@avinash-baraiya/pragma-interpreter

createEngine turns natural-language instructions into validated TableQuery results: a conservative deterministic parser first, a pluggable model interpreter for the rest, then validation, ambiguity handling, normalization and explanation.

import { createEngine, createModelInterpreter } from '@avinash-baraiya/pragma-interpreter';

const engine = createEngine({
  schema,
  timezone: 'Asia/Kolkata',
  interpreter: createModelInterpreter({ provider }),
});
const result = await engine.interpret('active customers older than 25, newest first', {
  currentState,
});

interpret() never rejects for runtime problems. It resolves to ok, needs_clarification, unsupported or error.

Docs: architecture · models · ambiguity