@juspay/neurolink/dist/voice/whistle/worker.d.ts

The pipe layer of an AI nervous system: one interface connecting provider neurons to your application, across three inference types — generate, stream and decide. `decide` returns typed, calibrated judgements from a non-generative model (~400ms, ~$0.00002

/**
 * Whistle worker thread: owns one instance of the Cactus "needle" WASM engine
 * with the Whistle model loaded, and runs its synchronous calls off the main
 * event loop. Messages are handled one at a time, in arrival order.
 *
 * `needle.js` is an Emscripten CommonJS build. It is evaluated through an
 * explicit CommonJS wrapper rather than `require()`d, because a model
 * directory that sits under a `"type": "module"` package would make Node load
 * it as ESM, where its `module.exports` assignment is lost.
 *
 * Self-contained — node builtins and type-only imports, nothing else — so the
 * thread starts quickly, never initialises the SDK's logger or config, and
 * also runs straight from the TypeScript source under Node's type stripping
 * (which does not map a sibling `./x.js` import onto `./x.ts`).
 *
 * @module voice/whistle/worker
 */
export {};