@jevable/core

The rule engine behind jevable: CEL expressions over a record (line, json) or a window (window) that can ask Jev semantic questions through judge.boolean, judge.choice and judge.score. Plain conditions decide first; Jev is asked only when they cannot, and the same question on the same material once.

import { Client, Engine, parseRecord, toCel } from "@jevable/core";

const engine = new Engine(new Client({ apiKey: process.env.TYPESAFE_API_KEY }));
const rule = engine.compile(`json.user != "bot" && judge.boolean(json.body, "Does this comment ask for a code change?") >= 0.7`);

const { line, json } = parseRecord(`{"user": "alice", "body": "can you rename this?"}`);
const { pass, calls, error } = await rule.match({ line, json: toCel(json) });

compile throws a RuleError for a rule that cannot work (unknown variable, unknown function, a judge call not compared with a threshold). @jevable/core/testing serves a fake Jev for tests.