfacing in production.

No dependencies beyond base — net/http-client and json ship with Racket. A client for one HTTP endpoint should not drag a dependency tree behind it.

MIT licensed. Source on github.com/IoTone/jevracket.